An automated planning method, system, and electronic equipment for patellofemoral joint femoral prostheses.

Abstract

This invention provides an automated planning method, system, and electronic device for patellofemoral prostheses, comprising: acquiring CT and MRI images of the patient's knee joint; segmenting the femoral region based on the CT images and identifying key anatomical points on the femur; segmenting the femoral cartilage region based on the MRI images; fusing the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces; obtaining the femoral prosthesis model based on the initial femoral model and key anatomical points; reconstructing the femoral cartilage in the initial femoral model to generate a target femoral model, thereby determining the implantation pose of the femoral prosthesis; and determining the femoral osteotomy area matching the femoral prosthesis based on the implantation pose. This method overcomes the shortcomings of traditional planning methods that rely on physician experience, such as high subjectivity, low efficiency, and difficulty in restoring the physiological morphology of the joint, significantly improving the accuracy of preoperative planning and providing technical support for robot-assisted patellofemoral joint replacement surgery.

Description

Technical Field

[0001] This invention relates to the field of automatic planning technology for patellofemoral joint femoral prostheses, specifically to an automatic planning method, system, and electronic device for patellofemoral joint femoral prostheses. Background Technology

[0002] Patellofemoral arthroplasty (PFA) is an effective treatment for isolated patellofemoral arthritis. By replacing the diseased femoral trochlea and patellar articular surface, it preserves the healthy tibiofemoral joint structure, offering advantages such as minimal trauma and rapid recovery. However, the success of this surgery is highly dependent on the precision of the femoral prosthesis placement. An ideal prosthesis position can restore normal patellar trajectory and joint biomechanics, thereby reducing the risk of postoperative complications such as patellar dislocation and wear.

[0003] Currently, preoperative planning methods in clinical practice rely heavily on physician experience and manual operation, lacking objective standards. Existing planning systems typically require physicians to manually mark key anatomical points (such as the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur) on two-dimensional or three-dimensional images, manually measure dimensions (such as the length of the Whiteside line), and repeatedly adjust the implantation position and posture of the prosthesis based on personal experience. This approach is not only time-consuming and labor-intensive, resulting in low planning efficiency, but its planning results are also easily affected by the physician's subjective judgment and skill level, making it difficult to guarantee consistency and repeatability between different surgeries.

[0004] Current technologies are ineffective in addressing pathological cartilage wear, and planning goals are often biased. For patients with existing cartilage wear, current planning systems typically match prostheses directly to the post-wear anatomical morphology. This results in planning schemes designed to adapt to the pathological anatomy rather than correct it, failing to restore the joint line to its normal physiological position. The patellar trajectory remains abnormal, creating a risk of postoperative functional impairment and early prosthesis failure. Therefore, there is a lack of an automated, personalized, and precise method for patellofemoral prosthesis planning to overcome over-reliance on surgeon experience, improve planning efficiency and quality, and provide reliable operational guidelines for robotic surgery.

[0005] Therefore, existing technologies still need further development. Summary of the Invention

[0006] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide an automatic planning method, system, and electronic device for patellofemoral joint femoral prostheses to solve the problems existing in the prior art.

[0007] To achieve the above-mentioned technical objectives, according to a first aspect of the present invention, the present invention provides an automatic planning method for a patellofemoral joint femoral prosthesis, comprising:

[0008] S100. Acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images. S200: The femoral region and the femoral cartilage region are fused to generate an initial femoral model including the femoral and cartilage surfaces. The femoral prosthesis model is obtained based on the initial femoral model and the key anatomical points. S300. Reconstruct the femoral cartilage in the initial femoral model to generate a target femoral model, and determine the implantation position of the femoral prosthesis based on the target femoral model. S400. Based on the implantation position, determine the femoral osteotomy area that matches the femoral prosthesis.

[0009] Specifically, the method for segmenting the femoral region based on the CT image and identifying key anatomical points on the femur includes: The CT image is segmented using a trained first neural network model, and the segmented femoral region is output. The second neural network model identifies key anatomical points on the femoral region in the CT image and outputs the coordinate information of the key anatomical points. The key anatomical points include the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur.

[0010] Specifically, when using the second neural network model to identify key anatomical points on the femoral region in the CT image, the key anatomical points are converted into spherical regions with the point as the center and a first preset value as the radius. The method for outputting the coordinate information of the key anatomical points includes: Calculate the coordinates of the centroid of the spherical region and use the coordinates of the centroid as the coordinates of the key anatomical point.

[0011] Specifically, the method for segmenting the femoral cartilage region based on the MRI image includes: The MRI image is segmented using a trained third neural network model, and the segmented femoral cartilage region is output.

[0012] Specifically, the method for fusing the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces includes: A three-dimensional image registration algorithm is used to calculate the spatial transformation relationship between the CT image and the MRI image. Based on the spatial transformation relationship, the coordinate system of the femoral cartilage region is transformed to the first coordinate system of the CT image. In the first coordinate system, the transformed femoral cartilage region is merged with the femoral region to generate an initial femoral model.

[0013] Specifically, the method for obtaining the femoral prosthesis model based on the initial femoral model and the key anatomical points includes: The initial femoral model is converted into a three-dimensional surface point cloud; On the three-dimensional surface point cloud, the lowest point of the patellar trochlea is taken as the starting point of the path, and the midpoint of the intercondylar fossa of the femur is taken as the ending point of the path. The shortest path algorithm is used to calculate the shortest path connecting the starting point and the ending point, and the shortest path is defined as the femoral trochlear groove path. The femoral prosthesis model is determined based on the described femoral trochlear groove path.

[0014] Specifically, the method for reconstructing the femoral cartilage in the initial femoral model to generate the target femoral model includes: The surface of the initial femoral model is subjected to surface interpolation and smoothing to reconstruct a continuous femoral cartilage surface, thereby obtaining the target femoral model.

[0015] Specifically, the method for reconstructing the femoral cartilage in the initial femoral model to generate the target femoral model further includes: Obtain a femoral cartilage model of the non-operating knee joint of the patient; The femoral cartilage model of the non-operational knee joint is mirrored and flipped along the sagittal plane to obtain the first femoral cartilage model. The first femoral cartilage model is non-rigidly registered with the initial femoral model on the surgical side; The registered cartilage model is used as the reconstructed femoral cartilage surface on the surgical side to generate the target femoral model.

[0016] Specifically, the method for determining the implantation position of the femoral prosthesis based on the target femoral model includes: Using the key anatomical points and the femoral trochlear groove path as a reference, the preset reference points on the femoral prosthesis are aligned with the key anatomical points on the femoral region, and the central axis of the femoral prosthesis is aligned with the femoral trochlear groove path to determine the initial position of the femoral prosthesis. Based on the initial position, the femoral prosthesis is rotated along the central axis to adjust the inversion and valgus angles of the femoral prosthesis, thereby obtaining the implantation position of the femoral prosthesis.

[0017] Specifically, rotating the femoral prosthesis along the central axis to adjust the varus / valgus angle of the femoral prosthesis, thereby obtaining the implantation position of the femoral prosthesis, includes: At least one point is preset on the surface edge line of the femoral prosthesis. Then, the femoral prosthesis is rotated along the central axis to minimize the distance between the point on the edge line of the femoral prosthesis and the surface of the target femoral model, thereby obtaining the adjusted inversion and eversion angle of the femoral prosthesis, and thus obtaining the determined implantation posture of the femoral prosthesis.

[0018] Specifically, determining the femoral osteotomy region matching the femoral prosthesis based on the implantation position includes: In three-dimensional space, the intersection of the femoral prosthesis with the target femoral model is calculated when the implantation pose is in the specified position. The region defined by the intersection is taken as the femoral osteotomy region that matches the femoral prosthesis.

[0019] According to a second aspect of the present invention, an automatic planning system for a patellofemoral joint femoral prosthesis is provided, comprising: Image processing module: used to acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images; Model building module: used to fuse the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces, and to obtain the femoral prosthesis model based on the initial femoral model and the key anatomical points; Prosthesis planning module: used to reconstruct the femoral cartilage in the initial femoral model, generate a target femoral model, and determine the implantation pose of the femoral prosthesis based on the target femoral model; Osteotomy planning module: used to determine the femoral osteotomy area that matches the femoral prosthesis based on the implantation pose.

[0020] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, which, when executed by the processor, implement the above-described automatic planning method for a patellofemoral prosthesis.

[0021] Beneficial effects: This invention provides an automated planning method and system for patellofemoral prostheses. By integrating multimodal knee CT and MRI image data, it automates the entire process from image segmentation, model fusion, cartilage function reconstruction to prosthesis planning and osteotomy scheme generation. First, the invention accurately segments the femoral structure and identifies key anatomical points based on CT images, while simultaneously segmenting the cartilage region using MRI images. An initial femoral model including bone and cartilage is constructed through coordinate system one and data fusion. Then, pathological cartilage defects are digitally reconstructed to generate a target femoral model representing the morphology of a healthy joint. Based on this, the matching prosthesis model and implantation position are automatically determined. Finally, a precise osteotomy area is automatically generated based on the geometric relationship between the prosthesis and the bone model. This effectively overcomes the shortcomings of traditional planning methods that rely on physician experience, such as high subjectivity, low efficiency, and difficulty in restoring the physiological morphology of the joint. It significantly improves the accuracy, repeatability, and personalization of preoperative planning, providing reliable technical support for robot-assisted patellofemoral joint replacement surgery and greatly enhancing the intelligence, usability, and reliability of this invention. Attached Figure Description

[0022] Figure 1 This is a flowchart of an automatic planning method for patellofemoral joint femoral prosthesis provided in a specific embodiment of the present invention; Figure 2 This is a schematic diagram of the composition of the automatic planning system for the patellofemoral joint femoral prosthesis provided in a specific embodiment of the present invention; Figure 3 This is a schematic diagram of femoral cartilage segmentation provided in a specific embodiment of the present invention; Figure 4 This is a schematic diagram of femoral segmentation provided in a specific embodiment of the present invention; Figure 5 This is a schematic diagram of key anatomical points in the femoral region provided in a specific embodiment of the present invention; Figure 6 This is a schematic diagram of an initial femoral model including the femur and cartilage surface provided in a specific embodiment of the present invention; Figure 7 This is a schematic diagram of the femoral trochlear groove path solving process provided in a specific embodiment of the present invention; Figure 8 This is a schematic diagram of a three-dimensional point cloud of a bone surface including the femur and cartilage, provided in a specific embodiment of the present invention. Figure 9 This is a schematic diagram of the three-dimensional point cloud of the femoral prosthesis provided in a specific embodiment of the present invention; Figure 10 This is a schematic diagram of the initial position of the femoral prosthesis provided in a specific embodiment of the present invention; Figure 11This is a schematic diagram illustrating the fine-tuning and adjustment of the inversion and valgus angles of the femoral prosthesis in a specific embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Other similar embodiments obtained by those skilled in the art based on the embodiments in this application without creative effort should all fall within the scope of protection of this application. Furthermore, directional terms mentioned in the following embodiments, such as "up," "down," "left," and "right," are only for reference to the directions in the accompanying drawings; therefore, the directional terms used are for illustrative purposes and not for limiting the invention.

[0024] This invention addresses the problems of low planning accuracy, poor efficiency, and high risk of postoperative complications in traditional patellofemoral joint replacement surgery, which relies heavily on surgeon experience. It proposes an automated planning scheme based on multimodal imaging and artificial intelligence. This invention constructs an initial femoral model that includes the current anatomical structure by fusing the skeletal precision of CT images with the cartilage details of MRI images. Furthermore, it restores the femoral model to its healthy state using a unique cartilage reconstruction technique, generating a target femoral model. Based on this, it automatically completes prosthesis selection, pose optimization, and osteotomy plan generation. This method significantly improves the objectivity, accuracy, and repeatability of planning, providing a reliable basis for robot-assisted surgery.

[0025] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0026] Example 1 Please see Figure 1 This embodiment provides an automatic planning method for a patellofemoral joint femoral prosthesis, including: S100, acquiring CT and MRI images of the patient's knee joint, segmenting the femoral region based on the CT images, identifying key anatomical points on the femur, and segmenting the femoral cartilage region based on the MRI images; S200, fusing the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces, and obtaining the femoral prosthesis model based on the initial femoral model and the key anatomical points; S300, reconstructing the femoral cartilage in the initial femoral model to generate a target femoral model, and determining the implantation pose of the femoral prosthesis based on the target femoral model; S400, determining the femoral osteotomy region matching the femoral prosthesis based on the implantation pose.

[0027] It is understandable that this embodiment integrates multimodal imaging data from knee CT and MRI, achieving full automation of the entire process from image segmentation, model fusion, cartilage function reconstruction to prosthesis planning and osteotomy plan generation. This effectively overcomes the shortcomings of traditional planning that relies on doctors' experience, such as strong subjectivity, low efficiency, and difficulty in restoring the physiological morphology of the joint. It significantly improves the accuracy, repeatability, and personalization of preoperative planning, providing reliable technical support for robot-assisted patellofemoral joint replacement surgery.

[0028] See Figure 1 The specific implementation steps of the automatic planning method for the patellofemoral joint femoral prosthesis in this embodiment are as follows: S100. Acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images. It should be noted that the patient's knee CT and MRI images are collected. Because CT images have high spatial resolution and can clearly show the bone structure, while MRI images have excellent contrast for soft tissues (such as cartilage), in this embodiment, it is necessary to use knee CT images to segment the femoral region and MRI images to segment the femoral cartilage region.

[0029] See Figure 3 and Figure 4 In this embodiment, the method for segmenting the femoral region based on the CT image and identifying key anatomical points on the femur includes: The CT image is segmented using a trained first neural network model, such as a 3D fully convolutional network of nnUNet, and the segmented femoral region is output. The second neural network model identifies key anatomical points on the femoral region in the CT image and outputs the coordinate information of the key anatomical points; wherein, the key anatomical points include the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur.

[0030] In a preferred embodiment, a trained neural network model is used to segment the femoral region in CT images and identify key anatomical points, resulting in the following: Figure 4 and Figure 5 The femoral region and two key anatomical points on the femur are shown in the diagram. The specific implementation steps are as follows: Step 1: Data Preparation and Processing We collected CT scan data from 600 knee joints, and professional physicians marked the femoral region, as well as two key anatomical points: the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur. For the key anatomical point identification task, the coordinates of the marked points are converted into spherical regions, and a spherical mask with a radius of 6mm and the marked points are generated as training labels. Step 2: Network Model and Training Femoral segmentation uses an nnU-Net 3D fully convolutional network as the first neural network model and is trained using the Dice loss function; The identification of key anatomical points can use the same network architecture as a second neural network model, but with a sphere mask as the learning target. After training, the coordinate information of two key anatomical points is obtained by calculating the centroid of the predicted sphere region.

[0031] The above preferred embodiments achieve precise segmentation of the femoral region and accurate identification of key anatomical points. In particular, the method of transforming the identification of key anatomical points into a spherical segmentation task effectively improves the robustness and accuracy of the localization, and provides a reliable anatomical reference benchmark for subsequent prosthesis planning.

[0032] Preferably, when using the second neural network model to identify key anatomical points on the femoral region in the CT image, the key anatomical points are converted into spherical regions with the point as the center and a first preset value as the radius. The method for outputting the coordinate information of the key anatomical points includes: Calculate the coordinates of the centroid of the spherical region and use the coordinates of the centroid as the coordinates of the key anatomical point.

[0033] Understandably, to improve positioning accuracy and robustness, when training the second neural network, the coordinates of the labeled key anatomical points can be converted into a spherical region with the key anatomical point as the center and a first preset value (e.g., 6 voxels) as the radius, which serves as the segmentation label. The goal of the neural network is to predict this spherical region. Therefore, during inference, the centroid of the spherical region output by the neural network is calculated, and its coordinates are used as the final coordinates of the key anatomical points. This transforms the regression problem of key anatomical points into a more stable segmentation problem. This approach can reduce the errors that may be caused by single-point recognition and improve the accuracy of key point positioning. By utilizing the contextual information of the spherical region, it effectively overcomes the positioning deviation caused by image noise or individual differences, and significantly improves the accuracy of key point recognition.

[0034] See Figure 3 In this embodiment, the method for segmenting the femoral cartilage region based on the MRI image includes: The MRI images are segmented using a trained third neural network model (which can also be a 3D fully convolutional network like nnUNet), outputting the segmented femoral cartilage region. To ensure segmentation quality, the collected MRI data should include the complete femoral condyle cartilage portion and undergo uniform data augmentation.

[0035] In a preferred embodiment, a method for automatically segmenting the femoral cartilage region in MRI images using a trained third neural network model can be employed. This third neural network model is specifically optimized for soft tissue segmentation and can accurately distinguish the boundary between cartilage and surrounding tissues, yielding results such as... Figure 3 The segmentation results of the left and right femoral cartilages shown are illustrated below. The specific implementation steps are as follows: Step 1: Data Preparation We collected 3D MRI data of 600 knee joints containing intact femoral condyle cartilage. Professional physicians performed 3D annotation of the femoral cartilage region to form standard segmentation labels. The raw data underwent intensity normalization and unified resolution preprocessing, and data enhancement methods such as rotation and scaling were used. Step 2: Network Model and Training The nnU-Net 3D fully convolutional network is used as the third neural network model. The network contains a 5-level encoder-decoder structure, uses 3×3×3 convolutional kernels, and employs instance normalization and LeakyReLU activation functions. The model training uses a combination of Dice loss and cross-entropy loss as the loss function and is optimized using the Adam optimizer. A five-fold cross-validation strategy is used during training. The final model is an ensemble of the prediction results of the five sub-models.

[0036] The above preferred embodiments achieve automatic and accurate segmentation of femoral cartilage. They employ a specially optimized 3D deep learning segmentation network adapted to the imaging characteristics of cartilage in MRI images. Through large-scale labeled data and effective data augmentation strategies, the integration of the model's generalization ability and prediction mechanism enhances the stability and accuracy of the segmentation results, providing key technical support for a fully automated planning process.

[0037] S200: The femoral region and the femoral cartilage region are fused to generate an initial femoral model including the femoral and cartilage surfaces. The femoral prosthesis model is obtained based on the initial femoral model and the key anatomical points. See Figure 6 In this embodiment, the method for generating an initial femoral model including the femoral bone and cartilage surface includes: A three-dimensional image registration algorithm is used to calculate the spatial transformation relationship between the CT image and the MRI image. Based on the spatial transformation relationship, the coordinate system of the femoral cartilage region is transformed to the first coordinate system of the CT image. In the first coordinate system, the transformed femoral cartilage region is merged with the femoral region to generate an initial femoral model.

[0038] Understandably, this embodiment employs a three-dimensional image registration algorithm (such as a registration method based on mutual information or deep learning) to calculate the spatial transformation matrix from MRI to CT images. This matrix is ​​used to transform the segmented femoral cartilage region to the coordinate system of the CT image. Subsequently, under a unified coordinate system, the transformed cartilage mask and the femoral mask are merged. During the merging process, Boolean operations can be used to ensure a natural transition between cartilage and bone, generating an initial femoral model. This technical solution comprehensively utilizes the skeletal geometric accuracy of CT and the cartilage morphology information of MRI to construct a more complete and realistic joint anatomy model, laying a data foundation for subsequent precise planning.

[0039] Furthermore, methods for obtaining femoral prosthesis models based on initial femoral models and key anatomical points include: The initial femoral model is converted into a three-dimensional surface point cloud; On the three-dimensional surface point cloud, taking the lowest point of the patellar trochlea as the starting point of the path and the midpoint of the intercondylar fossa of the femur as the ending point of the path, the shortest path connecting the starting point and the ending point is calculated using the shortest path algorithm, and the shortest path is defined as the femoral trochlear groove path; the femoral trochlear groove path is of great significance for determining the model and position of the femoral prosthesis, as it represents the sliding trajectory of the patella on the femur.

[0040] The femoral prosthesis model is determined based on the femoral trochlear groove path. Specifically, the femoral prosthesis model can be determined based on the length of the trochlear groove path or the projection length of the trochlear groove on the cross-section.

[0041] Understandably, the initial femoral model is converted into a three-dimensional surface point cloud using algorithms such as Marching Cubes. Starting from the lowest point of the patellar trochlea and ending at the midpoint of the intercondylar fossa, the shortest path connecting the start and end points is calculated using shortest path algorithms such as Dijkstra on the three-dimensional point cloud. This path is the femoral trochlear groove path (Whiteside line). By measuring the length and curvature characteristics of the femoral trochlear groove path, and combining this with parameters such as the width of the intercondylar fossa, the most suitable femoral prosthesis model is selected from a pre-set femoral prosthesis database. The selection process considers the matching degree between the prosthesis size and the patient's anatomical structure, ensuring that the prosthesis can restore the patient's normal knee joint function to the greatest extent. This approach automatically and objectively solves the key biomechanical path through algorithms, replacing the subjectivity and inaccuracy of traditional manual measurements, and providing a clear quantitative basis for prosthesis model selection.

[0042] It should be noted that by calculating the total three-dimensional spatial length of the femoral trochlear groove path (Whiteside line), a core parameter reflecting the developmental size of the patient's femoral trochlear groove is obtained. Matching this length value with a femoral prosthesis size database and selecting the prosthesis model with the closest size range directly reflects the longitudinal anatomical characteristics of the trochlear groove, providing an intuitive quantitative basis for femoral prosthesis size selection. Furthermore, projecting the three-dimensional femoral trochlear groove path onto a transverse plane and calculating its projected length on that plane effectively characterizes the width of the trochlear groove in the coronal plane. Matching this projected length with a prosthesis size database further improves the accuracy of size selection.

[0043] S300. Reconstruct the femoral cartilage in the initial femoral model to generate a target femoral model, and determine the implantation position of the femoral prosthesis based on the target femoral model. In a preferred embodiment, a method for reconstructing the femoral cartilage in an initial femoral model to generate a target femoral model includes: The surface of the initial femoral model is subjected to surface interpolation and smoothing to reconstruct a continuous femoral cartilage surface, thereby obtaining the target femoral model.

[0044] In another preferred embodiment, the method for reconstructing the femoral cartilage in the initial femoral model to generate the target femoral model further includes: Obtain a femoral cartilage model of the non-operating knee joint of the patient; The femoral cartilage model of the non-operational knee joint is mirrored and flipped along the sagittal plane to obtain the first femoral cartilage model. The first femoral cartilage model is non-rigidly registered with the initial femoral model on the surgical side; The registered cartilage model is used as the reconstructed femoral cartilage surface on the surgical side to generate the target femoral model.

[0045] It should be further noted that, due to cartilage wear in the patient, the initial femoral model exhibits pathological anatomy. To achieve functional reconstruction, cartilage restoration is necessary. This can be achieved through the following two preferred implementation methods: (1) Reconstruction based on surface interpolation: Based on the existing three-dimensional morphology of the cartilage surface of the initial femoral model, the geometric missing parts are filled by surface interpolation algorithms (such as radial basis function interpolation, NURBS), which can generate a smooth and continuous surface. The smoothing process can use the Laplacian operator to reduce the noise and irregularity generated during segmentation and reconstruction, and generate a continuous and smooth cartilage surface. Thus, based on the patient's own residual cartilage information, the articular surface that conforms to the physiological curvature can be intelligently inferred and reconstructed, providing an ideal matching target for prosthesis planning. (2) Reconstruction based on contralateral mirror registration: Obtain the healthy femoral cartilage model of the non-operational side of the patient's knee joint, i.e. the contralateral knee joint with worn cartilage. Mirror the model along the sagittal plane, i.e., flip the femoral cartilage model of the healthy knee joint in the left-right direction while keeping the front-back and up-down directions unchanged. This yields the ideal cartilage model of the operation side, i.e. the first femoral cartilage model. Subsequently, the ideal cartilage model is elastically registered with the initial femoral model of the operation side. This allows the ideal cartilage model to elastically adapt to the skeletal geometry of the femur on the operation side while maintaining its smooth shape. The registered cartilage model is the reconstructed cartilage surface. This approach utilizes the high symmetry of the bilateral joints of the human body to more accurately restore the patient's original anatomical structure and avoids biomechanical abnormalities caused by matching the worn shape.

[0046] It is understood that flexible registration (non-rigid registration) refers to a three-dimensional image registration technique, which differs from traditional rigid registration that only involves rotation and translation transformations. It can achieve more complex local deformations to accurately align two anatomical structures with different morphological characteristics. Specifically, in the cartilage reconstruction process of this embodiment, the technical implementation of flexible registration includes the following key steps: Transformation model selection: A free-form deformation model based on B-splines is adopted as the mathematical basis for elastic transformation. This model achieves a smooth, local deformation field by setting a control point grid in three-dimensional space and adjusting the displacement of the control points. Similarity measurement: Normalized cross-correlation or mutual information is used as a similarity measure to quantify the degree of anatomical feature matching between the first femoral cartilage model to be registered and the initial femoral model on the surgical side; Regularization constraint: Introducing a bending energy constraint term ensures the smoothness and physical rationality of the deformation field, prevents unnatural and drastic deformation, and thus maintains the anatomical rationality of the reconstructed cartilage surface; Optimization strategy: A multi-resolution optimization strategy is adopted to gradually optimize the deformation parameters from coarse to fine, which not only ensures the global convergence of registration, but also achieves accurate alignment of local details. Through flexible registration, this embodiment can smoothly adapt to the individualized geometric features of the femoral bone on the surgical side while maintaining the smooth shape and anatomical characteristics of the healthy cartilage model itself. It makes full use of the complete anatomical information of the contralateral healthy cartilage and respects the actual morphological differences of the surgical bone, thereby achieving personalized and high-precision cartilage reconstruction. This provides a reliable anatomical basis for subsequent prosthesis planning, overcomes the limitation of rigid registration in handling local morphological differences, and significantly improves the accuracy and clinical applicability of cartilage reconstruction.

[0047] Using any of the above methods, a target femoral model for prosthesis planning is finally generated, which represents the ideal shape of the joint in an unworn state.

[0048] In this embodiment, the method for determining the implantation position of the femoral prosthesis based on the above-described target femoral model includes: Using the key anatomical points and the femoral trochlear groove path as a reference, the preset reference points on the femoral prosthesis are aligned with the key anatomical points on the femoral region, and the central axis of the femoral prosthesis is aligned with the femoral trochlear groove path to determine the initial position of the femoral prosthesis. This alignment method can ensure that the basic position of the prosthesis conforms to the patient's anatomical structure. Based on the initial position, the femoral prosthesis is rotated along the central axis to adjust the inversion and valgus angles of the femoral prosthesis, thereby obtaining the implantation position of the femoral prosthesis; Furthermore, at least one point is pre-defined on the surface edge line of the femoral prosthesis. Then, the femoral prosthesis is rotated along its central axis to minimize the distance between the point on the femoral prosthesis edge line and the surface of the target femoral model. This yields the adjusted inversion / valgus angle of the femoral prosthesis, thus determining the implantation position of the femoral prosthesis. This optimization method ensures optimal fit between the femoral prosthesis and the patient's bone, reducing the risk of prosthesis loosening.

[0049] Preferably, according to the above technical solution, the process of determining the femoral prosthesis implantation position mainly consists of two steps: coarse registration and fine optimization. (1) Coarse registration to determine the initial pose: Based on the key anatomical points and the femoral trochlear groove path, the preset reference points on the femoral prosthesis model are aligned with the corresponding key anatomical points on the femur. The preset reference points here refer to the two points on the femoral prosthesis that correspond to the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur. At the same time, the central axis of the femoral prosthesis model is aligned with the calculated femoral trochlear groove path to determine the initial pose of the femoral prosthesis.

[0050] (2) Fine-tuning the inversion and valgus angles: Based on the initial pose, the only remaining determination of the inversion and valgus angles of the femoral prosthesis around its central axis (i.e., the Whiteside line) is to be made. In this embodiment, an optimization target is defined, which is to minimize the sum of the Euclidean distances from the two preset lower edge points of the femoral prosthesis to the cartilage surface of the target femoral model, such as... Figure 11 As shown, the optimal rotation angle is searched by an optimization algorithm (such as gradient descent), and the pose at this angle is used as the final implantation pose. This fine-tuning strategy ensures that the lower edge of the femoral prosthesis and the reconstructed cartilage surface achieve maximum conformal contact and smooth transition through mathematical optimization, effectively avoiding edge loading, suspension or over-embedding of the femoral prosthesis, and optimizing the initial stability and long-term mechanical environment of the prosthesis.

[0051] S400. Based on the implantation position, determine the femoral osteotomy area that matches the femoral prosthesis. The specific implementation method is as follows: In three-dimensional space, the intersection of the femoral prosthesis with the target femoral model is calculated when the femoral prosthesis is in the implantation pose state; the region defined by the intersection is taken as the femoral osteotomy region that matches the femoral prosthesis.

[0052] It should be noted that after determining the final implantation position of the femoral prosthesis, the geometric intersection of the femoral region (i.e., the patient's actual bone) in the femoral prosthesis model and the initial femoral model is calculated in three-dimensional space. This intersection physically represents the volume of bone that must be removed from the patient's femur to install the femoral prosthesis. The system automatically defines this region as the femoral osteotomy region and can directly output it to guide the surgical robot or navigation system to perform osteotomy. This method can accurately determine the range of bone tissue to be removed, ensuring that the prosthesis can be stably fixed on the femur after implantation. It achieves precision and minimization of osteotomy, preserving the patient's original bone to the greatest extent.

[0053] Please see Figures 3-11 The implementation steps of this embodiment are illustrated below with specific examples: Step 1: Train a neural network model to segment the cartilage of the femur in MRI images, obtaining results such as... Figure 3 The segmentation results of the left and right femoral cartilages are shown below, with specific steps as follows; (1) Data collection Knee MRI data were collected. While MRI data from the femoral head to the femoral condyle is generally not collected in practical applications, it is essential to ensure that the collected images include the cartilaginous portion of the femoral condyles in both legs. Furthermore, in addition to 3D MRI data, transverse, coronal, and sagittal planes each have a significant slice thickness. Therefore, the distribution of various data types needs to be uniform during dataset creation. Consequently, in this example, a total of 600 MRI images were collected.

[0054] (2) Data labeling The portion of the femoral cartilage region is marked in the image data collected above; (3) Establish a 3D fully convolutional neural network for femoral cartilage segmentation. In this example, nnUNet is used to predict the segmentation of MRI images. (4) The collected MRI images were used to train the nnUNet neural network to obtain a neural network model for segmenting the femoral cartilage region.

[0055] Step 2: Train a neural network model to segment the femur in CT images, obtaining results such as... Figure 4 The segmentation results of the femoral region shown are implemented using the following steps: (1) Data collection CT data was collected, specifically knee CT images of size 512×512×N, where N represents the number of CT slices and 512×512 represents the size of each CT slice. In actual use, complete CT data from the femoral head to the femoral condyle is generally not collected, but it is necessary to ensure that the collected CT images include the femoral condyle on the operated side. In this example, a total of 600 CT images were collected. (2) Data labeling The femoral region was marked in the collected CT image data; (3) Establish a 3D fully convolutional neural network for femoral segmentation, taking nnUNet as an example, to predict the segmentation of CT images; (4) Use the collected data to train the nnUNet neural network to obtain a neural network model for segmenting the femoral region.

[0056] Step 3: Train a neural network model to identify the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur, such as... Figure 5 As shown, the specific implementation steps are as follows: (1) Data collection CT data were collected, specifically knee joint CT images of size 512×512×N, where N represents the number of CT slices and 512×512 represents the size of each CT slice. In actual use, complete CT data from the femoral head to the femoral condyle is generally not collected, but it is necessary to ensure that the image includes the complete femoral condyle of the operated side. In this embodiment, a total of 600 CT images were collected. (2) Data labeling In the collected CT images, the lowest point of the patellar trochlea in the femur and the midpoint of the intercondylar fossa of the femur were marked, such as... Figure 5 As shown; (3) Data processing Read the CT image and convert it into a 3D matrix. Generate a zero-based 3D matrix with the same size as the CT image. Using the coordinates of the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur as the center of a sphere, render a small sphere with a radius of 6. Set all pixel values ​​within the sphere to 1. Then, reconstruct the 3D matrix into an image in volume coordinates and use this image as the image label, such as... Figure 5 As shown; (4) The collected CT image data was used to train the nnUNet neural network to obtain a neural network model for segmenting the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur; (5) Since the location information of the two key anatomical points in the training set exists in the form of small balls, when using the network model trained in step (4) for inference, the result also exists in the form of small balls. By obtaining the centroid of each small ball, the coordinate information of the two key anatomical points can be obtained.

[0057] Step 4: Unify the coordinates of MRI and CT images, and merge the segmented femoral cartilage region and femoral region; Unifying the coordinates of MRI and CT images can be achieved through a three-dimensional image registration algorithm. This algorithm can be based on transformation space, similarity measurement, search strategy, or deep learning. This invention does not further limit the three-dimensional image registration algorithm used, as long as it can unify the coordinate systems of the two images.

[0058] The registration results of CT and MRI images are converted into a transformation matrix, which is then applied to CT and MRI segmentation, such as... Figure 6 The image shows the merged femur and femoral cartilage, with the green area representing the femur and the red area representing the cartilage. Figure 6 As can be seen, there are missing portions of the cartilage, which means that wear and tear has occurred. Step 5: Calculate the femoral trochlear groove path (whiteside line) based on the femoral segmentation, the lowest point of the patellar trochlea, and the midpoint of the intercondylar fossa of the femur; like Figure 7 As shown, the segmented femoral region is used to calculate the femoral surface point cloud using a rasterization algorithm. This algorithm can be the Marching cube algorithm or other methods that can obtain the surface point cloud through segmentation. Starting from the lowest point of the patellar trochlea and ending at the intercondylar fossa of the femur, the shortest path is found in the three-dimensional point cloud obtained in the above steps. This shortest path is the path of the trochlear groove of the femur. The algorithm for finding this shortest path can be Dijkstra's algorithm, Bellman-Ford algorithm, or other methods that can solve the shortest path. This invention does not further limit the specific shortest path algorithm.

[0059] Step 6: Reconstruct the femoral cartilage that has been worn down due to degenerative changes or other reasons; Patients experience cartilage wear due to degenerative changes or other reasons. However, during the femoral prosthesis implantation process, it is necessary to restore the patient's cartilage to its pre-wear condition as much as possible using the femoral prosthesis. Therefore, it is necessary to simulate and restore the worn cartilage first. There are two approaches to cartilage restoration. One approach is to reconstruct the cartilage by smoothing the surrounding cartilage. This involves using three-dimensional surface interpolation and smoothing to reconstruct a continuous cartilage surface that conforms to physiological morphology for the pathological cartilage defect area in the femoral model, based on the intact cartilage contour around it. Another approach is to use a femoral cartilage model from the contralateral side of the cartilage wear side, i.e., the non-operational side femoral cartilage model, which is flipped along the sagittal plane and elastically registered before restoration. The elastic registration method can be based on artificial intelligence.

[0060] Step 7: Determine the implant type; In this example, the femoral prosthesis size is determined by the length of the whiteside line. Alternatively, the size can be determined by the length of the whiteside line's projection onto the cross-section. Specifically, as shown below... Figure 9 As shown; Step 8: Register the femur and femoral prosthesis using coarse registration; (1) Merge the reconstructed femoral cartilage and femur after the cartilage wear, and convert the merged image into a three-dimensional point cloud using a rasterization algorithm. This point cloud is the merged femoral surface, such as... Figure 8 As shown: (2) Obtain prosthesis data suitable for the femoral prosthesis model. This prosthesis data is also in the form of point cloud data, such as... Figure 9 As shown: (3) Using the lowest point of the patellar trochlea and the corresponding point of the midpoint of the intercondylar fossa of the femur on the femoral prosthesis as positioning points, and aligning the whiteside line with the midline of the femoral prosthesis, the initial pose of the femoral prosthesis is obtained. Coarse point cloud registration is then performed. At this point, all angles except for the varus / valgus angle of the femoral prosthesis can be determined, such as... Figure 10 As shown, in Figure 10 In the diagram, the green dot represents a point on the femoral surface, and the red dot represents a point on the femoral prosthesis corresponding to the green dot. This shows that the femoral prosthesis and femoral surface cannot transition smoothly in the initial position. Therefore, it is necessary to adjust the varus / valgus angle of the femoral prosthesis based on the initial position to achieve a better match between the femoral prosthesis and the femoral surface. Step 9: Since the universal prosthesis cannot achieve the same state as the patient's femur and cartilage before wear, the position of the prosthesis is finely adjusted according to the actual clinical application needs, that is, the inversion and valgus angle of the femoral prosthesis is finely optimized and adjusted. In this example, such as Figure 11As shown in the left figure, the center points of the two lower edges of the femoral prosthesis are taken as hard requirement points. That is, the center points of the lower edges of the femoral prosthesis on both sides must smoothly transition with the femoral surface. Therefore, the constraint condition that can be determined is to adjust the inversion and valgus angle of the femoral prosthesis to achieve the final state. This constraint condition can be simplified to solving for the sum of the distances from the two lower edges of the femoral prosthesis to the closest point on the femoral point cloud. In other words, it makes... Figure 10 The green and red dots in the image should overlap as much as possible to obtain the image shown. Figure 11 The final implantation pose is shown in the image on the right: Step 10: Automatically generate an osteotomy plan based on the femoral prosthesis position and femoral position; After confirming the final implantation position of the femoral prosthesis, the overlapping part of the femoral prosthesis and the femur is the osteotomy area. In other words, the intersection of the femoral prosthesis with the target femoral model in the implantation position is calculated, and the area defined by the intersection is taken as the femoral osteotomy area that matches the femoral prosthesis.

[0061] Understandably, traditional surgical methods rely heavily on the surgeon's anatomical knowledge and proficiency in operating surgical instruments. The intraoperative planning of the femoral prosthesis and osteotomy, combined with real-time fluoroscopic imaging, presents a challenge to both the surgeon's physical strength and professional skills. This invention, however, employs robot-assisted surgery and preoperative osteotomy planning, which not only reduces the surgeon's workload during surgery but also standardizes and replicates the surgical procedure. Furthermore, the automated planning method improves preoperative planning efficiency and reduces reliance on the surgeon's experience.

[0062] It should be noted that this embodiment provides an automated planning method for femoral prostheses in patellofemoral joint surgery. By integrating multimodal imaging data from knee CT and MRI, it achieves full automation from image segmentation, model fusion, cartilage function reconstruction to prosthesis planning and osteotomy scheme generation. This invention first precisely segments the femoral structure and identifies key anatomical points based on CT images, and simultaneously segments the cartilage region using MRI images. An initial femoral model including bone and cartilage is constructed through coordinate system one and data fusion. Then, pathological cartilage defects are digitally reconstructed to generate a target femoral model representing the morphology of a healthy joint. Based on this, the matching prosthesis model and implantation position are automatically determined. Finally, a precise osteotomy area is automatically generated based on the geometric relationship between the prosthesis and the bone model. This effectively overcomes the shortcomings of traditional planning methods that rely on physician experience, such as high subjectivity, low efficiency, and difficulty in restoring the physiological morphology of the joint. It significantly improves the accuracy, repeatability, and personalization of preoperative planning, providing reliable technical support for robot-assisted patellofemoral joint replacement surgery.

[0063] Example 2 Please see Figure 2This embodiment provides an automatic planning system for patellofemoral joint femoral prostheses, the system including an image processing module 100, a model construction module 200, a prosthesis planning module 300, and an osteotomy planning module 400.

[0064] The image processing module 100 is used to acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images. The model building module 200 is used to fuse the femoral region and the femoral cartilage region to generate an initial femoral model containing the femoral and cartilage surfaces, and to obtain the femoral prosthesis model based on the initial femoral model and the key anatomical points. The prosthesis planning module 300 is used to reconstruct the femoral cartilage in the initial femoral model, generate a target femoral model, and determine the implantation pose of the femoral prosthesis based on the target femoral model. The cartilage reconstruction process considers the cartilage thickness distribution in both the diseased and healthy areas of the patient. An interpolation algorithm is used to restore the ideal cartilage morphology. This module, based on the target femoral model, determines the implantation position of the femoral prosthesis. In the position determination process, the system comprehensively considers three aspects: mechanical axis alignment, rotational alignment, and joint space balance. Mechanical axis alignment ensures that the patient's lower limb force line is restored to normal after prosthesis implantation. Rotational alignment makes the anterior edge of the prosthesis parallel to the anterior femoral cortex to avoid malrotation of the prosthesis. Joint space balance is achieved by adjusting the position of the prosthesis to make the flexion-extension gap uniform and ensure postoperative joint mobility.

[0065] The osteotomy planning module 400 is used to determine the femoral osteotomy area that matches the femoral prosthesis based on the implantation pose.

[0066] In a preferred embodiment, the image processing module further includes an image preprocessing unit for denoising, enhancing, and standardizing the acquired CT and MRI images. The denoising process uses an adaptive filtering algorithm to effectively remove random noise from the images. The enhancement process improves image contrast through histogram equalization. The standardization process converts images acquired by different devices to a uniform grayscale range and resolution to ensure consistency in subsequent processing.

[0067] In a preferred embodiment, the prosthesis planning module also includes a prosthesis library that stores 3D models and parameter information of various specifications and models of patellofemoral joint femoral prostheses. The prostheses in the library are categorized according to attributes such as manufacturer, material, and size, facilitating rapid retrieval and matching by the system. When new prosthesis models become available, the administrator can update the prosthesis library to ensure the system always uses the latest prosthesis information.

[0068] In another preferred embodiment, the osteotomy planning module further includes an osteotomy simulation unit for simulating the osteotomy process in three-dimensional space and calculating the volume and shape of the bone tissue after osteotomy. This unit can evaluate the rationality of the osteotomy plan, such as whether the amount of osteotomy is too large or too small, whether the osteotomy surface is flat, etc. If a problem is found, the system will automatically adjust the osteotomy plan until the best effect is achieved.

[0069] This embodiment also includes a user interface, through which doctors can view and adjust the prosthesis planning scheme automatically generated by the system. The interface displays the patient's femoral model, prosthesis position and osteotomy plane in a three-dimensional visualization. Doctors can adjust the position and direction of the prosthesis by mouse or touch operation, and the system will update the osteotomy plan and expected results in real time.

[0070] It should be noted that this embodiment provides an automated planning system for patellofemoral prostheses, including an image processing module 100, a model building module 200, a prosthesis planning module 300, and an osteotomy planning module 400. It constructs a complete, closed-loop automated preoperative planning process for patellofemoral prostheses, from multimodal data input to final surgical plan output. This significantly improves the efficiency, accuracy, and repeatability of preoperative planning, effectively solving the technical problems of strong subjectivity, low efficiency, and difficulty in restoring joint physiological morphology inherent in traditional planning that relies on physician experience. It significantly improves the accuracy, repeatability, and personalization of preoperative planning, providing reliable technical support for robot-assisted patellofemoral joint replacement surgery.

[0071] In a preferred embodiment, this application also provides an electronic device, the electronic device comprising: The computer device includes a memory and a processor. The memory stores computer-readable instructions that, when executed by the processor, implement the automatic planning method for the patellofemoral joint femoral prosthesis. This computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the method of the present invention.

[0072] The electronic device in this embodiment is also equipped with a high-resolution display screen for displaying CT images, MRI images, and 3D reconstruction models, facilitating surgical planning and evaluation by doctors. The device also has an input interface that supports multiple input methods such as mouse, keyboard, and stylus, enabling doctors to perform interactive operations.

[0073] The automatic planning method for the patellofemoral prosthesis executed by the electronic device is the same as the method described in Example 1, including steps such as acquiring and segmenting the patient's knee CT and MRI images, fusing the femoral region and the femoral cartilage region to generate an initial femoral model, reconstructing the femoral cartilage to generate a target femoral model, determining the implantation pose of the femoral prosthesis, and determining the femoral osteotomy area.

[0074] In a preferred embodiment, the electronic device is also equipped with a medical image database interface, which can directly obtain the patient's CT and MRI data from the hospital's imaging system, simplifying the data import process. Simultaneously, the device also has a network communication module, supporting remote collaboration capabilities, enabling multiple doctors to participate in the surgical planning process simultaneously.

[0075] This invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0076] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0077] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0078] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. An automatic planning method for patellofemoral joint femoral prostheses, characterized in that, include: S100. Acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images. S200: The femoral region and the femoral cartilage region are fused to generate an initial femoral model including the femoral and cartilage surfaces. The femoral prosthesis model is obtained based on the initial femoral model and the key anatomical points. S300. Reconstruct the femoral cartilage in the initial femoral model to generate a target femoral model, and determine the implantation position of the femoral prosthesis based on the target femoral model. S400. Based on the implantation position, determine the femoral osteotomy area that matches the femoral prosthesis.

2. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 1, characterized in that, The method for segmenting the femoral region based on the CT image and identifying key anatomical points on the femur includes: The CT image is segmented using a trained first neural network model, and the segmented femoral region is output. The second neural network model identifies key anatomical points on the femoral region in the CT image and outputs the coordinate information of the key anatomical points. The key anatomical points include the lowest point of the patellar trochlea and the midpoint of the intercondylar fossa of the femur.

3. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 2, characterized in that, When using a second neural network model to identify key anatomical points on the femoral region in the CT image, the key anatomical points are converted into spherical regions with the point as the center and a first preset value as the radius. The method for outputting the coordinate information of the key anatomical points includes: Calculate the coordinates of the centroid of the spherical region and use the coordinates of the centroid as the coordinates of the key anatomical point.

4. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 1, characterized in that, The method for segmenting the femoral cartilage region based on the MRI image includes: The MRI image is segmented using a trained third neural network model, and the segmented femoral cartilage region is output.

5. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 1, characterized in that, The method for fusing the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces includes: A three-dimensional image registration algorithm is used to calculate the spatial transformation relationship between the CT image and the MRI image. Based on the spatial transformation relationship, the coordinate system of the femoral cartilage region is transformed to the first coordinate system of the CT image. In the first coordinate system, the transformed femoral cartilage region is merged with the femoral region to generate an initial femoral model.

6. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 2, characterized in that, The method for obtaining the femoral prosthesis model based on the initial femoral model and the key anatomical points includes: The initial femoral model is converted into a three-dimensional surface point cloud; On the three-dimensional surface point cloud, the lowest point of the patellar trochlea is taken as the starting point of the path, and the midpoint of the intercondylar fossa of the femur is taken as the ending point of the path. The shortest path algorithm is used to calculate the shortest path connecting the starting point and the ending point, and the shortest path is defined as the femoral trochlear groove path. The femoral prosthesis model is determined based on the described femoral trochlear groove path.

7. The automatic planning method for patellofemoral prostheses according to claim 6, characterized in that, The method for reconstructing the femoral cartilage in the initial femoral model to generate the target femoral model includes: The surface of the initial femoral model is subjected to surface interpolation and smoothing to reconstruct a continuous femoral cartilage surface, thereby obtaining the target femoral model.

8. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 6, characterized in that, The method for reconstructing the femoral cartilage in the initial femoral model to generate the target femoral model further includes: Obtain a femoral cartilage model of the non-operating knee joint of the patient; The femoral cartilage model of the non-operational knee joint is mirrored and flipped along the sagittal plane to obtain the first femoral cartilage model. The first femoral cartilage model is non-rigidly registered with the initial femoral model on the surgical side; The registered cartilage model is used as the reconstructed femoral cartilage surface on the surgical side to generate the target femoral model.

9. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 2, characterized in that, The method for determining the implantation position of the femoral prosthesis based on the target femoral model includes: Using the key anatomical points and the femoral trochlear groove path as a reference, the preset reference points on the femoral prosthesis are aligned with the key anatomical points on the femoral region, and the central axis of the femoral prosthesis is aligned with the femoral trochlear groove path to determine the initial position of the femoral prosthesis. Based on the initial position, the femoral prosthesis is rotated along the central axis to adjust the inversion and valgus angles of the femoral prosthesis, thereby obtaining the implantation position of the femoral prosthesis.

10. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 9, characterized in that, The step of rotating the femoral prosthesis along the central axis to adjust the varus / valgus angle of the femoral prosthesis, thereby obtaining the implantation position of the femoral prosthesis, includes: At least one point is preset on the surface edge line of the femoral prosthesis. Then, the femoral prosthesis is rotated along the central axis to minimize the distance between the point on the edge line of the femoral prosthesis and the surface of the target femoral model, thereby obtaining the adjusted inversion and eversion angle of the femoral prosthesis, and thus obtaining the determined implantation posture of the femoral prosthesis.

11. The automatic planning method for patellofemoral joint femoral prosthesis according to claim 1, characterized in that, The step of determining the femoral osteotomy region matching the femoral prosthesis based on the implantation position includes: In three-dimensional space, the intersection of the femoral prosthesis with the target femoral model is calculated when the implantation pose is in the specified position. The region defined by the intersection is taken as the femoral osteotomy region that matches the femoral prosthesis.

12. An automatic planning system for a patellofemoral joint femoral prosthesis, characterized in that, include: Image processing module: used to acquire CT and MRI images of the patient's knee joint, segment the femoral region based on the CT images, identify key anatomical points on the femur, and segment the femoral cartilage region based on the MRI images; Model building module: used to fuse the femoral region and the femoral cartilage region to generate an initial femoral model including the femoral and cartilage surfaces, and to obtain the femoral prosthesis model based on the initial femoral model and the key anatomical points; Prosthesis planning module: used to reconstruct the femoral cartilage in the initial femoral model, generate a target femoral model, and determine the implantation pose of the femoral prosthesis based on the target femoral model; Osteotomy planning module: used to determine the femoral osteotomy area that matches the femoral prosthesis based on the implantation pose.

13. An electronic device, characterized in that, include: Memory; The processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the automatic planning method for a patellofemoral prosthesis according to any one of claims 1 to 11.

Images