Hip joint prosthesis three-dimensional registration method based on CT image

Abstract

The invention discloses a hip joint prosthesis three-dimensional registration method based on a CT image, and relates to the technical field of hip joint prosthesis registration, and the method comprises the steps: S1, reconstructing a patient femur three-dimensional structure based on the CT image through 3D slicer software, and carrying out the segmentation and extraction of two-dimensional features of a target part of the three-dimensional structure; by constructing a complete technical process of three-dimensional reconstruction and two-dimensional feature extraction, precise positioning of anatomical mark points, comprehensive analysis of the proximal femur form and registration of a prosthesis size algorithm, the problems in the prior art are effectively solved, and by combining the ICP algorithm with a mark point positioning mode of global and local registration, the positioning accuracy of the femur is improved. Dependence on manual operation and empirical judgment is eliminated, the defect that individual anatomical difference adaptation is insufficient is overcome, accurate mapping of mark points is achieved, the medullary cavity form analysis method combining multi-plane intersection with two-dimensional and three-dimensional feature fusion makes up for the defect that in the prior art, medullary cavity form analysis is not comprehensive, and accurate data support is provided for registration.

Description

Technical Field

[0001] This invention relates to the field of hip joint prosthesis registration technology, specifically a 3D registration method for hip joint prostheses based on CT images. Background Technology

[0002] The hip joint is a core weight-bearing joint in the human body, connecting the trunk and lower limbs. It is a key structure for enabling motor functions such as walking, standing, running, and jumping. Its anatomical integrity and functional stability directly determine the body's mobility. When the hip joint is damaged or loses its function due to trauma, degeneration, or disease, hip replacement surgery is the main treatment for restoring its function. 3D registration of the hip prosthesis is the core of hip replacement surgery. It refers to the process of accurately matching the patient's proximal femur anatomy with the shape and size parameters of the hip prosthesis to determine the optimal prosthesis model, installation position, and orientation. The accuracy of this process directly affects the fit and stability of the prosthesis after implantation, as well as the stress distribution between the prosthesis and the femur, thus determining the patient's postoperative joint function recovery and prosthesis lifespan. It is also key to reducing postoperative complications such as prosthesis loosening, dislocation, and infection.

[0003] With its high resolution and high definition imaging advantages, CT technology can clearly present the anatomical details of the proximal femur, including the morphological features, size data and spatial relationships of key parts such as the femoral head, femoral neck, lesser trochanter and femoral medullary cavity. This provides accurate and comprehensive anatomical data support for 3D registration, becoming an important technical foundation for achieving precise prosthesis fitting. Compared with traditional imaging methods, it can better meet the needs of 3D registration for refined anatomical data.

[0004] However, existing 3D registration techniques for hip prostheses still have certain shortcomings. Traditional registration relies on manual measurement of anatomical parameters and experience in prosthesis selection, which is highly subjective and prone to errors. Some image-based registration methods lack a systematic anatomical landmark localization process, and the connection between global and local registration is not smooth, resulting in insufficient landmark mapping accuracy. The analysis of femoral medullary cavity morphology often relies solely on 3D or 2D features, without fully integrating the advantages of multiple algorithms, making it difficult to adapt to individual anatomical differences in patients. Ultimately, this results in poor prosthesis registration accuracy, directly affecting surgical outcomes and increasing the risk of postoperative complications. Therefore, developing a 3D registration method for hip prostheses based on CT images is of great significance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a 3D registration method for hip joint prostheses based on CT images. This method solves the problems of existing technologies by constructing a complete technical process: 3D reconstruction and 2D feature extraction, precise localization of anatomical landmarks, comprehensive analysis of proximal femoral morphology, and prosthesis size algorithm registration. It achieves precise landmark mapping through the ICP algorithm combined with global and local registration for landmark localization. It provides accurate data support for registration through a multi-plane intersection combined with 2D and 3D feature fusion for medullary cavity morphology analysis. Finally, it achieves precise matching between the prosthesis and the patient's femoral anatomy through the fusion of Jaccard and ORB adaptive weighting algorithms and the minimum distance selection principle, thus improving registration accuracy and efficiency.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a 3D registration method for hip joint prostheses based on CT images, the method comprising the following steps:

[0007] S1. Using 3D slicer software, the 3D structure of the patient's femur is reconstructed based on CT images, and the 2D features of the target part of the 3D structure are segmented and extracted.

[0008] S2. Perform anatomical landmark localization, import the 3D mesh of the patient's femur and the template bone and the landmarks of the template bone and realign them, perform initial global registration through the ICP algorithm to determine the initial landmarks, and then perform local registration for the femoral head, femoral neck and lesser trochanter area to obtain the precisely mapped landmarks.

[0009] S3. Based on the mapping of landmark points, analyze the proximal femoral morphology to determine the femoral head diameter, femoral head center, femoral neck axis, femoral medullary cavity morphology, proximal femoral axis, femoral offset, and femoral neck axis angle.

[0010] S4. Perform femoral stem prosthesis size registration algorithm, compare the femoral medullary cavity morphology of the patient with the femoral prosthesis morphology in the training database, and complete the 3D coordinate alignment by combining the comprehensive calculation of 2D features to select the best-fitting femoral stem prosthesis.

[0011] Furthermore, step S2, when locating anatomical landmarks, includes the following steps:

[0012] The position and orientation relationship between the template bone and the patient's femur are observed through the visualization and interactive function of the 3D slicer. If the difference is large, the registration results of 2D segmentation features are combined with manual fine-tuning to make the two initially aligned.

[0013] During the initial global registration, the geometric data of the template bone and the patient's femur are input according to the operation logic of the ICP algorithm. The template model is rotated, translated and scaled, and the geometric differences between the two are continuously calculated and the transformation parameters are iteratively adjusted.

[0014] Identify the key corresponding regions around the femoral head, femoral neck, and lesser trochanter. Recognize region boundaries based on skeletal anatomical features. Use a registration algorithm based on distance and geometric features to match vertices in local regions. During the matching process, calculate the local registration error using the following formula: ,in, This represents the total local registration error. This represents the number of vertices in the local region. The coordinates of the vertex of the local area of ​​the femur in the patient. Register the coordinates of the vertices in the corresponding region of the template bone. The vertex weight coefficient. The complexity of the anatomical structure of the region where the vertex is located is determined by statistical analysis of the vertex contribution of successful registration cases in the training set. Vertices in key parts of the anatomical structure correspond to higher weight coefficients, ultimately forming a precise mapping of the marker point.

[0015] Furthermore, step S3, in determining the morphology of the patient's femoral medullary cavity, includes the following steps:

[0016] Starting from a position below the lesser trochanter of the patient's femur, the multi-plane intersection is controlled according to a preset rule to obtain a continuous femoral medullary cavity boundary contour, and each contour is presented in the form of 3D view and 2D view respectively;

[0017] Using the Jaccard algorithm and the ORB algorithm, the boundary contours in the 2D view are comprehensively calculated according to the adaptive weight allocation rule, and the contour comprehensive features are obtained by fusing them using the following formula: ,in, For contour comprehensive feature values, Contour similarity features calculated by the Jaccard algorithm. Features for feature point matching calculated by the ORB algorithm. For adaptive weighting coefficients, The matching confidence is dynamically determined by the matching confidence of the two algorithms. The matching confidence is calculated by the number of matching point pairs output by each algorithm and the matching consistency test results. The algorithm with more matching point pairs and higher consistency corresponds to a higher weight coefficient. Based on this comprehensive feature, 2D features are extracted and 3D features are enhanced.

[0018] Perform geometric analysis on each boundary profile to determine its inscribed circle, and record the center coordinates and radius data of each inscribed circle;

[0019] Connect the centers of all inscribed circles, and obtain the bisector of the line through geometric calculations. Define the bisector as the proximal axis of the diseased femur.

[0020] Furthermore, step S4, when performing femoral stem prosthesis size registration algorithm, includes the following steps:

[0021] The femoral stem prosthesis morphology data in the training database were retrieved, and the geometric features of the medullary cavity of the patient's femur were compared with the prosthesis morphology data one by one. The implantation process of the femoral stem prosthesis in the femoral medullary cavity was simulated to determine the appropriate placement position and calculate the relative distance between the center of the prosthesis head and the center of the patient's femoral head.

[0022] For the 2D features extracted from the segmentation, they are input into the fusion calculation model of the Jaccard algorithm and the ORB algorithm. The comprehensive calculation is performed according to the adaptive weights to obtain the 2D registration results from the two dimensions of area matching degree and angle consistency, and complete the alignment of the 3D coordinates.

[0023] For each candidate femoral stem prosthesis, geometric parameters were calculated, including the center and radius of the circumcircle of the prosthesis outline. Combined with previously calculated relative distance data, the prosthesis fit score was calculated using the following formula: ,in, Score the fit of the implant. This refers to the relative distance between the center of the prosthesis head and the center of the patient's femoral head. The similarity of the morphology of the prosthesis and the femoral medullary cavity. This is the scoring weighting coefficient. Based on clinical follow-up data, the influence weights of different fitting parameters on the long-term stability of the prosthesis were analyzed. The contribution ratios of distance factors and morphological matching factors in a large number of postoperative cases were statistically analyzed to select the prosthesis with the highest fitting score as the best fitting prosthesis.

[0024] Furthermore, in step S1, when performing 2D feature segmentation and extraction on the 3D structural target area, the target area includes the femoral head region, femoral neck region, region around the lesser trochanter, and the projection region corresponding to the femoral medullary cavity. During the segmentation process, the edge detection function of the 3D slicer software is used to enhance the boundary distinction between the target area and the surrounding tissues. The segmentation operation follows the process of first contour recognition and then precise extraction to obtain a complete 2D feature image.

[0025] Furthermore, during the initial global registration in step S2, the ICP algorithm optimizes the distance error between the template marker point and the patient's femoral surface point by iteratively calculating and continuously adjusting the three transformation parameters—rotation angle, translation vector, and scaling ratio—until the distance error reaches a preset threshold. The patient's femoral surface point determined at this point is the initial marker point. During the iteration process, the distance error value after each parameter adjustment is recorded.

[0026] Furthermore, in step S3, when determining the diameter and center of the femoral head of the patient, the 3D grid point cloud data corresponding to the femoral head is first denoised to remove abnormal discrete points. Then, the least squares method is used to perform sphere fitting operation on the denoised point cloud data. The diameter of the fitted sphere is the diameter of the patient's femoral head, and the coordinates of the center of the sphere are the center of the patient's femoral head. The fitting process strictly follows the fitting rules between the point cloud data and the surface of the sphere.

[0027] Furthermore, in step S3, when determining the femoral neck axis of the patient, the original axis data of the femoral neck is first obtained based on the accurately mapped landmarks. Then, the original axis data is corrected by combining the coordinates of the femoral head center and the anatomical structure of the proximal femur using a vector correction algorithm. This eliminates the axis offset caused by the small error in the mapping of the landmarks, and obtains the accurate femoral neck axis. The anatomical standard parameters of the proximal femur are referenced during the correction process.

[0028] Furthermore, the training database used in step S4 contains complete morphological data of femoral stem prostheses of various specifications, including the contour curve of the prosthesis stem, cross-sectional geometric parameters, position coordinates of the prosthesis head center, and spatial structure data of the prosthesis as a whole. All data has been standardized and the data format is consistent with the operation requirements of the registration algorithm. The database is regularly updated with morphological data of newly added femoral stem prostheses of various specifications.

[0029] Furthermore, during local registration in step S2, a corresponding registration strategy is selected based on the anatomical characteristics of different key regions. The femoral head region adopts a registration method based on surface curvature, the femoral neck region adopts a registration method based on axis alignment, and the region around the lesser trochanter adopts a registration method based on feature point density. Each registration method is adapted to the anatomical characteristics of the region, and the geometric association rules of vertices within the region are followed during the registration process.

[0030] Compared with existing technologies, this CT image-based 3D registration method for hip joint prostheses has the following advantages:

[0031] This invention effectively solves the problems of existing technologies by constructing a complete technical process of 3D reconstruction and 2D feature extraction, precise localization of anatomical landmarks, comprehensive analysis of proximal femoral morphology, and prosthesis size registration. Through the ICP algorithm combined with global and local registration of landmarks, it eliminates reliance on manual operation and experience-based judgment, overcomes the shortcomings of insufficient adaptation due to individual anatomical differences, and achieves precise landmark mapping. The medullary canal morphology analysis method, which combines multi-plane intersection with 2D and 3D feature fusion, compensates for the incomplete analysis of medullary canal morphology in existing technologies, providing accurate data support for registration. By fusing Jaccard and ORB adaptive weighting algorithms and using the minimum distance selection principle, it achieves precise matching between the prosthesis and the patient's femoral anatomy, improving registration accuracy and efficiency, reducing the probability of postoperative complications, and extending the prosthesis's lifespan.

[0032] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0034] Figure 1 A flowchart of a 3D registration method for hip joint prostheses based on CT images;

[0035] Figure 2 A flowchart of a 3D registration method for hip joint prostheses based on CT images;

[0036] Figure 3 Flowchart for determining the morphology of the femoral medullary cavity in patients;

[0037] Figure 4 A flowchart for the algorithmic registration of the femoral stem prosthesis size. Detailed Implementation

[0038] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0039] This invention provides a 3D registration method for hip joint prostheses based on CT images, aiming to solve the problems of existing registration techniques such as reliance on human experience, insufficient accuracy of landmark positioning, and incomplete analysis of medullary canal morphology. It provides a precise prosthesis fitting solution for hip replacement surgery. The core of the method revolves around a complete technical process: "3D reconstruction and 2D feature extraction - precise positioning of anatomical landmarks - comprehensive analysis of proximal femoral morphology - prosthesis size algorithm registration." (See [link to relevant documentation]). Figure 1 and Figure 2 The specific technical solution is as follows:

[0040] First, 3D structural reconstruction and 2D feature extraction were performed. The patient's CT images were imported using 3D slicer software to reconstruct the 3D structure of the femur. Simultaneously, for target areas such as the femoral head, femoral neck, lesser trochanter region, and femoral medullary cavity projection area, the software's edge detection function was used to enhance boundary differentiation. Following the process of "contour recognition followed by precise extraction," 2D feature segmentation and extraction were completed, providing basic data for subsequent registration.

[0041] Secondly, precise localization of anatomical landmarks was implemented. A 3D mesh of the patient's femur, template bone, and template bone landmarks were imported. The position and orientation relationship between these elements were observed using a 3D slicer visualization function. When significant differences were found, initial alignment was achieved by combining 2D feature registration results with manual fine-tuning. The ICP algorithm was used for initial global registration. Rotation, translation, and scaling parameters were iteratively adjusted, and initial landmarks were determined based on a preset threshold for distance error. Subsequently, adaptive local registration strategies were adopted for different key regions: the femoral head region was registered based on surface curvature, the femoral neck region based on axis alignment, and the region around the lesser trochanter based on feature point density, ultimately forming precisely mapped landmarks.

[0042] Next, a comprehensive analysis of the proximal femur morphology was conducted. Based on precise landmarks, key parameters such as femoral head diameter, femoral head center, femoral neck axis, femoral offset, and femoral neck angle were determined. In the analysis of the femoral medullary canal morphology, starting from a position below the lesser trochanter landmark, continuous boundary contours were obtained by intersecting multiple planes according to a preset rule. The advantages of the Jaccard algorithm and the ORB algorithm were combined to extract 2D features and enhance 3D features. Geometric analysis was performed on each contour to determine the inscribed circle. The proximal femoral axis was defined by the bisector of the line connecting the centers of the inscribed circles, achieving a comprehensive and accurate depiction of the medullary canal morphology.

[0043] Finally, the femoral stem prosthesis size registration algorithm was completed. A training database containing complete morphological data of prostheses of various sizes was retrieved, and the patient's medullary canal geometry was compared with the prosthesis morphology one by one to simulate the prosthesis implantation process. Combining the segmented and extracted 2D features, an adaptive weighted fusion algorithm was used to complete 3D coordinate alignment from the dimensions of area matching and angle consistency. The relative distance between the prosthesis head center and the patient's femoral head center was calculated, and then combined with morphological matching similarity, the best-fitting femoral stem prosthesis was selected through a fit score.

[0044] This method eliminates the reliance on manual operation and experience-based judgment, effectively adapts to individual anatomical differences, improves registration accuracy and efficiency, reduces the risk of postoperative complications, and extends the lifespan of the prosthesis.

[0045] Example 1

[0046] This embodiment is applied to the preoperative planning of hip replacement surgery. It is designed for patients who need replacement surgery due to damage to the hip joint structure caused by trauma, degeneration, or disease. It solves the problems of traditional prosthesis registration relying on human experience, insufficient accuracy of landmark positioning, and incomplete analysis of medullary cavity morphology. By accurately matching the patient's proximal femoral anatomy with the hip joint prosthesis parameters, it provides a scientific prosthesis selection and implantation plan for the surgery, ensuring postoperative joint function recovery and reducing the risk of complications.

[0047] See Figure 1 and Figure 2 The specific implementation process of this embodiment is as follows:

[0048] First, 3D reconstruction of the femur and 2D feature segmentation extraction were performed. CT images of the patient's femur were acquired and imported into 3Dslicer software. Utilizing the software's 3D reconstruction function, based on the high-resolution anatomical data from the CT images, the complete 3D structure of the patient's femur was reconstructed, clearly presenting the spatial morphology of key areas such as the femoral head, femoral neck, lesser trochanter, and femoral medullary cavity. Simultaneously, 2D feature segmentation extraction was performed on target areas within the 3D structure, encompassing the femoral head region, femoral neck region, the area around the lesser trochanter, and the projected area corresponding to the femoral medullary cavity. During segmentation, the edge detection function of the 3D slicer software was enabled to enhance the distinction between the target area and surrounding soft tissues, avoiding interference from surrounding tissues. The operation process of first contour recognition and then precise extraction was strictly followed to ensure that the acquired 2D feature images were complete and clear, accurately reflecting the anatomical morphological characteristics of the target area, providing reliable basic data support for subsequent registration steps.

[0049] Next, precise localization of anatomical landmarks was performed. 3D mesh data of the patient's femur was imported, along with pre-set template bone and its standard anatomical landmark data. Using the interactive visualization function of the 3D slicer software, the position and orientation of the template bone and the patient's femur in space were observed in real time. If there were significant differences, the template bone was initially aligned with the patient's femur by combining the previously obtained 2D segmentation feature registration results with manual fine-tuning, laying the foundation for subsequent precise registration.

[0050] The initial global registration operation is performed, and the geometric data of the template bone and the patient's femur are input into the algorithm system according to the operation logic of the ICP algorithm. The algorithm performs rotation, translation, and scaling transformations on the template model, continuously calculates the geometric differences between the template bone and the patient's femur, and iteratively adjusts the three transformation parameters—rotation angle, translation vector, and scaling ratio—based on the difference results. The distance error between the template marker point and the patient's femur surface point is used as the optimization target until the distance error reaches a preset threshold. The patient's femur surface point determined at this point is the initial marker point. During the iteration process, the distance error value after each parameter adjustment is recorded synchronously for easy traceability and verification.

[0051] After initial global registration, local registration is performed on key corresponding areas around the femoral head, femoral neck, and lesser trochanter. Based on the anatomical features of the skeleton, the boundary range of each key area is accurately identified, and appropriate registration strategies are adopted for the anatomical characteristics of different areas: the femoral head area uses a registration method based on surface curvature, the femoral neck area uses a registration method based on axis alignment, and the area around the lesser trochanter uses a registration method based on feature point density, ensuring that the registration process conforms to the geometric association rules of the vertices within the area.

[0052] A registration algorithm based on distance and geometric features is used to match vertices in the local region. In the specific implementation of this embodiment, the local registration error is calculated using the following formula: ,in This represents the total local registration error. This represents the number of vertices in the local region. , , The coordinates of the apex of the local area of ​​the patient's femur. , , Register the coordinates of the vertices in the corresponding region of the template bone. , where is the vertex weight coefficient. The complexity of the anatomical structure of the region where the vertex is located is determined by statistical analysis of the vertex contribution of successful registration cases in the training set. Vertices in key parts of the anatomical structure correspond to higher weight coefficients. Through this error calculation, accurate matching of vertices is achieved, and finally, accurate mapping of the marker points is formed.

[0053] Subsequently, a comprehensive analysis of the proximal femur morphology was conducted. Based on the previously obtained precise mapping landmarks, the system automatically analyzed the proximal femur morphology of the patient and determined various key parameters. When determining the femoral head diameter and center, the 3D mesh point cloud data corresponding to the femoral head was first denoised to remove abnormal discrete points and avoid interfering with the calculation results. Then, the least squares method was used to perform spherical fitting calculations on the denoised point cloud data. During the fitting process, the fitting rules between the point cloud data and the sphere surface were strictly followed. The diameter of the fitted sphere is the diameter of the patient's femoral head, and the coordinates of the center of the sphere are the center of the patient's femoral head.

[0054] When determining the femoral neck axis, the original axis data of the femoral neck is first obtained based on the accurately mapped landmarks. Then, combined with the determined coordinates of the femoral head center and the anatomical structure of the proximal femur, the original axis data is corrected for deviation using a vector correction algorithm to eliminate axis offset caused by minor errors in landmark mapping. During the correction process, the anatomical standard parameters of the proximal femur are referenced to ensure an accurate femoral neck axis is obtained.

[0055] When determining the morphology of the femoral medullary cavity, see [reference needed]. Figure 3 Starting from a position below the lesser trochanter landmark of the patient's femur, multiple planes are intersected according to a preset rule to obtain continuous femoral medullary cavity boundary contours. Each contour is presented in both 3D and 2D views for comprehensive observation. The Jaccard and ORB algorithms are used to perform comprehensive calculations on the boundary contours in the 2D view according to an adaptive weight allocation rule. In the specific implementation of this embodiment, the contour comprehensive features are obtained through the following formula: ,in For contour comprehensive feature values, Contour similarity features calculated by the Jaccard algorithm. Features for feature point matching calculated by the ORB algorithm. These are adaptive weighting coefficients. The matching confidence is dynamically determined by the matching confidence of the two algorithms. The matching confidence is calculated by the number of matching point pairs output by each algorithm and the matching consistency test results. Algorithms with more matching point pairs and higher consistency correspond to higher weight coefficients.

[0056] Based on this comprehensive feature extraction, 2D features are extracted and 3D features are enhanced to improve the accuracy of the feature data. Geometric analysis is performed on each boundary contour to determine its inscribed circle. The center coordinates and radius data of each inscribed circle are recorded in detail. The centers of all inscribed circles are connected, and the bisector of this line is obtained through geometric calculations. This bisector is defined as the proximal femoral axis of the patient. Simultaneously, combining the above data, key parameters such as femoral offset and femoral neck angle are further determined, completing a comprehensive analysis of the proximal femoral morphology.

[0057] Finally, the femoral stem prosthesis size registration algorithm was completed. (See also...) Figure 4 The system retrieves femoral stem prosthesis morphological data from the training database. This database contains complete morphological data of femoral stem prostheses of various specifications, covering the contour curve of the prosthesis stem, cross-sectional geometric parameters, position coordinates of the prosthesis head center, and the overall spatial structure data of the prosthesis. All data has been standardized, and the data format is consistent with the computational requirements of the registration algorithm. The database is also regularly updated with new femoral stem prosthesis morphological data of newly added specifications to ensure the comprehensiveness and timeliness of the data.

[0058] The geometric features of the patient's femoral medullary cavity were compared one by one with the prosthesis morphology data in the database. A simulation system was used to model the implantation process of the femoral stem prosthesis within the femoral medullary cavity. Combining key parameters of the proximal femur, the appropriate placement of the prosthesis was determined, and the relative distance between the prosthesis head center and the patient's femoral head center was calculated. The previously segmented and extracted 2D features were then input into a fusion calculation model combining the Jaccard and ORB algorithms. A comprehensive calculation was performed with adaptive weights, obtaining 2D registration results from both area matching and angular consistency dimensions. Based on these results, precise 3D coordinate alignment was achieved.

[0059] For each candidate femoral stem prosthesis, geometric parameters are calculated, including the center and radius of the prosthesis's circumcircle. Combined with the previously calculated relative distance between the prosthesis head center and the patient's femoral head center, the prosthesis fit score is calculated using the following formula in this specific implementation: ,in Score the fit of the implant. This refers to the relative distance between the center of the prosthesis head and the center of the patient's femoral head. The similarity of the morphology of the prosthesis and the femoral medullary cavity. This represents the scoring weighting coefficient. Based on clinical follow-up data, the influence weights of different fitting parameters on the long-term stability of the prosthesis were determined by analyzing the contribution ratios of distance factors and morphological matching factors in a large number of postoperative cases. According to the calculated fitting score, the prosthesis with the highest score was selected as the best-fitting femoral stem prosthesis, completing the entire 3D registration process.

[0060] In summary, this embodiment achieves precise 3D registration of hip joint prostheses through a complete technical process, effectively eliminating reliance on manual operation and experience-based judgment in application scenarios. By combining the ICP algorithm with global and local registration, it achieves accurate mapping of anatomical landmarks, overcoming the shortcomings of insufficient adaptation due to individual anatomical differences. In femoral medullary cavity morphology analysis, the method of multi-plane intersection combined with 2D and 3D feature fusion compensates for the incompleteness of traditional technical analysis, providing accurate data support for registration. Through the fusion of Jaccard and ORB adaptive weighting algorithms and adaptation score screening, precise matching between the prosthesis and the patient's femoral anatomy is achieved.

[0061] Example 2

[0062] This embodiment is applied to hip replacement surgery scenarios with complex proximal femoral anatomy. It addresses the problems of poor compatibility and insufficient data support of traditional registration methods in such complex cases by optimizing the data acquisition and analysis process based on the aforementioned embodiment. This achieves precise matching between the prosthesis and the complex anatomical structure, providing reliable preoperative planning and intraoperative reference for high-difficulty hip replacement surgery.

[0063] See Figure 1 and Figure 2 The specific implementation process of this embodiment is as follows:

[0064] First, multimodal CT image acquisition and data preprocessing are performed. Plain CT images and enhanced CT images of the patient's femur are acquired, and the two types of images are imported into 3D slicer software for multimodal fusion. The enhanced CT image's clear representation of blood vessels and areas of bone density difference helps distinguish between diseased tissue and normal bone structure. Based on the 3D reconstruction process described in the previous embodiment, the anatomical information from the two types of CT images is integrated through the software's multimodal data fusion module to reconstruct the 3D structure of the patient's femur, ensuring the complete representation of the morphological details of deformed areas and irregular medullary cavities.

[0065] In the 2D feature segmentation and extraction stage, the selection range of the target area is maintained, and the software's adaptive threshold segmentation function is enabled. Combined with the density gradient data of enhanced CT images, the boundary between the target area and surrounding abnormal tissues is strengthened. The segmentation operation still follows the process of contour recognition followed by precise extraction. For irregular areas of medullary cavity morphology, a contour completion algorithm is added to fill in contour breaks caused by bone defects or deformities, ensuring that the acquired 2D feature images can fully reflect the anatomical morphological features under complex structures, providing more comprehensive basic data for subsequent registration steps.

[0066] When locating anatomical landmarks, an anatomical landmark pre-screening step is added to the template bone import and preliminary alignment process described in the previous embodiment. Using the anatomical structure recognition module of the 3D slicer, combined with clinical anatomical standards, landmarks that are not affected by deformities and have high recognizability are automatically selected as core reference points. Then, based on these core reference points, the approximate location range of fuzzy landmarks is deduced, reducing initial alignment errors.

[0067] The initial global registration still uses the ICP algorithm, with the same computational logic as described above, iteratively adjusting the transformation parameters until the distance error reaches a preset threshold. In the local registration phase, based on the aforementioned partitioned registration strategy, a registration method based on local surface fitting is adopted for deformed regions, achieving vertex matching by fitting the geometric features of local micro-surfaces. In the specific implementation of this embodiment, the local registration error is also calculated using the following formula: In addition, a dynamic error feedback mechanism is added. If the local registration error exceeds the threshold, it will automatically backtrack to the global registration stage, adjust the weight allocation of the core reference points, and recalculate until an accurate marker point mapping is formed.

[0068] When comprehensively analyzing the proximal femoral morphology, optimizations are made to address complex structures based on the aforementioned parameter determination process. When determining the femoral head diameter and center, in addition to noise reduction and spherical fitting, a validation step is added to verify the fitting results. This involves comparing the curvature changes in multiple local areas on the femoral head surface to verify the fit of the fitted sphere. When determining the femoral neck axis, bone density distribution shown on enhanced CT scans is used, prioritizing landmark data from dense bone regions for axis correction to improve axis accuracy.

[0069] When determining the morphology of the femoral medullary cavity, see [reference needed]. Figure 3 Building upon the aforementioned contour acquisition through multi-plane intersection, surface reconstruction is added to aid analysis. Using the software's surface reconstruction function, continuous boundary contours are fitted into a complete 3D surface of the medullary cavity. Combined with contour features from the 2D view, this enhances the comprehensiveness of the morphological analysis. When using the Jaccard algorithm and ORB algorithm for comprehensive computation, in this specific implementation, the contour synthesis features are obtained through the following formula: For irregular contours, the matching window size of the algorithm is dynamically adjusted to ensure the completeness of feature point extraction. The subsequent inscribed circle analysis and proximal femoral axis definition process is the same as described above, but multiple verification steps are added to ensure parameter accuracy through cross-validation on different planes.

[0070] When completing the femoral stem prosthesis size registration algorithm, please refer to Figure 4The training database, based on the aforementioned standard prosthesis data, adds morphological data of personalized prostheses, covering special specifications of prostheses suitable for different types of deformities. When comparing the patient's medullary cavity geometric features with the database data, geometric topological structure matching analysis is added, focusing on comparing topological features such as curvature changes and concavity / convexity distribution of the medullary cavity surface. A collision detection mechanism is added during the simulation of the implantation process to avoid interference between the prosthesis and deformed bone.

[0071] The 2D feature fusion calculation and 3D coordinate alignment process is the same as described above. When calculating the prosthesis fit score, mechanical simulation parameters are added as a supplement to the aforementioned formula. The stress distribution after prosthesis implantation is simulated using a finite element analysis module, and the stress uniformity parameter is calculated and incorporated into the morphological matching similarity assessment. In the specific implementation of this embodiment, the prosthesis fit score is still calculated using the following formula: The morphological matching similarity index has been integrated with the stress uniformity index. In addition to clinical follow-up data, postoperative biomechanical feedback data from complex cases were also considered to ensure that the scoring more closely reflects the actual needs of complex scenarios. Finally, the prosthesis with the highest fit score and uniform stress distribution was selected as the optimal fit prosthesis, completing the entire 3D registration process.

[0072] In summary, this embodiment, based on the aforementioned embodiments, optimizes the registration process for patients with complex proximal femoral anatomy. Through optimization measures such as multimodal CT image fusion, landmark pre-screening, dynamic error feedback, and mechanical simulation supplementation, it effectively solves the registration challenges caused by complex medullary canal morphology and ambiguous landmarks. Multimodal data fusion improves the completeness and accuracy of anatomical structure presentation, the dynamic optimization mechanism ensures the precision of landmark positioning and morphological analysis, and the database containing personalized prostheses and mechanical simulation analysis enhance the practicality and safety of prosthesis fitting.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for 3D registration of hip joint prostheses based on CT images, characterized in that, The method includes the following steps: S1. Using 3D slicer software, the 3D structure of the patient's femur is reconstructed based on CT images, and the 2D features of the target part of the 3D structure are segmented and extracted. S2. Perform anatomical landmark localization, import the 3D mesh of the patient's femur and the template bone and the landmarks of the template bone and realign them, perform initial global registration through the ICP algorithm to determine the initial landmarks, and then perform local registration for the femoral head, femoral neck and lesser trochanter area to obtain the precisely mapped landmarks. S3. Based on the mapping of landmark points, analyze the proximal femoral morphology to determine the femoral head diameter, femoral head center, femoral neck axis, femoral medullary cavity morphology, proximal femoral axis, femoral offset, and femoral neck axis angle. S4. Perform femoral stem prosthesis size registration algorithm, compare the femoral medullary cavity morphology of the patient with the femoral prosthesis morphology in the training database, and complete the 3D coordinate alignment by combining the comprehensive calculation of 2D features to select the best-fitting femoral stem prosthesis.

2. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, Step S2, when locating anatomical landmarks, includes the following steps: The position and orientation relationship between the template bone and the patient's femur are observed through the visualization and interactive function of the 3D slicer. If the difference is large, the registration results of 2D segmentation features are combined with manual fine-tuning to make the two initially aligned. During the initial global registration, the geometric data of the template bone and the patient's femur are input according to the operation logic of the ICP algorithm. The template model is rotated, translated and scaled, and the geometric differences between the two are continuously calculated and the transformation parameters are iteratively adjusted. Identify the key corresponding regions around the femoral head, femoral neck, and lesser trochanter. Recognize region boundaries based on skeletal anatomical features. Use a registration algorithm based on distance and geometric features to match vertices in local regions. During the matching process, calculate the local registration error using the following formula: ,in, This represents the total local registration error. This represents the number of vertices in the local region. The coordinates of the vertex of the local area of ​​the femur in the patient. Register the coordinates of the vertices in the corresponding region of the template bone. These are the vertex weight coefficients, which ultimately form the precise mapping of the marker points.

3. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, Step S3, in determining the morphology of the patient's femoral medullary cavity, includes the following steps: Starting from a position below the lesser trochanter of the patient's femur, the multi-plane intersection is controlled according to a preset rule to obtain a continuous femoral medullary cavity boundary contour, and each contour is presented in the form of 3D view and 2D view respectively; Using the Jaccard algorithm and the ORB algorithm, the boundary contours in the 2D view are comprehensively calculated according to the adaptive weight allocation rule, and the contour comprehensive features are obtained by fusing them using the following formula: ,in, For contour comprehensive feature values, Contour similarity features calculated by the Jaccard algorithm. Features for feature point matching calculated by the ORB algorithm. To adapt the weighting coefficients, 2D features are extracted based on this comprehensive feature, and the 3D features are enhanced. Perform geometric analysis on each boundary profile to determine its inscribed circle, and record the center coordinates and radius data of each inscribed circle; Connect the centers of all inscribed circles, and obtain the bisector of the line through geometric calculations. Define the bisector as the proximal axis of the diseased femur.

4. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, Step S4, when performing femoral stem prosthesis size registration algorithm, includes the following steps: The femoral stem prosthesis morphology data in the training database were retrieved, and the geometric features of the medullary cavity of the patient's femur were compared with the prosthesis morphology data one by one. The implantation process of the femoral stem prosthesis in the femoral medullary cavity was simulated to determine the appropriate placement position and calculate the relative distance between the center of the prosthesis head and the center of the patient's femoral head. For the 2D features extracted from the segmentation, they are input into the fusion calculation model of the Jaccard algorithm and the ORB algorithm. The comprehensive calculation is performed according to the adaptive weights to obtain the 2D registration results from the two dimensions of area matching degree and angle consistency, and complete the alignment of the 3D coordinates. For each candidate femoral stem prosthesis, geometric parameters were calculated, including the center and radius of the circumcircle of the prosthesis outline. Combined with previously calculated relative distance data, the prosthesis fit score was calculated using the following formula: ,in, Score the fit of the implant. This refers to the relative distance between the center of the prosthesis head and the center of the patient's femoral head. The similarity of the morphology of the prosthesis and the femoral medullary cavity. Using the scoring weight coefficient, the prosthesis with the highest fit score is selected as the best fit prosthesis.

5. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, In step S1, when performing 2D feature segmentation and extraction on the 3D structural target area, the target area includes the femoral head region, femoral neck region, region around the lesser trochanter, and the projection region corresponding to the femoral medullary cavity. During the segmentation process, the edge detection function of the 3D slicer software is used to enhance the boundary distinction between the target area and the surrounding tissue. The segmentation operation follows the process of first contour recognition and then precise extraction to obtain a complete 2D feature image.

6. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, In step S2, during the initial global registration, the ICP algorithm optimizes the distance error between the template marker point and the patient's femoral surface point. It iteratively calculates and continuously adjusts the three transformation parameters: rotation angle, translation vector, and scaling ratio until the distance error reaches a preset threshold. The patient's femoral surface point determined at this point is the initial marker point. During the iteration process, the distance error value after each parameter adjustment is recorded.

7. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, In step S3, when determining the diameter and center of the femoral head, the 3D grid point cloud data corresponding to the femoral head is first denoised to remove abnormal discrete points. Then, the least squares method is used to perform sphere fitting on the denoised point cloud data. The diameter of the fitted sphere is the diameter of the femoral head, and the center coordinates of the sphere are the center of the femoral head. The fitting process strictly follows the fitting rules between the point cloud data and the surface of the sphere.

8. The method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, In step S3, when determining the femoral neck axis of the patient, the original axis data of the femoral neck is first obtained based on the accurately mapped landmarks. Then, the original axis data is corrected by combining the coordinates of the femoral head center and the anatomical structure of the proximal femur using a vector correction algorithm. This eliminates the axis offset caused by small errors in the mapping of landmarks, thus obtaining an accurate femoral neck axis. The anatomical standard parameters of the proximal femur are referenced during the correction process.

9. A method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, The training database used in step S4 contains complete morphological data of femoral stem prostheses of various specifications, including the contour curve of the prosthesis stem, cross-sectional geometric parameters, position coordinates of the prosthesis head center, and spatial structure data of the prosthesis as a whole. All data has been standardized and the data format is consistent with the operation requirements of the registration algorithm. The database is updated regularly with morphological data of newly added femoral stem prostheses of different specifications.

10. A method for 3D registration of hip joint prostheses based on CT images according to claim 1, characterized in that, During local registration in step S2, a corresponding registration strategy is selected based on the anatomical features of different key regions. The femoral head region adopts a registration method based on surface curvature, the femoral neck region adopts a registration method based on axis alignment, and the region around the lesser trochanter adopts a registration method based on feature point density. Each registration method is adapted to the anatomical features of the region, and the geometric association rules of vertices within the region are followed during the registration process.

Images