System and method for orthopedic prosthesis design

Abstract

A computer-implemented method for automated orthopedic prosthesis design includes the steps of receiving at least one medical image of a patient's bone, generating a three-dimensional (3D) model of the bone from the at least one medical image, determining a defective region in the bone, generating a defect reconstruction model, generating a 3D prosthesis model corresponding to the defect reconstruction model, and generating a 3D prosthesis model corresponding to the 3D prosthesis model. Wherein the 3D prosthesis model is patient-specific.

Description

Technical Field

[0001] The present invention relates to a system and method for orthopedic prosthesis design. Background Art

[0002] Lower limb bone defects pose rather complex challenges in orthopedic medicine, especially when the bone defects affect critical weight-bearing structures of the patient (i.e., human), such as the femur, tibia, and fibula. These defects can be caused by various etiological factors, such as trauma, neoplastic diseases, congenital anomalies, open fractures, infections, and failed joint replacement surgeries. These bone defects may require complex treatment methods due to their impact on the patient's mobility and quality of life.

[0003] One treatment method is to use 3D printed prostheses to provide personalized solutions for patients with bone defects. However, this process involves several time-consuming steps, significantly affecting the treatment schedule. The first major time-consuming phase is medical image segmentation, where CT scans must be carefully processed to create an accurate three-dimensional representation of the patient's anatomy. This segmentation process can take from several hours to several days, depending on the complexity of the defect and the quality of the imaging data. This segmentation, i.e., 3D modeling process, requires skilled professionals to manually or semi-automatically separate the bone structure from the surrounding tissues and determine the precise boundaries of the defect.

[0004] After segmentation, the prosthesis design phase is another challenging phase. The design phase requires a significant time investment and the involvement of expert engineers and medical professionals to create such a prosthesis. The engineers and medical professionals must collaborate to create an orthopedic prosthesis design that not only perfectly fits the anatomical defect but also meets the biomechanical requirements for weight-bearing and functionality. The design phase typically takes from several days to several weeks because each iteration must be carefully reviewed and potentially modified to ensure that the final prosthesis provides the best functional outcome while minimizing the risk of complications. This meticulous attention to detail, although time-consuming, is crucial for ensuring the long-term success of orthopedic prostheses. Summary of the Invention

[0005] The present invention provides a system and / or method for orthopedic prosthesis design and construction that can create orthopedic prostheses faster than conventional methods.

[0006] The present invention relates to a system and method for orthopedic prosthesis design and construction. In particular, the present invention relates to a system and method for automating orthopedic prosthesis design and construction that provides patient-specific, i.e., customized, prostheses. The prosthesis can be an implant that can be implanted into a patient.

[0007] In one example, the present invention also relates to a system and method for generating a three-dimensional (3D) model of a patient's skeleton based on one or more two-dimensional (2D) medical images.

[0008] According to one aspect, the present invention provides a computer-implemented method for automated orthopedic prosthesis design, comprising:

[0009] · Receiving at least one medical image of a patient's bone,

[0010] · Generating a three-dimensional (3D) model of the bone from at least one medical image,

[0011] · Determining a defective region in the bone,

[0012] · Generating a defect reconstruction model,

[0013] · Generating a 3D prosthesis model corresponding to the defect reconstruction model, wherein the 3D prosthesis model is patient-specific.

[0014] The advantages of the method are that it provides an automated method for segmenting medical images of a patient's bone and automatically generating a customized prosthesis to correct the bone defect. The method provides a faster and more efficient medical image segmentation method than manual segmentation methods. The method also has the advantage that the 3D prosthesis model it generates is more suitable for the patient's anatomy than current manual prosthesis manufacturing methods.

[0015] The 3D prosthesis model corresponds to an orthopedic prosthesis that can be placed on the patient's bone to correct the bone defect.

[0016] In one example, the method includes an additional step of exporting the 3D prosthesis model for 3D printing an orthopedic prosthesis corresponding to the shape and configuration of the 3D prosthesis model.

[0017] In one example, the reconstruction model is a 3D model corresponding to the normal bone portion without bone defects.

[0018] In one example, the reconstruction model indicates a prosthesis that can be attached to the bone in that region. The reconstruction model can be similar in shape, size, and configuration to a prosthesis (i.e., implant) that can be attached to the patient's bone.

[0019] Optionally, the orthopedic prosthesis can be an implant that can be implanted into the patient's body. The implant, i.e., the orthopedic prosthesis, can be surgically attached to the patient's bone after the bone defect has been surgically removed. The implant, i.e., the prosthesis, can enable the surgically modified bone to heal and function normally once the prosthesis generated by the method described herein is surgically attached to the bone and implanted into the patient's body.

[0020] In one example, the reconstruction model can be displayed on a user interface of the 3D bone model to illustrate how the corrected bone should look.

[0021] In one example, the method includes receiving multiple medical images. In one example, each medical image may include a CT scan of a bone or an X-ray of a bone or other medical images.

[0022] In one example, the method includes the following additional steps:

[0023] · Perform volume segmentation on the medical images,

[0024] · Generate a 3D model of the bone based on the volume segmentation.

[0025] In one example, the volume segmentation step includes the following additional steps:

[0026] · Preprocess at least one received medical image by normalization and / or standardization,

[0027] · Extract context information from at least one medical image,

[0028] · Reconstruct spatial details by upsampling,

[0029] · Generate a 3D model of the bone based on the context information and the reconstructed spatial details.

[0030] The advantage of the method is that it provides an improved automated method for segmenting medical images of new patients. The automated segmentation method is faster than the currently used manual method.

[0031] In one example, the step of extracting context information includes the following steps:

[0032] · Perform 3D convolution on at least one medical image,

[0033] · Perform batch normalization on at least one image and / or the 3D convolution,

[0034] · Perform ReLU activation.

[0035] In one example, the method includes an additional step of determining a surgical path for removing a bone defect and displaying the determined surgical path on a user interface.

[0036] In one example, the step of determining the surgical path includes the following additional steps:

[0037] · Determine an incision for removing the bone defect, where determining the incision includes determining the size and shape of the incision for removing the bone defect,

[0038] · Display the determined incision on the user interface,

[0039] · Determine the position of the reconstruction model to correct the bone defect,

[0040] · Determine a fixation arrangement for fixing the prosthesis to the bone to correct the bone defect.

[0041] In one example, the 3D prosthesis model is generated based on the determined incision, reconstruction model, and fixation configuration.

[0042] In one example, the 3D prosthesis model is customized in shape and configuration to match the patient's bone, such that the 3D prosthesis model is patient-specific.

[0043] In one example, the method steps are performed by a 3D U-net model including an encoder path, a decoder path, and skip connections connecting the corresponding encoder and decoder layers, wherein the encoder path is adapted to capture context information and the decoder path is adapted to reconstruct spatial information.

[0044] In one example, the method includes the step of 3D printing a prosthesis based on the derived 3D prosthesis model. The 3D prosthesis model can be exported in a suitable format for processing and printing by a 3D printer. The 3D printed prosthesis can be printed from a biocompatible material.

[0045] The advantage of the method is that it can produce a customized prosthesis (i.e., an orthopedic implant) that is better suited to the patient's anatomy, thereby reducing the surgical time, improving the prosthesis fit, and enhancing the postoperative outcome.

[0046] According to a second aspect, the present invention provides a machine learning model for automated orthopedic prosthesis design and construction, particularly for one or more of the above methods, the machine learning model comprising:

[0047] · An encoder path configured to capture context information,

[0048] · A decoder path adapted to reconstruct spatial information,

[0049] · Skip connections connecting the corresponding encoder and decoder layers.

[0050] The advantage of the machine learning model is that it can automatically process medical images and generate a customized prosthesis model that can be printed. This provides a time saving in image segmentation, provides a more accurate prosthesis model, and can improve cost-effectiveness and speed up the creation of the prosthesis (i.e., the implant).

[0051] In one example, the encoder path of the machine learning model includes multiple layers, the decoder path includes multiple layers, and the machine learning model is a 3D U-net model.

[0052] In one example, the encoder path of the machine learning model includes multiple 3D convolutional layers, each followed by or including batch normalization and ReLU activation. Optionally, the encoder path can include average pooling after each convolutional layer.

[0053] In one example, the encoder path is a contracting path and the decoder path is an expanding path. In one example, skip connections connect corresponding encoder and decoder paths. In one example, skip connections facilitate the reverse flow of data.

[0054] In one example, the input data may include 3D CT scans. The 3D CT scans may be preprocessed by normalization and standardization before being input into the 3D U-net model.

[0055] In one example, the encoder path is configured to extract context information using 3D convolutions, batch normalization, and ReLU activation.

[0056] According to a third aspect, the present invention provides a data processing device for automated orthopedic prosthesis design, comprising means for performing one or more of the methods described above.

[0057] In one example, the data processing device may be configured to execute a machine learning model described in one or more of the methods above.

[0058] In one example, the data processing device may be configured to execute a 3D U-net model that processes input medical images and is capable of accurately segmenting new patient medical images to optimize orthopedic prosthesis design.

[0059] According to a fourth aspect, the present invention provides a computer program comprising instructions that, when executed by a computer, cause the computer to perform one of the methods described above.

[0060] According to a fifth aspect, the present invention provides a computer-readable medium comprising instructions that, when executed by a computer, cause the computer to perform one or more of the methods described above.

[0061] According to a sixth aspect, the present invention provides a computer-implemented method for training a machine learning model for automated orthopedic prosthesis design, particularly the machine learning model described herein, comprising:

[0062] · Receiving an input training dataset comprising a plurality of medical images and ground truth segmentations, the plurality of medical images comprising images of bones with bone defects;

[0063] · Passing the input training dataset through the convolutional layers of a 3D U-net model,

[0064] · Optimizing the model parameters using a combination of Dice coefficient loss and binary cross-entropy,

[0065] · Enhance the generalization ability by applying one or more mutation techniques,

[0066] · Monitor the model performance based on the validation dataset to ensure convergence and reduce overfitting.

[0067] In one example, during training, the encoder path uses 3D convolution, batch normalization, and ReLU activation. During training, the decoder path performs an upsampling operation and utilizes skip connections for detail preservation.

[0068] In one example, the training dataset for the method of training a machine learning model includes:

[0069] · Input data, where the input data includes

[0070] · Input data, where the input data includes multiple images or models with one or more bone defects,

[0071] · Output data, where the output data includes multiple images or models of normal bones.

[0072] According to the seventh aspect, the present invention provides a system for automated orthopedic prosthesis design,

[0073] including:

[0074] · A computing device including a processor and a storage unit, the processor and the storage unit being operatively coupled,

[0075] · The storage unit is adapted to store instructions that, when executed by the processor, cause the computing device to perform the method for automated orthopedic prosthesis design described herein.

[0076] According to the eighth aspect, the present invention provides a system for automated orthopedic prosthesis design, including:

[0077] · A segmentation module configured to:

[0078] o Receive at least one medical image of a patient's bone,

[0079] o Generate a three-dimensional (3D) model of the bone from at least one medical image,

[0080] o Determine the defective regions in the bone,

[0081] · An orthopedic prosthesis design module configured to:

[0082] o Generate a defect reconstruction model,

[0083] o Generate a 3D prosthesis model corresponding to the defect reconstruction model, where the 3D prosthesis model is patient-specific.

[0084] In one example, the orthopedic prosthesis design module is configured to export a 3D prosthesis model for 3D printing a prosthesis corresponding to the shape and configuration of the 3D prosthesis model.

[0085] In one example, the reconstruction model is a 3D model corresponding to a normal bone portion without bone defects.

[0086] In one example, the reconstruction model indicates a prosthesis that can be attached to the bone in the region. The reconstruction model can be similar in shape, size, and configuration to a prosthesis (i.e., an implant) that can be attached to the patient's bone.

[0087] The orthopedic prosthesis can be an implant that can be implanted into a patient. The implant, i.e., the orthopedic prosthesis, can be surgically attached to the patient's bone after the bone defect has been surgically removed. The implant, i.e., the prosthesis, can enable the surgically modified bone to heal and function normally once the prosthesis generated by the method described herein is surgically attached to the bone and implanted into the patient.

[0088] In one example, the system includes a user interface, such as a display screen. In one example, the reconstruction model can be presented on the user interface of the 3D bone model to illustrate how the corrected bone should look.

[0089] In one example, the segmentation module is configured to:

[0090] · Perform volumetric segmentation on medical images,

[0091] · Generate a three-dimensional (3D) model of the bone based on the volumetric segmentation.

[0092] In one example, the system includes:

[0093] · A medical image preprocessing module configured to:

[0094] o Preprocess at least one received medical image by normalization and / or standardization,

[0095] · A segmentation module configured to:

[0096] o Extract context information from at least one medical image,

[0097] o Reconstruct spatial details by upsampling,

[0098] o Generate a three-dimensional (3D) model of the bone based on the context information and the reconstructed spatial details.

[0099] In one example, the system includes a surgical planning module configured to: determine a surgical path for removing the bone defect and present the determined surgical path on the user interface.

[0100] In one example, the surgical planning module is configured to:

[0101] · Determine the size and shape of the incision for removing the bone defect,

[0102] · Display the determined incision on the user interface,

[0103] · Determine the position of the reconstruction model to correct the bone defect,

[0104] · Determine the fixation configuration for fixing the prosthesis to the bone to correct the bone defect.

[0105] In one example, the orthopedic prosthesis design module is configured to generate a 3D prosthesis model based on the determined incision, reconstruction model, and fixation configuration.

[0106] In one example, the 3D prosthesis model is customized in shape and configuration to match the patient's bone, such that the 3D prosthesis model is patient-specific.

[0107] In one example, the system includes a machine learning model, and the machine learning model includes:

[0108] · An encoder path configured to capture context information,

[0109] · A decoder path adapted to reconstruct spatial information,

[0110] · Skip connections connecting the corresponding encoder and decoder layers,

[0111] · Wherein the segmentation module and the orthopedic prosthesis design module are part of the machine learning module.

[0112] In one example, the encoder path includes multiple layers, the decoder path includes multiple layers, and the machine learning model is a 3D U-net model.

[0113] In one example, the 3D U-net model includes an encoder path, a decoder path, and skip connections connecting the corresponding encoder and decoder layers, wherein the encoder path is adapted to capture context information and the decoder path is adapted to reconstruct spatial information.

[0114] In one example, the encoder path includes multiple 3D convolutional layers, each followed by or including batch normalization and ReLU activation. Optionally, the encoder path may include average pooling after each convolutional layer.

[0115] In one example, the system includes a 3D printer. The 3D printer is configured to receive the exported 3D prosthesis model and print an orthopedic prosthesis corresponding to the 3D prosthesis model.

[0116] According to a ninth aspect, the present invention provides an artificial intelligence (AI)-assisted orthopedic prosthesis design system. The system can also be used for surgical planning. The system includes an AI-assisted module that focuses on patient-specific orthopedic prosthesis (i.e., prosthesis) design. The module uses AI algorithms and utilizes a comprehensive range of complex data including high-resolution medical imaging, biomechanical dynamics, and detailed patient history. By analyzing this information, the system generates an accurate three-dimensional model that conforms to the patient's anatomical profile, facilitating the creation of an orthopedic prosthesis (i.e., prosthesis) that seamlessly restores optimal function.

[0117] According to one aspect, the present invention relates to an artificial intelligence-assisted orthopedic prosthesis design and surgical planning system. In one example, the system provides a medical image segmentation method and a bone model defect simulation method.

[0118] In one example, the system may include a machine learning algorithm for bone defect repair prediction and a machine learning algorithm for the design of surgical implants with fixation plates.

[0119] In one example, the system is a computer-implemented system for real-time monitoring and adjustment of orthopedic prosthesis function.

[0120] In one example, a personalized surgical planning method based on patient-specific anatomical and functional data is provided.

[0121] In one example, a method for optimizing prosthesis material selection based on patient-specific biomechanical characteristics is provided.

[0122] In one example, the system is suitable for implementing a machine learning algorithm for optimizing surgical incision and implant placement positions in orthopedic surgery.

[0123] The term "comprising" (and its grammatical variants) is used herein in an inclusive sense of "having" or "including", rather than in the sense of "consisting only of". The term "medical image" herein refers to a two-dimensional image of a patient's bone or a patient's anatomical part. The medical image can be a CT scan or an X-ray or other suitable images. Optionally, each medical image may include multiple two-dimensional views of a patient's bone or a patient's anatomical part. The term "prosthesis" or "orthopedic prosthesis" defines an orthopedic device that is connected or implanted on or in a bone for correcting a bone defect or restoring the bone to its normal shape after the bone defect has been surgically removed. These terms can be used interchangeably to define this element. BRIEF DESCRIPTION OF THE DRAWINGS

[0124] Embodiments of the present invention will now be described by way of example and with reference to the accompanying drawings, in which:

[0125] Figure 1 An embodiment of the system for orthopedic prosthesis design is shown.

[0126] Figure 2 Shows a schematic diagram of a computing device implementing an orthopedic prosthesis design system.

[0127] Figure 3 Shows another embodiment of a system for orthopedic prosthesis design.

[0128] Figure 4 Shows an example of a 3D U-net model for automated orthopedic prosthesis design.

[0129] Figure 5 Shows an embodiment of an automated method for orthopedic prosthesis design.

[0130] Figure 6 Shows another embodiment of an automated method for orthopedic prosthesis design.

[0131] Figure 7 Shows sub-method steps of an image preprocessing step.

[0132] Figure 8 Shows sub-method steps of an extracting context information step.

[0133] Figure 9 Shows sub-method steps of a determining surgical path step.

[0134] Figure 10 Shows a training method of a machine learning model for automated orthopedic prosthesis design.

[0135] Figure 11 Shows for training Figure 4 An example of training data for the 3D U-net model shown.

[0136] Figure 12 Shows a comparison of the output of the 3D U-net model with the ground truth and the input image.

[0137] Figure 13 Shows a comparison of the 3D bone model generated by the 3D U-net model with the ground truth bone model.

[0138] Figure 14 Shows a table of quantitative results of training the 3D U-net model.

[0139] Figures 15 to 21 Shows the operations of a system and method for automated prosthesis design. Detailed Description

[0140] Reference Figure 1, which shows an embodiment of the present invention. This embodiment is configured to provide a system 100 for automated orthopedic prosthesis design, including: a segmentation module 102 configured to receive at least one medical image of a patient's bone, generate a three-dimensional (3D) model of the bone from the at least one medical image, and determine the defective area in the bone; and an orthopedic prosthesis design module 104 configured to generate a defect reconstruction model and generate a 3D prosthesis model corresponding to the defect reconstruction model, where the 3D prosthesis model is patient-specific. The prosthesis model (i.e., the implant model) corresponds to the structure or component for correcting the bone defect.

[0141] The advantage of this system is that it provides an automated system that can segment medical images of a patient's bone and automatically generate a customized prosthesis to correct the bone defect or provide a correction component that attaches to the bone after the bone defect is surgically removed.

[0142] In one example, the orthopedic prosthesis design module 104 is configured to export the 3D prosthesis model to 3D print a prosthesis corresponding to the shape and configuration of the 3D prosthesis model. The system may include a 3D printer 110 operatively coupled to the orthopedic prosthesis design module via a wired connection or a wireless connection. The 3D printer 110 may be configured to receive the 3D prosthesis model. The 3D prosthesis model may be exported in an appropriate file format for use in 3D printing an orthopedic prosthesis corresponding to the 3D prosthesis model. The 3D printer 110 may be configured to print the orthopedic prosthesis.

[0143] The 3D printer 110 may be configured to print the prosthesis using an appropriate biocompatible material (such as titanium or ceramic or other appropriate biocompatible materials). The 3D printed orthopedic prosthesis (i.e., the orthopedic implant) may be coupled to the bone to correct the determined bone defect. The prosthesis, i.e., the orthopedic implant, may be surgically installed in the patient's bone after the bone defect is removed.

[0144] The segmentation module 102 and the orthopedic prosthesis design module 104 may be operatively coupled. The segmentation module 102 is configured to receive at least one medical image, but preferably receives multiple medical images 10 as input. The segmentation module 102 may receive multiple medical images of the patient's bone. The segmentation module 102 may receive multiple medical images of the lower limb bones (such as the femur, tibia, or fibula or other lower limb bones). In one example, the medical image may be a CT scan image.

[0145] The segmentation module 102 is configured to process the received medical images to segment the medical images and generate a 3D model of the skeleton. In particular, the segmentation module 102 is configured to perform volumetric segmentation on the medical images. The segmentation module 102 is configured to generate a 3D model of the skeleton based on the volumetric segmentation. The 3D skeleton model is reconstructed from the segmentation process of the medical images. The medical images may include two-dimensional images from different perspectives, such as front view, back view, side view, etc. The 3D model is generated from the two-dimensional views, indicating the lower limb skeleton visible in the medical images.

[0146] The segmentation module 102 is also configured to identify bone defects in the skeleton. The bone defects can be identified in the region of interest of the skeleton, and the bone defects can be highlighted. The segmentation model 102 is adapted to output a 3D skeleton model showing the bone defects.

[0147] The system 100 may include a user interface 112. The user interface 112 may be a screen or a display, such as an LCD screen or an LED screen. The 3D skeleton model may be displayed on the user interface after automatically segmenting multiple medical images. In addition, the 3D prosthesis model (i.e., implant model) and the reconstruction model may be displayed after each generation. In one example, the defect reconstruction model is a 3D model corresponding to the normal bone part without bone defects.

[0148] As Figure 1 shown, the system 100 includes a medical image preprocessing module 106. The medical image preprocessing module 106 is configured to receive multiple medical images 10. The preprocessing module 106 is also configured to preprocess the received medical images (i.e., at least one medical image) by normalization and / or standardization. The preprocessing module 106 is adapted to process the medical images before the segmentation module processes the images.

[0149] The segmentation module 102 is also configured to extract context information from multiple medical images and reconstruct spatial information. The spatial details can be reconstructed by upsampling. The segmentation module 102 is also configured to generate a 3D model of the skeleton (i.e., 3D skeleton model) based on the context information and the reconstructed spatial details. The 3D skeleton model is an accurate 3D representation of the skeleton in the medical images.

[0150] The system 100 involves preprocessing input data (such as medical images), applying data augmentation techniques and outputting a 3D model of the skeleton. The system 100 provides a powerful tool for accurately segmenting the patient's anatomy from medical images, which is beneficial to medical practitioners. The system 100 also has the advantage that it provides a tool that can segment medical images faster and more efficiently compared to current manual practices.

[0151] In one example, the system includes a surgical planning module configured to: determine a surgical path for removing the bone defect and display the determined surgical path on the user interface.

[0152] As Figure 1 shown, an example form of system 100 also includes a surgical planning module 108. The surgical planning module 108 is configured to identify the location, size, and shape of an incision for removing a determined bone defect. The determined incision, i.e., the location, size, and shape of the incision, can be presented on the user interface 112. The surgical planning module 108 is configured to determine the location of the reconstruction model to correct the bone defect, i.e., to determine how and where to place the reconstruction model to repair the bone defect, e.g., after the bone defect has been resected.

[0153] The surgical planning module 108 is further configured to determine a fixation configuration for fixing an orthopedic prosthesis to the bone to correct the bone defect and enable normal function of the bone. The fixation configuration can include screws or bolts or adhesives or any other suitable means for fixing the prosthesis to the bone. The surgical planning module 108 can also determine an appropriate surgical path for removing the bone defect and placing the reconstruction model and the prosthesis in the gap or void left after removing the bone defect.

[0154] In one example, the location, size, and shape of the determined incision can be displayed on the user interface 112. Additionally, the location of the reconstruction model and the fixation configuration can also be displayed after being determined by the surgical planning module 108.

[0155] In another example, the bone defect can be repaired by attaching a prosthesis to the bone. In this alternative form, the surgical planning module 108 can also calculate the optimal location and manner for fixing the prosthesis to the bone to correct the bone defect.

[0156] The orthopedic prosthesis design module 104 can be configured to generate a 3D prosthesis model based on the determined incision, reconstruction model, and fixation configuration. The 3D prosthesis model is customized in shape and configuration to match the patient's bone, such that the 3D prosthesis model is patient-specific.

[0157] In this example embodiment, the modules can be implemented by a computing device 200 (i.e., a computer) having a suitable user interface and suitable other components. The computing device 200 can be implemented by any computing architecture, including a portable computer, a tablet computer, a stand-alone personal computer (PC), a smart device, an Internet of Things (IoT) device, an edge computing device, a client / server architecture, a "dumb" terminal / host architecture, a cloud computing-based architecture, or any other suitable architecture. The computing device can be appropriately programmed to implement the present invention.

[0158] In one example form, the modules described herein can be hardware modules such as operatively coupled ICs or microprocessors. Alternatively, the modules can be software modules that can be implemented on a computing device 200 as described herein. The software modules can be in accordance with Figure 1The architecture arrangement shown and can be operatively coupled. In another alternative form, the module can be a combination of hardware and software elements.

[0159] Figure 2 FIG. shows a schematic diagram of a computing device 200 (i.e., a computer or a computing server) arranged to implement an example embodiment of a system for automated orthopedic prosthesis design. In this embodiment, the computing device 200 includes appropriate components required to receive, store, and execute appropriate computer instructions. The components may include a processor 202 (i.e., a processing unit) including a central processing unit (CPU), a math coprocessor unit (math processor), a graphics processing unit (GPU), or a tensor processing unit (TPU) for tensor or multi-dimensional array calculations or operations, a random access memory (RAM) 206, and input / output devices such as a disk drive 208, an input device 210 such as an Ethernet port, a USB port, etc. A display 212 such as an LCD or LED display, or any other appropriate display may be part of the computing device 200. Optionally, the display 212 may be separate or independent from the computing device 200. The display 212 may be a user interface 112 or may display the user interface 112.

[0160] The computing device 200 may include instructions stored in the ROM 204, RAM 206, or disk drive 208 and executable by the processing unit 202 (i.e., the processor). One or more communication links 214 may be provided, which may be connected to one or more computing devices in various ways, such as servers, personal computers, terminals, wireless or handheld computing devices, Internet of Things (IoT) devices, smart devices, edge computing devices. At least one communication link may be connected to an external computing network via a telephone line or other type of communication link.

[0161] The computing device 200 may include a storage device such as a disk drive 208, which may include a solid-state drive, a hard disk drive, an optical drive, a tape drive, or a remote or cloud-based storage device. The computing device 200 may use a single disk drive or multiple disk drives, or a remote storage service. The computing device 200 may also have an appropriate operating system that resides in the disk drive or the ROM of the computing device 200.

[0162] The computing device may also include one or more databases 220 suitable for storing one or more data segments. The database may store bone model data. One or more databases may store the received input images and may also store the 3D bone models generated by the system 100.

[0163] A computer or computing device 200 may also provide the necessary computing power to operate or interface with a machine learning network, such as a neural network, to provide various functions and outputs. The neural network may be implemented locally or may also be accessed in part or in whole via a server or cloud-based service. The machine learning network may also be untrained, partially trained, or fully trained and / or may also be retrained, adapted, or updated over time. The computing device may include one or more GPUs operatively coupled to a CPU (i.e., a processor). The computing device may include additional hardware elements operatively coupled to the CPU and / or GPU to provide the computing device components required to implement the machine learning network or machine learning model. The learning network or model may be stored in a storage unit, such as a ROM.

[0164] Figure 3 An alternative embodiment of a system 300 for orthopedic prosthesis design is shown. The system 300 includes a computing device 302. The computing device includes a processor and a storage unit, the processor and the storage unit being operatively coupled. The storage unit is adapted to store instructions that, when executed by the processor, cause the computing device to: receive at least one medical image of a patient's bone, generate a three-dimensional (3D) model of the bone from the at least one medical image, determine a defective region in the bone, generate a defect reconstruction model, and generate a 3D prosthesis model corresponding to the defect reconstruction model, wherein the 3D prosthesis model is patient-specific.

[0165] The computing device 302 may be the same as the computing device 200. The computing device 302 is configured to store and execute a machine learning model or an AI model or a machine learning network that is configured to: receive at least one medical image of a patient's bone, generate a three-dimensional (3D) model of the bone from the at least one medical image, determine a defective region in the bone, generate a defect reconstruction model, and generate a 3D prosthesis model corresponding to the defect reconstruction model, wherein the 3D prosthesis model is patient-specific. The functions of the foregoing modules may be performed by a machine learning model 400. The machine learning model 400 is trained to automatically segment medical images, generate 3D bone models, identify bone defects, and generate an orthopedic prosthesis model that can be implanted to repair the bone defect. The machine learning model 400 may be executed by a processing unit 202 and may be stored in a storage unit 304.

[0166] The system 300 may also include a 3D printer 310 operatively coupled to the computing device 302. The computing device 302 may transmit the orthopedic prosthesis model generated by the machine learning model 400 to the 3D printer 310. The 3D printer 310 may be adapted to 3D print an orthopedic prosthesis (i.e., an orthopedic implant) based on the exported prosthesis model generated by the machine learning network 400 (i.e., the AI model 400).

[0167] System 300 may also include a user interface 312 on which a 3D bone model and a prosthesis model can be displayed. The user interface 312 may be a display similar to the display 212.

[0168] Figure 4 An example embodiment of a machine learning model 400 (or machine learning network or AI model) stored in the storage unit of the computing device 302 or 200 and executed by the processing unit of the computing device is shown. Referring Figure 4 , the machine learning model 400 includes an encoder path 402 configured to capture context information, a decoder path 404 suitable for reconstructing spatial information, and skip connections connecting the corresponding encoder and decoder layers.

[0169] The advantage of the machine learning model is that it can automatically process medical images and generate a customized prosthesis model that can be printed. This provides a time saving in image segmentation, provides a more accurate prosthesis model, and can improve cost effectiveness and speed up the creation of a prosthesis (i.e., an implant).

[0170] In one example, the machine learning model 400 may be a 3D U-net model. The 3D U-net model is a "U" shaped model. The encoder path 402 is preferably a contracting path and the decoder path 404 is preferably an expanding path. The encoder path 402 includes multiple layers, such as convolution layers. The decoder path 404 includes multiple layers, such as convolution layers. The convolution layers may have different sizes, i.e., different dimensions.

[0171] Each layer in the encoder path 402 may be followed by or include batch normalization and ReLU activation. Optionally, the encoder path 402 may include average pooling after each convolution layer. As Figure 4 shown, the machine learning model 400 may include seven convolution layers in the encoder path 402 and two convolution layers in the decoder path 404. The skip connections connect the corresponding encoder and decoder paths. In one example, the skip connections facilitate the reverse flow of data. In one example, the input data may include 3D CT scans. The 3D CT scans may be preprocessed by normalization and standardization before being input into the 3D U-net model.

[0172] Figure 5Shows an exemplary embodiment of a method 500 for automated orthopedic prosthesis design. The method includes a plurality of steps. Step 502 includes receiving at least one medical image of a patient's bone. Preferably, multiple medical images are received. Step 504 includes generating a three-dimensional (3D) model of the bone from at least one medical image. Step 506 includes determining defective regions in the bone. Step 508 includes generating a defect reconstruction model. Step 510 includes generating a 3D prosthesis model corresponding to the defect reconstruction model, wherein the 3D prosthesis model is patient specific, i.e., customized, to fit the patient and their bone defect.

[0173] Method 500 may include an additional step 512 of exporting the 3D prosthesis model for 3D printing an orthopedic prosthesis corresponding to the shape and configuration of the 3D prosthesis model. Optionally, method 500 may include the step of 3D printing (i.e., printing using a 3D printing method) an orthopedic prosthesis based on the exported 3D model.

[0174] Figure 6 Shows an exemplary embodiment of a method for automated orthopedic prosthesis design. The method starts at step 602. Step 602 includes receiving multiple medical images of a patient's bone with a bone defect. The bone may be a lower limb bone, such as a weight-bearing bone. Step 604 includes performing volume segmentation on the received medical images. Step 606 includes generating a 3D model of the bone based on context information and reconstructed spatial details. The 3D model of the bone may be generated based on volumetric segmentation.

[0175] Once the 3D model of the bone is generated, step 608 is executed. Step 608 includes identifying a bone defect in the 3D bone model. Step 610 includes determining a surgical path for removing the bone defect. The surgical path may be displayed or presented on the user interface 112 or the display 212 of the computing device 200.

[0176] Step 612 follows step 610. Step 612 includes generating a defect reconstruction model. The reconstruction model is a 3D model of the shape that needs to be attached to the bone to restore the bone to its normal shape after the defect is removed. The reconstruction model may be presented on the user interface 112 or the display 212. Step 614 includes generating a 3D prosthesis model (i.e., an implant model). The 3D prosthesis model is generated based on the determined incision, reconstruction model, and fixation arrangement. The 3D prosthesis model (i.e., the implant model) may be patient specific, i.e., customized, to fit the patient and address the specific bone defect the patient has. Step 616 includes exporting the 3D prosthesis model.

[0177] The volume segmentation step in step 604 includes additional steps.Figure 7 Shows the sub - method steps of the image pre - processing step. Step 620 includes pre - processing the received medical image by normalization and / or standardization. In one example, a batch normalization operation can be performed on the received image. Step 622 includes extracting contextual information from the medical image. Step 624 includes reconstructing spatial details by up - sampling.

[0178] In one example, the step of extracting contextual information in step 622 includes Figure 8 The additional steps shown. Step 630 includes performing 3D convolution on the received medical image. Step 632 includes performing batch normalization on the received image and / or the 3D convolution. Step 634 includes performing ReLU activation after each convolution process.

[0179] The step of determining the surgical path in step 610 includes Figure 9 The additional sub - steps shown. Referring to Figure 9 , step 640 includes identifying an incision (or other surgical operation) to remove the bone defect. Identifying the incision includes determining the size and shape of the incision for removing the bone defect. Step 642 includes presenting the determined incision on the user interface 112, such as the display 212, to illustrate the proposed incision to medical professionals. Step 644 includes determining the position of the reconstruction model to correct the bone defect. Step 646 includes determining the fixation configuration to fix the prosthesis (i.e., implant) to the bone to correct the bone defect.

[0180] The advantage of this method is that it provides an improved automated method for segmenting medical images of new patients. The automated segmentation method is faster than the currently used manual method. The advantage of this method is that it can produce customized prostheses (i.e., orthopedic implants) that are more suitable for the patient's anatomy, thereby reducing the surgical time, improving the prosthesis fit, and enhancing the postoperative outcome.

[0181] Method 600 can optionally include the step of 3D printing an orthopedic prosthesis based on the prosthesis model. The prosthesis can be printed from a suitable biocompatible material. Methods 500 and 600 are particularly suitable for generating orthopedic prosthesis designs (i.e., implant designs) applicable to the patient's lower limb. Methods 500 and 600 can be executed by the computing device 200. In one example, the steps of method 500 or method 600 are executed by a machine learning model on the computing device 200. For example, methods 500 and 600 can be executed by a 3D U - net model (such as Figure 4 the model 400 shown) stored on and executed by the computing device 200.

[0182] In one example form, computing device 200 may include a computer program that includes instructions that, when executed by the computer, cause the computer to perform the steps of method 500 or method 600, as described.

[0183] In one example form, a system for orthopedic prosthesis design includes a computer-readable medium that includes instructions that, when executed by a computing device (such as device 200), cause the computing device to perform the steps of method 500 or method 600.

[0184] Figure 10 An example of a computer-implemented method 700 for training a machine learning model 400 (such as a 3D U-net model for segmenting medical images and generating orthopedic prosthesis models) is shown. The training method begins at step 702. Step 702 includes receiving a training data set that includes a plurality of medical images of bones having bone defects and ground truth segmentations. The training data set may include input data and output data. The input data includes a plurality of images or models having one or more bone defects. The output data includes a plurality of images or models of normal bones. The input data and the output data may be input through the U-net model 400 to train the model.

[0185] Step 704 includes passing the training data set through the convolutional layers of the 3D U-net model. Step 706 includes optimizing the model parameters using a combination of Dice coefficient loss and binary cross-entropy. Step 708 includes enhancing the generalization ability by applying one or more augmentation techniques. Step 710 includes monitoring the model performance based on a validation data set to ensure convergence and reduce overfitting.

[0186] During training, the encoder path uses 3D convolution, batch normalization, and ReLU activation. During training, the decoder path performs upsampling operations and utilizes skip connections for detail retention. The network parameters are optimized using a combination of Dice coefficient loss and binary cross-entropy. Additionally, the training data is augmented by applying techniques such as random rotation, elastic deformation, and intensity variation to enhance the generalization ability of the training data and the model. The 3D U-net model 400 can be trained using supervised learning.

[0187] Figure 11 An example of training data and how it is prepared is shown. The training data can be prepared by using 3D bone models generated manually or by other appropriate operations. Some of the 3D bone models have parts removed manually to simulate bone defects. The original unmodified 3D bone models are grouped as output data, and the models with simulated bone defects are grouped as input data for training the 3D U-net model.

[0188] The training data batch size is selected according to the GPU memory, and the batch contains multiple 3D models. The performance of the model 400 is monitored on the validation dataset to ensure convergence and avoid overfitting. Early stopping mechanism based on validation metrics is implemented for model training. The training of the 3D U-net model 400 can be stopped based on the validation metrics. The trained model 400 can accurately segment new input images for 3D reconstruction of the bone and defect regions.

[0189] Figure 12 An example of the comparison between the model output and the ground truth is shown. Refer to Figure 12 , the input image 800 generates the predicted segmentation result 802, which is compared with the ground truth 804. The predicted data 802 is very close to the ground truth 804. This indicates that the U-net model trained according to the present invention is well-trained. Figure 13 A further output comparison between the 3D bone model generated by the U-net model 400 and the ground truth is shown. Figure 13 The 3D model on the left side of each pane in Figure 13 shows that the 3D U-net model 400 described herein provides the ability to accurately segment and 3D reconstruct bones from two-dimensional input medical images. Figure 14 The quantitative results of training the U-net model 400 according to the present invention are shown. The results show the Dice Similarity Coefficient (DSC), Intersection over Union (IoU), and Hausdorff Distance (HD) of the performance of the model 400.

[0190] Figures 15 to 21 The operations of the system and method for automated prosthesis design are shown. The operations shown can be performed by the aforementioned U-net model 400. Figure 15 The import of medical images 902, 904, 906 is shown. The bone model 908 is generated by the 3D U-net from medical images (such as CT scans). Refer to Figure 16 , the region of interest 910 with bone defects is identified. The U-net model 400 can create an incision 912 and display the incision on the user interface (such as interface 112), as Figure 17 shown. The incision 912 indicates the surgical operation to remove the bone defect. As Figure 18 shown, the U-net model 400 automatically generates a reconstruction model 914, that is, the U-net model 400 performs automatic defect reconstruction. The reconstruction model 914 is shown to fit the incision 912 to restore the normal shape of the bone. The U-net model determines the fixation structure 916 for installing the prosthesis (i.e., implant), as Figure 19 shown. As Figure 19As shown, the fixing structure is a screw 916, and the U-net model 400 generates four screw positions. The U-net model 400 is suitable for automatically generating a prosthesis model 918 (i.e., an implant model), such as Figure 20 shown. The prosthesis can be formed to fit the incision. The implant model 918 can be exported, for example, as a 3D CAD file, such as Figure 21 shown.

[0191] The system and method for automated orthopedic prosthesis design provide improvements in surgical outcomes and prosthesis design. The U-net model can continuously learn and improve the design based on learning from outcome data, thereby improving efficiency, cost-effectiveness, and personalized orthopedic care. The system and method combine deep learning and manufacturing techniques to create customized prostheses (i.e., implants). The system and method further reduce surgical time, enhance the fit of the prosthesis, and postoperative outcomes. The AI model used can continuously learn to improve accuracy and cost-effectiveness.

[0192] Another example of a system for automated orthopedic prosthesis design will be described. In this example, the system integrates artificial intelligence (AI) to provide highly personalized orthopedic prosthesis design and surgical planning solutions, especially for the unique anatomical needs of each patient.

[0193] In this example, the system can include an AI-assisted module focused on patient-specific orthopedic prosthesis / implant design. This module uses AI algorithms and utilizes a comprehensive range of complex data including high-resolution medical imaging, biomechanical dynamics, and detailed patient history. By analyzing this information, the system generates an accurate three-dimensional model that conforms to the patient's anatomical profile, facilitating the creation of an orthopedic prosthesis (i.e., implant) that seamlessly restores optimal function.

[0194] The system can also include an AI-assisted medical image segmentation module that employs advanced image processing techniques. This module is capable of accurately segmenting medical images into different anatomical regions, thereby facilitating the creation of a highly accurate three-dimensional model. By improving the accuracy and efficiency of the overall prosthesis design process, this AI-driven segmentation module significantly enhances the effectiveness of the system.

[0195] The system also integrates a preoperative surgical planning software module. This module uses predictive models to predict the surgical path, enabling surgeons to simulate and plan surgical procedures with extremely high precision. By conducting comprehensive preoperative simulations, potential challenges can be identified, and the most effective surgical strategy can be developed based on the patient's individual prosthesis design and physiological characteristics. This AI-assisted surgical planning module enables surgeons to make informed decisions, thereby improving surgical outcomes, enhancing patient satisfaction, and accelerating recovery.

[0196] The system can include an AI self - learning mechanism. Through iterative learning, the system continuously improves its prediction accuracy and design precision by absorbing postoperative results. This self - improvement process further enhances the system's effectiveness and reliability, strengthening its position in orthopedic care.

[0197] The system enhances the functionality of the implant, aiming to significantly reduce the probability of implant - related complications.

[0198] In this example, the system can include a dynamic, interactive planning module capable of adapting to the variable nature of surgical procedures. This module enhances the surgeon's preparation, resulting in a more efficient and accurate surgical process.

[0199] The system combines patient - specific anatomical data with advanced predictive analytics, thereby reducing the risks associated with malpositioning and other surgical difficulties. This integration helps to reduce the frequency of postoperative complications and the need for subsequent revision surgeries.

[0200] One benefit of the described system is the reduction of the patient's recovery time, as precisely designed prosthetics / implants and carefully planned surgical procedures contribute to shortening the hospital stay and enabling the patient to resume daily activities more quickly. The system's simplified surgical approach reduces the time spent in the operating room and minimizes the consumption of medical resources, thus saving economic costs for healthcare institutions and patients.

[0201] The AI system can have a continuous improvement mechanism where the artificial intelligence module learns from each surgical intervention and optimizes its algorithms. This mechanism ensures the application of state - of - the - art techniques in surgical implant design and surgical procedure planning. The system can be configured to facilitate the collection and analysis of postoperative data, advancing the understanding of surgical implant performance and patient outcomes.

[0202] In this additional example, the system for automating orthopedic prosthesis design introduces a higher degree of personalization, prediction accuracy, and efficiency, as well as advanced training capabilities and continuous data - driven improvement.

[0203] Although not required, the embodiments described with reference to the accompanying drawings can be implemented as an application programming interface (API) or as a series of libraries for developers to use, or can be included in another software application, such as a terminal or personal computer operating system or a portable computing device operating system. Generally, as program modules include routines, programs, objects, components, and data files that assist in performing specific functions, those skilled in the art will understand that the functionality of a software application can be distributed among multiple routines, objects, or components to achieve the same desired functionality.

[0204] It will also be understood that if the methods and systems of the present invention are implemented entirely or in part by a computing system, any suitable computing system architecture may be utilized. This will include stand-alone computers, network computers, and dedicated hardware devices. When using the terms "computing system" and "computing device", these terms are intended to cover any suitable arrangement of computing hardware capable of implementing the described functionality.

[0205] Any reference to prior art contained herein should not be taken as an admission that such information is common general knowledge unless otherwise stated.

[0206] In addition, note that example forms may be described as processes depicted in flowcharts, flow diagrams, structural diagrams, or block diagrams. Although a flowchart may describe operations as a sequential process, many operations may be performed in parallel or simultaneously. In addition, the order of operations may be rearranged. The process terminates when its operations are complete. A process may correspond to a method, function, procedure, subroutine, subprogram, etc. In a computer program. When a process corresponds to a function, its termination corresponds to the function returning to the calling function or the main function.

[0207] One or more of the components and functions shown in the figures may be rearranged and / or combined into a single component or embodied as multiple components without departing from the scope of the present invention. Additional elements or components may also be added without departing from the scope of the present invention. In addition, the features described herein may be implemented in software, hardware, business methods, and / or combinations thereof.

[0208] Those skilled in the art will understand that many variations and / or modifications may be made to the specific embodiments shown without departing from the spirit or scope of the present invention.

[0209] Accordingly, the present examples and forms should be considered illustrative rather than restrictive in all respects.

Claims

1. A computer-implemented method for automated orthopedic prosthesis design, characterized in that, Comprising the following steps: Receiving at least one medical image of a patient's bone, Generating a 3D model of the bone from the at least one medical image, Determining a defective area in the bone, Generating a defect reconstruction model, and Generating a 3D prosthesis model corresponding to the defect reconstruction model, wherein the 3D prosthesis model is patient-specific.

2. The method according to claim 1, wherein Further comprising the step of exporting the 3D prosthesis model to print an orthopedic prosthesis corresponding to the shape and configuration of the 3D prosthesis model by means of 3D printing.

3. The method according to claim 1, wherein the defect reconstruction model is a 3D model corresponding to a part of a normal bone without bone defects.

4. The method according to claim 1, characterized in that Further comprising the following steps: Performing volume segmentation on the medical image, and Generating a 3D model of the bone based on the volume segmentation.

5. The method according to claim 4, characterized in that Wherein the step of performing volume segmentation on the medical image comprises the following additional steps: Preprocessing the received at least one medical image by normalization and / or standardization, extracting context information from the at least one medical image, Reconstructing spatial details by upsampling, and Generating a 3D model of the bone based on the context information and the reconstructed spatial details.

6. The method according to claim 5, characterized in that Wherein the step of extracting context information comprises the following steps: Performing 3D convolution on the at least one medical image, Performing batch normalization on the at least one medical image and / or the 3D convolution, and Performing ReLU activation.

7. The method according to claim 1, wherein Further comprising the additional steps of determining a surgical path for removing the bone defect and displaying the determined surgical path on a user interface.

8. The method according to claim 1, characterized in that, Wherein the step of determining the surgical path comprises the following additional steps: Determining an incision for removing the bone defect, wherein determining the incision comprises determining the size and shape of the incision for removing the bone defect, Displaying the determined incision on the user interface, Determining the position of the reconstruction model to correct the bone defect, and Determining a fixation configuration for fixing the orthopedic prosthesis to the bone to correct the bone defect.

9. The method according to claim 8, characterized in that, Wherein the 3D prosthesis model is generated based on the determined incision, the reconstruction model, and the fixation configuration.

10. The method according to claim 3, characterized in that, Wherein the 3D prosthesis model is customized in shape and configuration to match the patient's bone, such that the 3D prosthesis model is patient-specific.

11. A machine learning model for automated orthopedic prosthesis design, characterized in that, Including a method for any one of claims 1-10, the machine learning model comprising: An encoder path configured to capture context information, A decoder path suitable for reconstructing spatial information, and Skip connections connecting corresponding encoder layers and decoder layers.

12. The machine learning model according to claim 11, wherein Wherein the encoder path comprises multiple layers, the decoder path comprises multiple layers, and the machine learning model is a 3D U-net model.

13. A data processing device for automated orthopedic prosthesis design, characterized in that, Including an apparatus for performing the method according to any one of claims 1-10.

14. A computer program, characterized in that, Including instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1-10.

15. A computer-readable medium, characterized in that, Including instructions that, when executed by the computer, cause the computer to perform the method according to any one of claims 1-10.

16. A computer-implemented method for training a machine learning model for automated orthopedic prosthesis design, characterized in that, Including the machine learning model according to claim 11, the method comprising the following steps: Receive an input training dataset including multiple medical images and ground truth segmentation, where the multiple medical images include images of a bone with a bone defect; Pass the input training dataset through the convolutional layers of a 3D U-net model, Optimize the model parameters using a combination of Dice coefficient loss and binary cross-entropy, Enhance the generalization ability by applying one or more mutation techniques, and Monitor the model performance based on a validation dataset to ensure convergence and reduce overfitting.

17. A training data set for the method according to claim 16, characterized in that, Comprising: Input data, where the input data includes multiple images or models with one or more bone defects, Output data, where the output data includes multiple images or models of normal bones.

18. A system for automated orthopedic prosthesis design and construction, characterized in that, Comprising: A computing device including a processor and a storage unit, the processor and the storage unit being operatively coupled, The storage unit is adapted to store instructions that, when executed by the processor, cause the computing device to perform the method of any one of claims 1-10 and 16.

19. A system for automated orthopedic prosthesis design and construction, characterized in that, Comprising: A segmentation module and an orthopedic prosthesis design module, Wherein the segmentation module is configured to: Receive at least one medical image of a patient's bone, Generate a 3D model of the bone from the at least one medical image, and Determine the defective area in the bone, Wherein the orthopedic prosthesis design module is configured to: Generate a defect reconstruction model, and Generate a 3D prosthesis model corresponding to the defect reconstruction model, where the 3D prosthesis model is patient-specific.

20. The system according to claim 19, wherein Wherein the orthopedic prosthesis design module is configured to export the 3D prosthesis model to print an orthopedic prosthesis corresponding to the shape and configuration of the 3D prosthesis model using 3D printing.

21. The system according to claim 19, wherein Wherein the defect reconstruction model is a 3D model corresponding to the normal bone part without bone defects.

22. The system according to claim 19, wherein Wherein the segmentation module is configured to: Perform volume segmentation on the medical image, and Generate the 3D model of the bone based on the volume segmentation.

23. The system according to claim 22, characterized in that, Comprising: A medical image preprocessing module, where the medical image preprocessing module is configured to: Preprocess the received at least one medical image by normalization and / or standardization, and the segmentation module is configured to: Extract context information from the at least one medical image, Reconstruct spatial details by upsampling, and Generate the 3D model of the bone based on the context information and the reconstructed spatial details.

24. The system according to claim 19, wherein Comprising a surgical planning module, which is configured to: determine the surgical path for removing the bone defect and display the determined surgical path on a user interface.

25. The system according to claim 19, wherein Wherein the surgical planning module is configured to: Determine the size and shape of the incision for removing the bone defect, Display the determined incision on the user interface, Determine the position of the reconstruction model to correct the bone defect, and Determine the fixation configuration for fixing the orthopedic prosthesis to the bone to correct the bone defect.

26. The system according to claim 25, wherein, Wherein the orthopedic prosthesis design module is configured to generate the 3D prosthesis model based on the determined incision, the reconstruction model, and the fixation configuration.

27. The system according to claim 19, wherein Wherein the 3D prosthesis model is customized in shape and configuration to match the patient's bone, such that the 3D prosthesis model is patient-specific.

28. The system according to claim 19, wherein Comprising a machine learning model, which includes: An encoder path configured to capture context information, A decoder path adapted to reconstruct spatial information, and Skip connections connecting corresponding encoder layers and decoder layers, Wherein the segmentation module and the orthopedic prosthesis design module are part of a machine learning module.

29. The system according to claim 28, wherein, Wherein the encoder path within the machine learning model includes multiple layers, the decoder path includes multiple layers, and the machine learning model is a 3D U-net model.

Images