An AI-based system and method for diagnosis and surgical planning of orthopedic external fixation devices.
By combining artificial intelligence systems with deep learning and biomechanical constraints, the standardization and three-dimensional planning problems of orthopedic external fixation system diagnosis have been solved, enabling precise stent configuration and preoperative visual planning, and reducing the risk of postoperative complications.
Patent Information
- Application Number
- CN202511927166.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-11-17
AI Technical Summary
The diagnosis of orthopedic external fixation devices in the current technology relies on subjective experience and lacks standardized interaction methods. The selection of external fixation devices lacks quantitative biomechanical basis, and two-dimensional images cannot be used for three-dimensional precise planning, resulting in a high risk of misdiagnosis and postoperative complications.
An AI-based diagnostic and surgical planning system is employed, which uses a deep learning segmentation model combined with Gaussian heatmaps and biomechanical constraints to achieve accurate fracture identification and scaffold configuration. A statistical deformation model is used to reconstruct a three-dimensional skeletal model, providing a virtual surgical planning report.
It enables collaborative diagnosis between doctors and AI, ensuring that stent selection is scientific and reasonable, reducing the risk of postoperative complications, providing three-dimensional visualization planning, reducing radiation exposure, and improving diagnostic accuracy and surgical outcomes.
Smart Images

Figure CN123120199A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer-aided surgical technology, specifically to an artificial intelligence-based system and method for diagnosing and planning orthopedic external fixation devices. Background Technology
[0002] External fixation techniques are crucial for orthopedic treatment of complex fractures, correction of bone deformities, and bone lengthening surgeries. In traditional clinical practice, doctors primarily rely on X-ray films or two-dimensional digital imaging for diagnosis and preoperative planning. However, the existing diagnostic and treatment procedures have the following significant shortcomings: First, diagnosis relies on subjective experience and lacks standardized interaction methods. Although there are studies using deep learning to identify fractures, the recognition rate of pure AI models often drops when faced with complex situations such as comminuted fractures or overlapping fracture lines, and missegmentation is frequent. Existing correction methods usually involve doctors manually drawing lines on the image layers for annotation. This annotation cannot be fed back to the algorithm core, causing the system's subsequent measurement calculations to still be based on incorrect AI segmentation data, resulting in a "human-machine disconnect."
[0003] Secondly, the selection of external stents lacks quantitative biomechanical basis. Doctors mainly rely on personal experience to select stent models (such as ring diameter and link length) and screw layout. Due to the lack of quantitative calculations on fracture end displacement, soft tissue swelling, and overall stent system stiffness, complications such as stents that are too small and compress the skin, or stents with insufficient stiffness leading to excessive micromovement at the fracture ends and affecting healing are prone to occur.
[0004] Finally, two-dimensional imaging lacks the predictive power of three-dimensional space. External fixation devices are three-dimensional structures, while X-rays are two-dimensional projections. Surgeons cannot accurately determine, based solely on two-dimensional images, whether the device will collide with the limb in three-dimensional space, or whether the screw path will penetrate the neurovascular safety zone. Although CT scans can provide three-dimensional information, their high radiation dose and cost make them unsuitable as a routine preoperative planning method. Therefore, there is an urgent need for an artificial intelligence-based diagnostic and surgical planning system and method for orthopedic external fixation devices. Summary of the Invention
[0005] The purpose of this invention is to provide an orthopedic external fixation system and surgical planning system and method based on artificial intelligence, so as to solve the technical problems in the prior art, such as the fragmented diagnostic correction mechanism, lack of mechanical basis for selection, and inability to perform three-dimensional precise planning based on two-dimensional images.
[0006] To achieve the above objectives, the present invention provides the following technical solution: An artificial intelligence-based method for diagnosis and surgical planning of orthopedic external fixation devices includes the following steps: Two-dimensional medical image data of the target lesion is acquired, physical reference objects in the two-dimensional medical image data are identified, and a mapping relationship between pixel size and physical size is established based on the physical reference objects to complete image calibration and preprocessing. The preprocessed two-dimensional medical image data is input into a pre-trained deep learning segmentation model, which outputs a first segmentation mask containing the probability distribution of the fracture area. In response to a correction command input by the user through a human-computer interaction interface, the system captures the trajectory coordinate sequence drawn by the user and generates a spatial constraint feature vector based on the trajectory coordinate sequence. The spatial constraint feature vector is weighted and fused with the first segmentation mask. Based on the fused data, the local region is re-segmented to generate the corrected target fracture parameters. Based on the target fracture parameters, geometric quantitative features are extracted, and combined with the pre-set external fixation product database and biomechanical constraints, an external fixation configuration scheme is generated through a parameterized constraint solving algorithm. A three-dimensional support model is constructed based on the external fixation support configuration scheme. A pre-set three-dimensional bone model containing the average shape and principal components of deformation is called. By calculating the deviation between the projection contour of the pre-set three-dimensional bone model on the two-dimensional plane and the bone edge in the two-dimensional medical image data, the deformation parameters of the pre-set three-dimensional bone model are iteratively adjusted to complete the three-dimensional reconstruction. The three-dimensional support model is spatially registered with the three-dimensional morphology of the reconstructed bone of the affected area, and a visualization planning report containing virtual surgical effects is generated.
[0007] Furthermore, the step of generating a spatial constraint feature vector based on the trajectory coordinate sequence specifically includes: generating a Gaussian heatmap centered on the trajectory coordinate sequence; and mapping the pixel values in the Gaussian heatmap to energy constraint values or confidence weight matrices as the spatial constraint feature vector.
[0008] Furthermore, the step of generating an external fixation device configuration scheme by combining a pre-set external fixation device product database and biomechanical constraints using a parametric constraint solving algorithm specifically includes: Mechanical stability constraints are established, requiring the diameter of the external fixation ring assembly to be greater than the sum of the physical diameter of the bone and the estimated thickness of the soft tissue, while maintaining a preset safety buffer distance; anatomical safety constraints are established, based on the human cross-sectional anatomical atlas, to determine the permissible needle insertion sector that avoids nerve and blood vessel tissues. A stiffness assessment model is constructed to simulate the micromotion stability of fracture ends under different support configurations, and the optimal configuration scheme is generated within the solution space that satisfies the above constraints.
[0009] This invention also provides an artificial intelligence-based orthopedic external fixation system for diagnosis and surgical planning, comprising: The image calibration and preprocessing module is used to establish the mapping relationship between pixels and physical dimensions; the human-machine collaborative intelligent diagnosis module includes a deep learning inference unit and an interactive correction unit, which is used to generate accurate fracture geometric parameters through a dynamic weighted correction algorithm driven by Gaussian heatmap. The biomechanical constraint formulation generation module is used to automatically calculate the configuration scheme of external fixation devices based on mechanical stability constraints, anatomical safety constraints, and stiffness assessment models. The virtual surgical simulation and registration module is used to reconstruct the three-dimensional morphology of the affected bone based on the statistical deformable model (SSM) technology and generate digital reconstructed radiographic projection (DRR) images. The comprehensive report generation module is used to output planning reports.
[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: By employing a dynamic correction algorithm based on Gaussian heatmaps, each drawing by the doctor is not merely a visual marker, but is transformed into a mathematical energy constraint or weight matrix. This directly drives the underlying graph cut or region growing algorithm to recalculate the segmentation boundary. This mechanism leverages the efficiency of AI while incorporating the doctor's professional judgment, effectively solving the problem of inaccurate AI identification in complex bone fragmentation areas.
[0011] By introducing biomechanical constraints, the selection of external stents has been made more scientific and quantitative, changing the previous reliance on doctors' experience to estimate stent models. By establishing a multi-dimensional parametric solution model that includes geometric constraints, soft tissue safety buffering, anatomical obstacle avoidance, and system stiffness assessment, it can be ensured that the recommended stent is not only appropriately sized (avoiding skin compression), but also mechanically stable enough (promoting bone healing) and can automatically avoid high-risk areas of nerves and blood vessels, significantly reducing the risk of postoperative complications.
[0012] By utilizing the Statistical Deformation Model (SSM) and projection optimization technology, a patient-specific three-dimensional skeletal model was successfully reconstructed based on conventional two-dimensional X-ray images. This allows doctors to intuitively assess the scaffold assembly effect and needle insertion path in three-dimensional space before surgery without requiring patients to undergo CT scans at high cost and with high radiation exposure. This achieves a two-dimensional input and three-dimensional planning experience.
[0013] By registering and comparing the actual postoperative images with the preoperative plan, the execution error can be calculated and fed back to the recommendation algorithm. This allows the system to continuously optimize its recommendation strategy as case data accumulates, gradually aligning with actual clinical practice. Attached Figure Description
[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.
[0015] Figure 1 This is a schematic diagram of the overall hardware architecture and functional module connections of the system described in the embodiment of the present invention; Figure 2 This is a general flowchart of the method described in the embodiments of the present invention; Figure 3 This is a schematic diagram illustrating the interactive correction logic principle of the diagnostic pen based on Gaussian heatmap in an embodiment of the present invention; Figure 4 This is a schematic diagram of the geometric cross-section principle of the mechanical constraint conditions of the external fixation frame in an embodiment of the present invention. The main component symbols are explained as follows: 1—Bone (corresponding to the tibia in this embodiment); 2—Soft tissue boundary (corresponding to the skin surface in this embodiment); 3—External fixation frame ring assembly; 4—Anatomical danger zone (corresponding to the anterior tibial neurovascular bundle sector in this embodiment). —Bone physical diameter ( ); —Estimated thickness of soft tissue ( ); —Safe buffer distance ( ); —Relative radius of external fixation bracket ( ).
[0016] Figure 5 This is a schematic diagram illustrating the 2D-3D projection optimization principle based on a statistical deformation model in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating an application scenario of the client interaction interface in an embodiment of the present invention; Figure 7 This is a schematic diagram comparing the superposition effect of the virtual external fixator and the affected bone before and after planning in an embodiment of the present invention (the left image is before planning, and the right image is after virtual surgical planning). Figure 7 Explanation of key component symbols: 71—Bone at the affected site (corresponding to the tibia and fracture site in the embodiment); 72—Virtual external fixation frame (overall structure); 73—Frame connection assembly (including connecting rods and virtual screw paths). Detailed Implementation
[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0018] like Figure 1 As shown in the figure, the artificial intelligence-based orthopedic external fixation system for diagnosis and surgical planning described in this embodiment is deployed on a server containing a high-performance computing unit and a doctor's operating terminal.
[0019] Server-side: Equipped with an NVIDIA RTX 3090 or higher performance GPU for running deep learning segmentation models (such as U-Net or Mask R-CNN) and performing heavy 3D rendering and projection calculations.
[0020] Client: A PC or tablet used by doctors, equipped with a high-resolution display and human-computer interaction devices (mouse or stylus) to display medical images and receive drawing instructions from doctors.
[0021] This embodiment uses the diagnosis and treatment process of a patient with a tibial fracture as an example. The main workflow of the system is as follows: Figure 2 As shown, the specific steps include: Step S1: The image acquisition, calibration, and preprocessing system acquires anteroposterior and lateral X-ray images of the patient's affected area (tibia) from the hospital's PACS system via a DICOM interface. Image calibration: The system automatically identifies a pre-placed metal ball or ruler in the image as a physical reference. Assume the diameter of the metal ball is known. (For example, 25mm), which occupies a pixel width in the image. Then establish pixel-physical size mapping coefficients (mm / pixel), the system will use this coefficient Globally applied to the backbone diameter in subsequent step S3 and soft tissue thickness The calculations ensure that all geometric parameters are true physical dimensions.
[0022] Preprocessing: Adaptive histogram equalization (CLAHE) is performed on the original image to enhance the contrast between bone edges and soft tissue, and the image is uniformly resampled to... or Resolution.
[0023] Step S2: Human-machine collaborative intelligent diagnosis, which specifically includes the following sub-steps: AI initial screening: The pre-processed image is input into the pre-trained Mask R-CNN model, which outputs the first segmentation mask, containing a probability distribution map of the fracture line location and bone fragment regions. .
[0024] Doctor intervention: If the AI identification is biased (for example, missing a subtle spiral fracture line), the doctor can click the "diagnostic pen" tool on the interface to draw a red line at the missed area; The logical principle of the dynamic correction algorithm is as follows: Figure 3 As shown: The system captures the trajectory coordinate sequence drawn by the doctor. , with trajectory Generate a Gaussian heatmap centered on the target element. : in This represents the shortest Euclidean distance from the pixel to the trajectory. Parameters used to control the diffusion range of the heat map.
[0025] Weighted fusion: Gaussian heatmaps are used as strong prior weights and superimposed onto the original AI probability map. In the formula, it is: in, This is the fusion coefficient.
[0026] Local resegmentation: Using the trajectory points drawn by the doctor as "foreground seeds", the graph cut algorithm is used to find the minimum energy cut path on the updated probability graph, so that the segmentation boundary automatically snaps to the real anatomical edge, rather than simply covering the lines. The whole process achieves coarse drawing and precise segmentation.
[0027] Step S3: The biomechanical constraint-based formulation generation system automatically extracts geometric features and calculates the framework formulation based on the corrected segmentation results, specifically: Parameter extraction: Calculation of the physical diameter of the backbone (e.g., 30mm), maximum soft tissue thickness (e.g., 45mm), fracture end displacement .
[0028] Perform constraint solving: such as Figure 4 The geometric cross-section principle shown indicates that the mechanical stability constraint requires a certain diameter for the external fixed support ring. (Marked R in the diagram) must satisfy the formula: In the diagram, 1 represents a bone (diameter corresponding to...). Mark 2 indicates the soft tissue boundary (thickness corresponds to). Mark 3 indicates the external fixation ring, and mark 4 indicates the vascular and nerve danger zone to be avoided; The mechanical stability constraint is: the diameter of the external fixed support ring. Must meet: in, For a safe buffer distance (set to 20mm), the system selects a suitable ring model (such as a 160mm ring) from the database.
[0029] Anatomical safety constraints: The system has a built-in tibial cross-sectional anatomical atlas that defines the sectors where the tibial nerve and anterior tibial vessels are located (e.g., the anterior 15°-45° region). The system automatically excludes needle insertion paths that cross this sector.
[0030] Stiffness assessment: The system simulates the axial stiffness and shear stiffness of the fracture ends under the current configuration based on a preset simplified beam element model. If the stiffness is lower than the threshold (e.g., for unstable fractures, stiffness > 2000 N / mm is required), the system automatically increases the number of links or changes the link angle (e.g., recommending an upgrade from "3-link configuration" to "4-link configuration").
[0031] Output formula: The system's final recommendation is: Taylor space bracket, 2 x 160mm full rings, 1 x 150mm half ring, 6 Schanz screws, and specific linkage length adjustment parameters.
[0032] Step S4: 3D Virtual Planning on 2D Images (SSM Reconstruction Technology) To display realistic 3D assembly effects on 2D images, the system performs the following operations: To solve the perspective projection problem, the system adopts... Figure 5 The SSM projection optimization process shown is as follows: A general tibial statistical deformation model (also known as the statistical shape model SSM) is retrieved from the database. This model represents the average shape. Weighted eigenvectors Linear combination: in This is the deformation weighting parameter.
[0033] Perform projection optimization: transform the model Through perspective projection matrix Projecting onto a 2D plane yields the contour. .calculate Compared to the actual bone edges in the X-ray The chamfer distance between them is used as an error function. The weights are iteratively adjusted using gradient descent. The spatial pose (rotation, translation) of the model causes errors. After minimization and convergence, a "patient-specific 3D skeletal model" that closely matches the patient's X-ray is obtained.
[0034] On the reconstructed 3D bone model, a 3D external fixation frame model is automatically assembled according to the recipe in step S3. Using digital reconstructed radiography (DRR) technology, the attenuation process of X-rays passing through the 3D scene is simulated to generate a 2D image with a semi-transparent perspective effect. This DRR image is then alpha-blended and superimposed with the original X-ray film, allowing the doctor to see the occlusion relationship between the anterior and posterior rings of the frame, and whether the screws have passed through the center of the medullary cavity in three-dimensional space.
[0035] Step S5: Result Output and Closed-Loop System Generates a PDF report containing the above diagnostic data, stent BOM (Bill of Materials), and virtual surgical diagram, specifically including: The postoperative X-ray images are imported back into the system. The system automatically registers the postoperative images with the preoperative plan, calculates the deviation between the actual needle insertion point and the planned point (e.g., a deviation of 3mm), and stores this data in the background for training subsequent recommendation weights (for example, if it is found that doctors tend to insert one more screw than the system recommends for a certain type of fracture, the system will update the recommendation strategy through reinforcement learning).
[0036] In an example scenario (the client interface looks like this) Figure 6-7 (as shown) The doctor treated a young patient with a spiral fracture of the right tibia.
[0037] The doctor then imported the X-ray, and the AI system automatically identified the main fracture line, but missed a tiny extended crack below it. The doctor clicked on the "diagnostic pen" and gently drew a line on the crack. The system responded by using a Gaussian heatmap algorithm to automatically capture and highlight the precise edge of the crack, updating the diagnosis. Based on the corrected data, the system calculated the bone diameter to be 32 mm and the soft tissue swelling thickness to be 40 mm, based on the mechanical constraint formula ( The system automatically recommended the 160mm Taylor space stent and warned the doctor: "There are blood vessels in the 30-degree area in front, it is recommended to avoid it." The screen displays an image showing the effect after the bracket is installed, such as... Figure 7As shown, the system's output visualization planning report includes a comparison view of before planning (left image) and after planning (right image). In the right image, the system uses SSM technology to reconstruct the three-dimensional morphology of the affected bone 71 and assembles the virtual external fixator 72 on the outside of the bone according to the calculated formula parameters. The doctor can intuitively observe the spatial layout of the support connection components 73 (including telescopic rods and screws). The planning results shown in the figure indicate that the installation position of the support avoids the soft tissue danger zone and the fracture ends are well aligned, verifying the effectiveness of the algorithm of this invention. If the user adjusts the position of the virtual screw in the interactive interface (e.g., moves it upwards), the system uses 3D projection technology to adjust it by dragging with the mouse. The system uses 3D projection technology to refresh the image in real time and indicates that the mechanical stability after adjustment is still qualified. The doctor clicked "Confirm," printed out the report, and took it into the operating room.
[0038] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence, characterized in that, The method includes the following steps: Two-dimensional medical image data of the target lesion is acquired, physical reference objects in the two-dimensional medical image data are identified, and a mapping relationship between pixel size and physical size is established based on the physical reference objects to complete image calibration and preprocessing. The preprocessed two-dimensional medical image data is input into a pre-trained deep learning segmentation model, which outputs a first segmentation mask containing the probability distribution of the fracture area. In response to a correction command input by the user through a human-computer interaction interface, the system captures the trajectory coordinate sequence drawn by the user and generates a spatial constraint feature vector based on the trajectory coordinate sequence. The spatial constraint feature vector is weighted and fused with the first segmentation mask. Based on the fused data, the local region is re-segmented to generate the corrected target fracture parameters. Based on the target fracture parameters, geometric quantitative features are extracted, and combined with the pre-set external fixation product database and biomechanical constraints, an external fixation configuration scheme is generated through a parameterized constraint solving algorithm. A three-dimensional support model is constructed based on the external fixation support configuration scheme. A pre-set three-dimensional bone model containing the average shape and principal components of deformation is called. By calculating the deviation between the projection contour of the pre-set three-dimensional bone model on the two-dimensional plane and the bone edge in the two-dimensional medical image data, the deformation parameters of the pre-set three-dimensional bone model are iteratively adjusted to complete the three-dimensional reconstruction. The three-dimensional support model is spatially registered with the three-dimensional morphology of the reconstructed bone of the affected area, and a visualization planning report containing virtual surgical effects is generated.
2. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 1, characterized in that, The generation of spatial constraint feature vectors based on the trajectory coordinate sequence specifically includes: Obtain the spatial location information of each coordinate point in the trajectory coordinate sequence; generate a Gaussian heatmap centered on the trajectory coordinate sequence; The pixel values in the Gaussian heatmap are mapped to energy constraint values or confidence weight matrices, and the energy constraint values or the confidence weight matrix are defined as the spatial constraint feature vectors. The correction instructions include drawing operations for the direction of the fracture line or delineating operations for the boundaries of fracture fragments.
3. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 2, characterized in that, The step of weightedly fusing the spatial constraint feature vector with the first segmentation mask, and performing local region resegmentation calculation based on the fused data, specifically includes: The pixel classification probability values in the first segmentation mask are weighted point-to-point with the confidence weight matrix to generate an updated pixel classification confidence map. Using the points on the trajectory coordinate sequence as seed points, the boundary of the target region is recalculated on the updated pixel classification confidence map using a graph cut algorithm or a region growing algorithm. Alternatively, an active contour model can be constructed, and the energy constraint value can be introduced into the energy function of the active contour model as an external force constraint to drive the initial contour to converge to the trajectory coordinate sequence, thereby obtaining an accurate segmentation boundary.
4. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 1, characterized in that, The extraction of geometric quantitative features based on the target fracture parameters specifically includes: Based on the corrected target fracture parameters and the mapping relationship between pixel size and physical size, the physical displacement, angular angle, physical diameter of bone shaft and estimated thickness of soft tissue at the fracture ends are calculated. Identify anatomical landmarks in the skeletal image and establish a local coordinate system for surgical planning based on the anatomical landmarks.
5. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 4, characterized in that, The process of generating an external fixation device configuration scheme by combining a pre-set external fixation device product database and biomechanical constraints using a parametric constraint solving algorithm specifically includes: Retrieve the geometric parameters and material properties of the external fixation bracket assembly from the external fixation bracket product database; Mechanical stability constraints are constructed by setting the diameter of the external fixation ring assembly to be greater than the sum of the physical diameter of the bone and the estimated thickness of the soft tissue, while maintaining a preset safety buffer distance; anatomical safety constraints are constructed by determining the permissible needle insertion sector that avoids nerve and blood vessel tissue based on the built-in human cross-sectional anatomical atlas; and a stiffness assessment model is constructed by simulating the micromotion stability of the fracture ends under different support configurations based on the physical displacement of the fracture ends and the material property parameters. Within the solution space that satisfies the mechanical stability constraints, the anatomical safety constraints, and the preset stiffness threshold, a multi-objective optimization algorithm is used to calculate the recommended stent model, link length, number of screws, and spatial angle of screw implantation to generate the external fixation stent configuration scheme.
6. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 1, characterized in that, The process involves calling a pre-defined 3D skeletal model containing average shape and principal components of deformation. By calculating the deviation between the projected contour of the pre-defined 3D skeletal model on a 2D plane and the bone edges in the 2D medical image data, the deformation parameters of the pre-defined 3D skeletal model are iteratively adjusted to complete the 3D reconstruction. Specifically, this includes: The statistical deformation model, which is pre-constructed based on multi-sample skeletal data using principal component analysis, is invoked. The statistical deformation model is represented by a linear combination of an average shape vector and a weighted eigenvector. The initial three-dimensional model corresponding to the average shape vector is mapped to the imaging plane of the two-dimensional medical image data through perspective projection transformation to establish an initial projection contour corresponding to the initial three-dimensional model. Edge detection is performed on the two-dimensional medical image data to extract pixel gradient features of bone edges; Construct an objective function to measure the shape difference between the initial projected contour and the pixel gradient features of the bone edge; The weight parameters of the feature vector and the spatial pose parameters of the statistical deformation model are iteratively adjusted using an optimization algorithm. The initial three-dimensional model is driven to deform and displace by minimizing the value of the objective function until the value of the objective function converges to a preset threshold. The converged model is then determined as the three-dimensional morphology of the reconstructed bone of the affected area.
7. The method for diagnosis and surgical planning of orthopedic external fixation devices based on artificial intelligence according to claim 1, characterized in that, The method also includes a postoperative assessment and feedback step: Acquire two-dimensional medical imaging data of the patient after surgery; The postoperative two-dimensional medical image data is compared with the virtual surgical effect in the visualization planning report to perform image registration and difference analysis, and the actual execution error is calculated. The actual execution error is fed back into the parameterized constraint solving algorithm, and the recommended weight parameters of the external fixation bracket configuration scheme are updated using machine learning algorithms.
8. An artificial intelligence-based orthopedic external fixation device diagnostic and surgical planning system for performing the method according to any one of claims 1 to 7, characterized in that, The system includes: The image calibration and preprocessing module is used to acquire two-dimensional medical image data of the target lesion and establish a size mapping relationship based on physical reference objects in the image; The human-machine collaborative intelligent diagnostic module includes a deep learning inference unit and an interactive correction unit; The deep learning inference unit is used to generate an initial fracture segmentation mask; The interactive correction unit is used to receive correction instructions and generate spatial constraint feature vectors, drive the segmentation algorithm to dynamically weight and correct the initial fracture segmentation mask, and output accurate fracture geometric parameters. The biomechanical constraint formula generation module is used to store external fixation bracket product data and anatomical rules, and automatically calculate an external fixation bracket configuration scheme including component models and layout parameters based on the fracture geometry parameters, combined with mechanical stability constraints, anatomical safety constraints and stiffness assessment models. The virtual surgical simulation and registration module is used to generate a three-dimensional support model based on the external fixation support configuration scheme; and includes a three-dimensional reconstruction unit, which is configured to call a preset statistical deformation model and calculate the three-dimensional bone morphology parameters by minimizing the difference between the model projection contour and the image edge features; the virtual surgical simulation and registration module is also used to assemble the three-dimensional support model onto the calculated three-dimensional bone morphology and generate a digital reconstructed radiographic projection image; The comprehensive report generation module is used to output a comprehensive report that includes diagnostic conclusions, a list of external fixation device configuration schemes, and a visual surgical planning diagram.
9. The system according to claim 8, characterized in that, The system further includes a postoperative closed-loop feedback module, used to compare postoperative imaging data with preoperative planning data, calculate the deviation value, and use the deviation value to optimize the decision parameters of the biomechanical constraint formula generation module.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.
