Orthopedic image intelligent segmentation and operation path planning system

Abstract

The invention discloses an intelligent orthopedic image segmentation and surgical path planning system, which relates to the technical field of medical image processing and comprises an image acquisition module, a multi-modal characteristic mode recognition module, an intelligent segmentation module, a surgical path planning module, an interactive display module and a data storage module. According to the method, accurate extraction and fusion of homologous features are realized through a multi-modal feature mode recognition technology, the problem of insufficient segmentation precision caused by existing multi-modal image feature conflicts is solved, a generated high-precision segmentation result provides a reliable basis for path planning, and by constructing a path constraint mechanism in which the segmentation result is associated with an anatomical mode, the segmentation precision is improved. According to the method, deep collaboration of image segmentation and surgical path planning is realized, the problem that the path is inconsistent with actual anatomy due to disjunction of the image segmentation and the surgical path planning is avoided, surgical risks under a complex structure are effectively avoided through real-time identification and path parameter adjustment of an abnormal anatomy mode, and the problem that existing planning is insufficient in safety is solved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of medical image processing, in particular to a kind of orthopedic image intelligent segmentation and surgical path planning system. BACKGROUND

[0002] Orthopedics is an important branch of medical field, mainly studies the disease diagnosis, treatment and rehabilitation of movement system such as bone, joint, muscle and ligament, covers fracture repair, tumor resection, joint replacement, spinal correction and various operation types, and the core goal is to restore the normal physiological function and structural integrity of movement system, and orthopedic image is the key basis for diagnosis and operation planning of orthopedic diseases, which is the image data of human movement system obtained by CT, MRI, X-ray and other imaging devices, which can directly present the anatomic information such as bone shape, lesion position and soft tissue distribution, and provide important support for doctors to judge the disease and develop surgical plan.

[0003] Image intelligent segmentation is the core link of orthopedic image processing, which is to accurately separate target structures such as bone, nerve, blood vessel and lesion tissue from complex image data, to provide clear and accurate anatomic basis for surgical path planning; surgical path planning is based on the anatomic structure information after segmentation, combined with surgical target and clinical experience, to develop the optimal path of instrument insertion and lesion resection, both complement each other, the accuracy of image intelligent segmentation directly determines the reliability of surgical path planning, and the demand of surgical path planning reversely guides the key direction of image segmentation, and the collaborative effect directly affects the accuracy, safety and efficiency of orthopedic surgery.

[0004] With the development of medical technology, the requirement of orthopedic surgery for precision and minimally invasive is increasing, but the existing technology still has certain defects, the imaging principle difference of multi-modal image leads to feature conflict, the existing fusion method is mostly simple superposition, which is difficult to effectively extract homologous anatomic features, affects the segmentation accuracy, image intelligent segmentation and surgical path planning are independent of each other, the segmentation result is not fully associated with the mode characteristics of anatomic structure for path constraint, which leads to the disconnection between planned path and actual anatomic condition, in complex anatomic structure such as spine and joint, the spatial relationship of nerve, blood vessel and bone is complex, the existing planning scheme is difficult to accurately avoid risk area, and the path safety is insufficient, therefore, it is of great significance to develop a kind of orthopedic image intelligent segmentation and surgical path planning system. SUMMARY

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent segmentation and surgical path planning system for orthopedic images. It can accurately extract and fuse homologous features through multimodal feature pattern recognition technology, and the generated high-precision segmentation results provide a reliable basis for path planning. By constructing a path constraint mechanism that associates segmentation results with anatomical patterns, it achieves deep synergy between image segmentation and surgical path planning, avoiding the problem of paths not matching actual anatomy caused by the disconnect between the two. Through real-time identification of abnormal anatomical patterns and adjustment of path parameters, it effectively avoids surgical risks under complex structures, and ultimately achieves a significant improvement in the accuracy, safety and efficiency of orthopedic surgery.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent segmentation and surgical path planning system for orthopedic images, the system comprising: an image acquisition module, a multimodal feature pattern recognition module, an intelligent segmentation module, a surgical path planning module, an interactive display module, and a data storage module;

[0007] The image acquisition module is used to acquire CT and MRI image data of the patient's surgical area and then transmit them to the multimodal feature pattern recognition module.

[0008] The multimodal feature pattern recognition module is based on pattern recognition technology. It performs cross-domain recognition of bone structure density patterns in CT images and soft tissue signal patterns in MRI images, extracts homologous feature patterns corresponding to bones, nerves, and blood vessels, weakens feature conflicts, and transmits homologous feature data to the intelligent segmentation module.

[0009] The intelligent segmentation module receives homogeneous feature data, generates high-precision segmentation results through feature fusion algorithms, and transmits the segmentation results to the surgical path planning module.

[0010] The surgical path planning module reads the segmentation results, identifies anatomical patterns and constructs path planning constraints, automatically calculates path safety parameters, adjusts planning parameters in real time when abnormal anatomical patterns are identified, and transmits the planned path information to the interactive display module.

[0011] The interactive display module is used to display relevant information and support manual fine-tuning. The fine-tuned information is fed back to the surgical path planning module and the data storage module. The data storage module is used to store various types of data.

[0012] Furthermore, the cross-domain recognition process of the multimodal feature pattern recognition module includes:

[0013] First, grayscale correction and edge enhancement are performed on the CT and MRI images respectively to remove image noise;

[0014] Then, a convolutional neural network is used to extract the bone structure density feature vector from the CT images. Soft tissue signal feature vectors of MRI images ;

[0015] The cosine similarity algorithm is used to filter out homologous features representing the same anatomical structure from two types of vectors. Based on the weight allocation rules of imaging modalities, a formula for calculating the weight of homologous feature fusion is constructed: ,in, For the first The fusion weights of homologous features, From or The intensity value of the feature signal extracted from it. For modal fit coefficients, This represents the anatomical structure complexity coefficient. The total number of homologous features. Based on experimental data on the imaging sensitivity of CT and MRI for different anatomical structures, the following criteria were determined. Based on the anatomical complexity of bones, nerves, and blood vessels, the system presets and stores them, assigns corresponding weighting coefficients to homologous features, and weakens feature conflicts caused by non-homologous features.

[0016] Furthermore, the feature fusion algorithm of the intelligent segmentation module includes:

[0017] The homologous features output by the multimodal feature pattern recognition module are classified according to the anatomical structure types of bones, nerves, and blood vessels;

[0018] We construct feature subspaces for various anatomical structures and unify the features of each subspace to the same dimension through principal component analysis.

[0019] A weighted summation fusion strategy is used to integrate multi-dimensional features, and a feature fusion calculation formula is constructed: ,in, The integrated characteristics after fusion For the first Homology features in multiple dimensions For feature weights, The edge enhancement coefficient, For the first The gradient values ​​of each feature, The total number of features after unifying the dimensions. The contribution of each feature to the segmentation of anatomical structures was determined based on the analysis results. Based on the edge sharpness requirements of clinical image segmentation, a segmentation model is generated that includes the spatial coordinates, boundary contours, and relative positional relationships of each anatomical structure.

[0020] Furthermore, the anatomical pattern association identified by the surgical path planning module includes the spatial correspondence between the morphological patterns of key skeletal parts and the course patterns of surrounding nerves and blood vessels. The key skeletal parts include the pedicle, articular surfaces, fracture ends, and prosthesis implantation target points. By establishing a three-dimensional spatial coordinate mapping model, each anatomical structure in the segmentation results is converted into three-dimensional coordinate data, and the relative positional parameters of each anatomical structure with respect to distance and angle are quantitatively calculated. The quantitative parameters are used as the core basis for the path planning constraints.

[0021] Furthermore, the path safety parameters calculated by the surgical path planning module include the three-dimensional coordinates of the path start point, the three-dimensional coordinates of the end point, the travel angle, the penetration depth, the preset safety distance standards for nerves and blood vessels, and the quantitative parameters associated with the anatomical model. The module uses a traversal algorithm to select parameter combinations that meet the safety distance standards to form an initial planned path. The initial planned path contains at least two alternative path parameter sets.

[0022] Furthermore, the process of identifying abnormal anatomical patterns by the surgical path planning module includes:

[0023] A database of normal anatomical patterns is pre-built, which contains standard pattern data of bones, nerves, and blood vessels for people of different ages and body types;

[0024] Extract morphological features, location features, and topological relationship features of anatomical patterns from the segmentation results;

[0025] The extracted features are compared with standard patterns in the database to construct a pattern similarity calculation formula: ,in, The similarity between the pattern to be identified and the standard pattern. The first of the patterns to be identified 1 eigenvalue, For the standard mode of the first 1 eigenvalue, This is a correction factor for individual differences. For the total number of features, Based on the statistical results of individual differences from large-sample clinical anatomical data, when the similarity is lower than a preset threshold, it is judged as an abnormal anatomical pattern, triggering a path parameter adjustment mechanism.

[0026] Furthermore, the path parameter adjustment mechanism includes:

[0027] A mapping relationship between abnormal anatomical pattern types and adjustment strategies is established in advance. Abnormal anatomical pattern types include abnormal nerve course, abnormal blood vessel distribution, abnormal bone morphology, and anatomical structure displacement caused by lesion invasion.

[0028] After identifying the abnormality type, the corresponding adjustment strategy is invoked. When there is a neural pathway abnormality, the path angle and avoidance distance are adjusted. When there is a skeletal morphology abnormality, the path depth and starting point position are adjusted.

[0029] During the adjustment process, parameters are iteratively modified according to a preset step size, and the relative positions of the adjusted parameters and each anatomical structure are verified in real time to ensure that they meet the safety constraints.

[0030] Furthermore, the interactive display module provides a visual operation interface, which includes a 3D display area for segmentation results, a path planning parameter display area, and a manual adjustment toolbar. The manual adjustment toolbar supports dragging path nodes and inputting specific parameter values ​​to modify path parameters or correct segmentation boundaries. After the fine-tuning operation is triggered, the relevant information is synchronized to the surgical path planning module in real time for recalculation, and is also synchronized to the data storage module to record the modification trajectory.

[0031] Furthermore, the data storage module is divided into an original image data area, a feature data area, a segmentation result area, a planned path area, and an operation log area. Each data area establishes an associated index through the patient's unique identifier, surgical site, and data generation time. It supports multi-condition combination queries and data tracing through patient identifier, surgical date, and data type keywords. It has a timed automatic backup function, with backup data stored in an independent storage unit. It also supports data export in DICOM and JSON formats.

[0032] Compared with existing technologies, this intelligent segmentation and surgical path planning system for orthopedic imaging has the following advantages:

[0033] This invention achieves accurate extraction and fusion of homologous features through multimodal feature pattern recognition technology, solving the problem of insufficient segmentation accuracy caused by multimodal image feature conflicts in existing technologies. The generated high-precision segmentation results provide a reliable foundation for path planning. By constructing a path constraint mechanism that associates segmentation results with anatomical patterns, deep collaboration between image segmentation and surgical path planning is achieved, avoiding the problem of paths not matching actual anatomy caused by the disconnect between the two. Through real-time identification of abnormal anatomical patterns and adjustment of path parameters, surgical risks under complex structures are effectively avoided, solving the problem of insufficient safety in existing planning. Ultimately, this invention achieves a significant improvement in the accuracy, safety, and efficiency of orthopedic surgery, providing strong support for precise minimally invasive orthopedic treatment.

[0034] 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

[0035] 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.

[0036] Figure 1 This is a schematic diagram of a smart segmentation and surgical path planning system for orthopedic imaging.

[0037] Figure 2 This is a flowchart of a smart segmentation and surgical path planning system for orthopedic images. 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 an intelligent segmentation and surgical path planning system for orthopedic images. Its core is to achieve precision and safety in orthopedic surgery through multi-module collaboration. (See also...) Figure 1 and Figure 2 The complete technical solution is as follows:

[0040] The system consists of an image acquisition module, a multimodal feature pattern recognition module, an intelligent segmentation module, a surgical path planning module, an interactive display module, and a data storage module. Each module performs its own function and works in deep collaboration.

[0041] The image acquisition module is responsible for acquiring CT and MRI images of the patient's surgical area and transmitting them to the multimodal feature pattern recognition module. Based on pattern recognition technology, the multimodal feature pattern recognition module first performs grayscale correction and edge enhancement preprocessing on the two images to remove noise. Then, it extracts bone density features from the CT images and soft tissue signal features from the MRI images. By filtering for homologous features and assigning fusion weights, it weakens feature conflicts and transmits the processed homologous feature data to the intelligent segmentation module.

[0042] After receiving the homologous feature data, the intelligent segmentation module classifies the data according to the anatomical structure types of bones, nerves, and blood vessels, constructs a feature subspace and unifies the dimensions, integrates multi-dimensional features through a weighted summation fusion strategy, and generates a high-precision segmentation result that includes the spatial coordinates, boundary contours, and relative positional relationships of each anatomical structure. This result is then transmitted to the surgical path planning module.

[0043] After reading the segmentation results, the surgical path planning module identifies the spatial correspondence between key skeletal components and surrounding nerves and blood vessels, constructs path planning constraints, and automatically calculates safety parameters such as the path's start and end points, travel angle, and penetration depth, generating at least two alternative initial planned paths. Simultaneously, this module identifies abnormal anatomical patterns by comparing them with a pre-set database of normal anatomical patterns. For different types, such as abnormal nerve courses and abnormal blood vessel distribution, it invokes corresponding adjustment strategies to adjust the path parameters in real time, ensuring path safety.

[0044] The interactive display module provides a visual operation interface, including a 3D display area for segmentation results, a path planning parameter display area, and a manual adjustment toolbar. It supports doctors to drag path nodes or input parameters for manual fine-tuning, and the fine-tuning information is fed back to the surgical path planning module and data storage module in real time.

[0045] The data storage module is divided into multiple data areas, and an associated index is established through information such as the patient's unique identifier. It supports multi-condition combined queries and data traceability, has a scheduled automatic backup function, and supports data export in a specified format. It completely stores various types of raw data, processing results, and operation logs.

[0046] The entire system achieves deep collaboration between image segmentation and surgical path planning through multi-module coordination, effectively improving the accuracy and safety of orthopedic surgery.

[0047] Example 1

[0048] This embodiment is applied to spinal fracture repair surgery. Addressing the complex anatomy, dense neurovascular structure, and significant individual differences in the spinal region, it utilizes an intelligent orthopedic image segmentation and surgical path planning system to achieve precise image processing and safe surgical path planning. This provides surgeons with clear anatomical references and scientific path guidance. (See also...) Figure 1 and Figure 2 This technology addresses issues in traditional spinal surgery such as insufficient image segmentation accuracy, disconnect between path planning and actual anatomy, and weak risk avoidance capabilities, thereby facilitating the implementation of minimally invasive and precise surgeries.

[0049] During the preparation phase for spinal fracture repair surgery, the system's image acquisition module is first activated to acquire CT and MRI images of the patient's spinal fracture area using CT and MRI equipment, respectively. CT images focus on capturing the density distribution characteristics of the spinal bones, clearly presenting the morphology, displacement, and structure of key areas such as the pedicles at the fracture ends; MRI images accurately reflect the signal characteristics of surrounding soft tissues, clarifying the course of nerves and blood vessels and their spatial relationship with the fracture site. The image acquisition module simultaneously transmits the two types of raw image data to the multimodal feature pattern recognition module.

[0050] After receiving the image data, the multimodal feature pattern recognition module first performs grayscale correction and edge enhancement preprocessing on the CT and MRI images respectively, effectively removing noise interference and improving image quality. Then, it extracts the bone structure density feature vector from the CT images using a convolutional neural network. Soft tissue signal feature vectors of MRI images Then, the cosine similarity algorithm is used to compare the two types of feature vectors and filter out the homologous features that represent the same anatomical structure, such as the density features that correspond to a certain pedicle and the signal features of the surrounding soft tissue.

[0051] Based on the weight allocation rules of imaging modalities, a formula for calculating the weight of homologous feature fusion is constructed. .in For the first The fusion weights of homologous features, From or The intensity value of the feature signal extracted from it. These are modal adaptation coefficients, used to adapt to the different imaging modal characteristics of CT and MRI. This is the anatomical structure complexity coefficient, set according to the morphological complexity of different anatomical structures. This represents the total number of homologous features. This calculation method assigns a corresponding weighting coefficient to each selected homologous feature, weakening feature conflicts caused by non-homologous features, ensuring effective fusion of homologous features, and finally transmitting the processed homologous feature data to the intelligent segmentation module.

[0052] After receiving homologous feature data, the intelligent segmentation module classifies the data according to the anatomical structure types of bones, nerves, and blood vessels, constructing independent feature subspaces for each of the three types. Principal component analysis is used to unify the features of each subspace to the same dimension, eliminating fusion interference caused by dimensional differences. A weighted summation fusion strategy is then employed to integrate multi-dimensional features, constructing a feature fusion computation mechanism. .in The integrated characteristics after fusion For the first Homology features in multiple dimensions These are feature weights, used to highlight features that contribute significantly to anatomical structure segmentation. The edge enhancement coefficient, For the first The gradient values ​​of these features, when combined, can enhance the boundary contours of anatomical structures. This represents the total number of features after unifying the dimensions. This calculation generates a high-precision segmentation model that includes the spatial coordinates, boundary contours, and relative positional relationships of the spinal skeleton, nerves, and blood vessels, and the segmentation result is then transmitted to the surgical path planning module.

[0053] After reading the segmentation results, the surgical path planning module first identifies anatomical pattern associations, focusing on analyzing the spatial correspondence between key skeletal sites such as the pedicles and fracture ends and the course patterns of surrounding nerves and blood vessels. By establishing a three-dimensional spatial coordinate mapping model, each anatomical structure in the segmentation results is converted into three-dimensional coordinate data. The relative positional parameters, such as distances and angles between each structure, are quantitatively calculated, and these quantitative parameters serve as the core basis for path planning constraints. Based on preset safety distance standards for nerves and blood vessels, and combined with the quantitative parameters, a traversal algorithm is used to filter parameter combinations that meet safety requirements. Path safety parameters, such as the three-dimensional coordinates of the starting point, the three-dimensional coordinates of the ending point, the travel angle, and the penetration depth, are calculated, forming at least two alternative initial planned paths.

[0054] Simultaneously, this module calls upon a pre-built database of normal anatomical patterns. This database contains standard pattern data of the spine, skeleton, nerves, and blood vessels for people of different ages and body types. It extracts the morphological, positional, and topological features of the current patient's anatomical pattern from the segmentation results and compares them with the standard patterns in the database. The similarity is calculated using a pattern similarity calculation formula. Determine if any abnormal anatomical patterns exist. The similarity between the pattern to be identified and the standard pattern. The first of the patterns to be identified 1 eigenvalue, For the standard mode of the first 1 eigenvalue, This is an individual difference correction factor, used to correct for normal anatomical differences between individuals. This represents the total number of features. When the similarity is lower than a preset threshold, it is determined to be an abnormal anatomical pattern. Based on the type of abnormality, the corresponding adjustment strategy is invoked. If it is an abnormal neural pathway, the path angle and avoidance distance are adjusted. If it is an abnormal skeletal morphology, the path depth and starting point position are adjusted. During the adjustment process, the parameters are iteratively modified according to a preset step size, and the parameters are checked in real time to see if they meet the safety constraints. Finally, the optimal planned path is determined and transmitted to the interactive display module.

[0055] The interactive display module presents the segmentation results in three dimensions, clearly showing the spatial relationships of the various anatomical structures of the spine, while also displaying various parameters of the planned path. Surgeons can manually adjust the planned path by dragging path nodes or entering specific parameter values ​​through the toolbar, and can also correct the segmentation boundaries. Once a fine-tuning operation is triggered, relevant information is synchronized in real time to the surgical path planning module for recalculation, ensuring that the adjusted path still meets safety requirements. Simultaneously, the modified trajectory is recorded in the data storage module.

[0056] The data storage module categorizes and stores various types of data into raw image data area, feature data area, segmentation result area, planned path area, and operation log area. It establishes associated indexes through patient unique identifier, surgical site, and data generation time, supports multi-condition combination query and data traceability, and automatically backs up data to independent storage units at preset times. It also supports data export in DICOM and JSON formats, which facilitates subsequent case analysis and data sharing.

[0057] In summary, this embodiment achieves efficient processing of multimodal images and precise planning of surgical paths in spinal fracture repair surgery through the collaborative work of various system modules. The application of multimodal feature pattern recognition technology effectively solves the feature conflict problem and improves image segmentation accuracy; the path constraint mechanism that associates segmentation results with anatomical patterns ensures the consistency between the planned path and the actual anatomy; and the abnormal anatomical pattern recognition and parameter adjustment functions significantly reduce surgical risks.

[0058] Example 2

[0059] This embodiment is applied to artificial joint replacement surgery. Addressing the challenges of irregular bone morphology, tight soft tissue coverage, complex neurovascular branches, and extremely high precision requirements for prosthesis implantation in the joint region, it utilizes an intelligent orthopedic image segmentation and surgical path planning system to achieve precise processing of multimodal images, clear segmentation of anatomical structures, and scientific planning of the prosthesis implantation path. (See [link to relevant documentation]). Figure 1 and Figure 2 This technology addresses issues in traditional joint replacement surgery such as poor image feature fusion, disconnect between path planning and anatomical structure, and inadequate handling of abnormal anatomical structures, providing technical support for minimally invasive and precise surgical procedures.

[0060] During the preoperative preparation phase of total joint replacement surgery, the system's image acquisition module is activated to acquire CT and MRI images of the patient's joint area to be replaced using CT and MRI equipment, respectively. CT images clearly show the morphology and structure of the joint bones, bone density distribution, and key bony landmarks required for prosthesis implantation. MRI images accurately capture the signal characteristics of soft tissues such as muscles and ligaments around the joint, clarifying the branching of nerves and blood vessels and their spatial relationship with the joint surface. The image acquisition module simultaneously transmits both types of raw image data to the multimodal feature pattern recognition module.

[0061] After receiving the image data, the multimodal feature pattern recognition module first performs grayscale correction and edge enhancement preprocessing on the CT and MRI images respectively, effectively filtering out noise interference and improving image clarity and contrast. Then, it extracts the bone structure density feature vector from the CT images using a convolutional neural network. Soft tissue signal feature vectors of MRI images The cosine similarity algorithm is used to compare the two types of feature vectors one by one, filtering out homologous features representing the same anatomical structure, such as the bone density features and surrounding soft tissue signal features of corresponding articular surfaces. Based on the weight allocation rules of imaging modalities, a weight calculation formula for homologous feature fusion is constructed. Each homologous feature is assigned a corresponding weighting coefficient to weaken the conflict caused by non-homologous features and ensure the effective fusion of homologous features. The processed homologous feature data is then transmitted to the intelligent segmentation module.

[0062] After receiving homologous feature data, the intelligent segmentation module classifies the data according to the anatomical structure types of bones, nerves, and blood vessels, constructing independent feature subspaces for each of the three types of structures. Principal component analysis is used to unify the features of each subspace to the same dimension, eliminating dimensional differences between features in different subspaces. A weighted summation fusion strategy is employed to integrate multi-dimensional features, constructing a feature fusion calculation formula. The system enhances the boundary contours of anatomical structures, generates a high-precision segmentation model that includes the spatial coordinates, boundary contours, and relative positional relationships of joints, bones, nerves, and blood vessels, and transmits the segmentation results to the surgical path planning module in real time.

[0063] After reading the segmentation results, the surgical path planning module prioritizes identifying anatomical pattern associations, focusing on analyzing the spatial correspondence between key skeletal components such as articular surfaces and prosthesis implantation targets and the course of surrounding nerves and blood vessels. By establishing a three-dimensional spatial coordinate mapping model, each anatomical structure in the segmentation results is converted into three-dimensional coordinate data, and relative positional parameters such as distances and angles between each structure are quantitatively calculated. These quantitative parameters serve as the core basis for path planning constraints.

[0064] Based on a preset neurovascular safety distance standard, and combined with quantitative parameters, a traversal algorithm is used to screen parameter combinations that meet safety requirements. Path safety parameters such as the three-dimensional coordinates of the starting point and ending point, travel angle, and penetration depth are calculated, forming at least two alternative initial planned paths. Simultaneously, this module calls a pre-built database of normal anatomical patterns, encompassing standard patterns of joints, bones, nerves, and blood vessels in people of different age groups and body types. From the segmentation results, the module extracts the anatomical morphological features, positional features, and topological relationship features of the current patient's pattern, and performs a comprehensive comparison with the standard patterns in the database. The similarity is calculated using a pattern similarity calculation method. Determine if there are any abnormal anatomical patterns.

[0065] When the similarity is lower than the preset threshold, it is judged as an abnormal anatomical pattern. The corresponding adjustment strategy is called according to the abnormality type. If it is an abnormal blood vessel distribution, the path avoidance distance and travel angle are adjusted. If it is an abnormal bone morphology, the starting point position and penetration depth of the path are adjusted. During the adjustment process, the parameters are iteratively modified according to the preset step size. The relative position of the adjusted parameters and each anatomical structure is checked in real time to see if they meet the safety constraints. Finally, the optimal prosthesis implantation planning path is determined and transmitted to the interactive display module.

[0066] The interactive display module presents the segmentation results in a 3D visualization, clearly showing the spatial layout and relative relationships of the various anatomical structures of the joint, while simultaneously displaying various parameters of the planned path. Surgeons can manually adjust the toolbar, dragging path nodes or entering specific parameter values ​​to fine-tune the planned path, and can also correct the segmentation boundaries.

[0067] After a fine-tuning operation is triggered, relevant information is synchronized in real time to the surgical path planning module for recalculation, ensuring that the adjusted path still meets safety and accuracy requirements. Simultaneously, the entire modification trajectory is recorded in the data storage module. The data storage module categorizes and stores various types of data according to the original image data area, feature data area, result area, planned path area, and operation log area. It establishes a correlation index based on the patient's unique identifier, surgical site, and data generation time, supporting multi-condition combined queries and data traceability. Data is automatically backed up to an independent storage unit at preset times, and data can be exported in DICOM and JSON formats, providing comprehensive data support for postoperative case analysis and academic research.

[0068] In summary, this embodiment achieves high-quality processing of multimodal images, high-precision segmentation of anatomical structures, and accurate planning of surgical paths in artificial joint replacement surgery through the efficient collaboration of various system modules. Multimodal feature pattern recognition technology effectively resolves feature conflicts between different modalities of images, improving the accuracy of segmentation results; the path constraint mechanism that associates segmentation results with anatomical patterns ensures a high degree of fit between the planned path and the actual anatomical structure; and the abnormal anatomical pattern recognition and dynamic adjustment functions significantly enhance the safety and adaptability of the surgical path.

[0069] 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. An intelligent segmentation and surgical path planning system for orthopedic images, characterized in that, The system includes: an image acquisition module, a multimodal feature pattern recognition module, an intelligent segmentation module, a surgical path planning module, an interactive display module, and a data storage module; The image acquisition module is used to acquire CT and MRI image data of the patient's surgical area and then transmit them to the multimodal feature pattern recognition module. The multimodal feature pattern recognition module is based on pattern recognition technology. It performs cross-domain recognition of bone structure density patterns in CT images and soft tissue signal patterns in MRI images, extracts homologous feature patterns corresponding to bones, nerves, and blood vessels, weakens feature conflicts, and transmits homologous feature data to the intelligent segmentation module. The intelligent segmentation module receives homogeneous feature data, generates high-precision segmentation results through feature fusion algorithms, and transmits the segmentation results to the surgical path planning module. The surgical path planning module reads the segmentation results, identifies anatomical patterns and constructs path planning constraints, automatically calculates path safety parameters, adjusts planning parameters in real time when abnormal anatomical patterns are identified, and transmits the planned path information to the interactive display module. The interactive display module is used to display relevant information and support manual fine-tuning. The fine-tuned information is fed back to the surgical path planning module and the data storage module. The data storage module is used to store various types of data.

2. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The cross-domain recognition process of the multimodal feature pattern recognition module includes: First, grayscale correction and edge enhancement are performed on the CT and MRI images respectively to remove image noise; Then, a convolutional neural network is used to extract the bone structure density feature vector from the CT images. Soft tissue signal feature vectors of MRI images ; The cosine similarity algorithm is used to filter out homologous features representing the same anatomical structure from two types of vectors. Based on the weight allocation rules of imaging modalities, a formula for calculating the weight of homologous feature fusion is constructed: ,in, For the first The fusion weights of homologous features, From or The intensity value of the feature signal extracted from it. For modal fit coefficients, This represents the anatomical structure complexity coefficient. This represents the total number of homologous features. It assigns corresponding weighting coefficients to homologous features to weaken feature conflicts caused by non-homologous features.

3. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The feature fusion algorithm of the intelligent segmentation module includes: The homologous features output by the multimodal feature pattern recognition module are classified according to the anatomical structure types of bones, nerves, and blood vessels; We construct feature subspaces for various anatomical structures and unify the features of each subspace to the same dimension through principal component analysis. A weighted summation fusion strategy is used to integrate multi-dimensional features, and a feature fusion calculation formula is constructed: ,in, The integrated characteristics after fusion For the first Homology features in multiple dimensions For feature weights, The edge enhancement coefficient, For the first The gradient values ​​of each feature, To obtain the total number of features after unifying the dimensions, a segmentation model is generated that includes the spatial coordinates, boundary contours, and relative positional relationships of each anatomical structure.

4. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The anatomical pattern association identified by the surgical path planning module includes the spatial correspondence between the morphological patterns of key skeletal parts and the course patterns of surrounding nerves and blood vessels. Key skeletal parts include pedicles, articular surfaces, fracture ends, and prosthesis implantation targets. By establishing a three-dimensional spatial coordinate mapping model, each anatomical structure in the segmentation results is converted into three-dimensional coordinate data. The relative positional parameters of each anatomical structure with respect to distance and angle are quantitatively calculated, and the quantitative parameters are used as the core basis for the path planning constraints.

5. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The surgical path planning module calculates path safety parameters including the three-dimensional coordinates of the path start point, the three-dimensional coordinates of the end point, the travel angle, the penetration depth, preset safety distance standards for nerves and blood vessels, and quantitative parameters associated with the anatomical model. The module uses a traversal algorithm to select parameter combinations that meet the safety distance standards to form an initial planned path. The initial planned path contains at least two alternative path parameter sets.

6. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The process of identifying abnormal anatomical patterns by the surgical path planning module includes: A database of normal anatomical patterns is pre-built, which contains standard pattern data of bones, nerves, and blood vessels for people of different ages and body types; Extract morphological features, location features, and topological features of the anatomical pattern from the segmentation results; The extracted features are compared with standard patterns in the database to construct a pattern similarity calculation formula: ,in, The similarity between the pattern to be identified and the standard pattern. The first of the patterns to be identified 1 eigenvalue, For the standard mode of the first 1 eigenvalue, This is a correction factor for individual differences. The total number of features is used. When the similarity is lower than a preset threshold, it is determined to be an abnormal dissection pattern, triggering the path parameter adjustment mechanism.

7. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 6, characterized in that, The path parameter adjustment mechanism includes: A mapping relationship between abnormal anatomical pattern types and adjustment strategies is established in advance. Abnormal anatomical pattern types include abnormal nerve course, abnormal blood vessel distribution, abnormal bone morphology, and anatomical structure displacement caused by lesion invasion. After identifying the abnormality type, the corresponding adjustment strategy is invoked. When there is a neural pathway abnormality, the path angle and avoidance distance are adjusted. When there is a skeletal morphology abnormality, the path depth and starting point position are adjusted. During the adjustment process, parameters are iteratively modified according to a preset step size, and the relative positions of the adjusted parameters and each anatomical structure are verified in real time to ensure that they meet the safety constraints.

8. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The interactive display module provides a visual operation interface, which includes a 3D display area for segmentation results, a path planning parameter display area, and a manual adjustment toolbar. The manual adjustment toolbar supports dragging path nodes and inputting specific parameter values ​​to modify path parameters or correct segmentation boundaries. After the fine-tuning operation is triggered, the relevant information is synchronized to the surgical path planning module in real time for recalculation, and is also synchronized to the data storage module to record the modification trajectory.

9. The intelligent segmentation and surgical path planning system for orthopedic images according to claim 1, characterized in that, The data storage module is divided into an original image data area, a feature data area, a segmentation result area, a planned path area, and an operation log area. Each data area is associated with an index based on the patient's unique identifier, surgical site, and data generation time. It supports multi-condition queries and data tracing based on patient identifier, surgical date, and data type keywords. It has a scheduled automatic backup function, with backup data stored in an independent storage unit. It also supports data export in DICOM and JSON formats.

Images