A 3D osteotomy preoperative intelligent planning method and system

Abstract

The present invention provides a 3D osteotomy correction preoperative intelligent planning method and system, which establishes a local coordinate system according to a bone deformity model, uses a model registration module to complete local registration of the distal end of the bone, uses a spatial range window function to draw a deformity function image of the deformed bone model compared with a reconstructed target model, and roughly determines the deformity range of the deformed bone to complete automatic diagnosis of bone deformity; the purpose of determining the automatic diagnosis of bone deformity is to determine an approximate range of an osteotomy plane for multi-stage multi-objective optimization of clinical goals. Different osteotomy planes will generate different bone fragments, and the resulting osteotomy correction schemes are also different. Therefore, the automatic diagnosis of bone deformity determines the approximate range of the osteotomy plane to provide a parameter range basis for multi-stage multi-objective optimization.

Description

Technical Field

[0001] The present invention relates to the field of intelligent diagnosis and treatment systems, and in particular to a method and system for intelligent pre-operative planning of 3D osteotomy and correction. Background Art

[0002] Congenital and acquired deformities of the human limbs can lead to loss of limb length, imbalance of force lines, limited joint movement, joint degeneration and pain. In particular, deformities of the lower limbs can even lead to serious complications such as osteoarthritis of adjacent joints, pelvic tilt and scoliosis. Therefore, surgical osteotomy and correction to restore the normal anatomical morphology and force line relationship of the limb bones is a commonly used treatment option in clinical practice. During the surgical correction process, the traditional approach is for clinicians to follow the basic principles of osteotomy and correction based on their own clinical experience and imaging data such as X-rays and CT scans, and manually calculate the osteotomy position and correction angle to achieve the purpose of limb correction. However, there are often defects such as the inability to scientifically predict the surgical effect before surgery, which in turn affects the formulation of the optimal surgical plan, insufficient surgical accuracy, and poor homogeneity of surgery performed by different doctors.

[0003] The existing preoperative planning systems are mainly two-dimensional and three-dimensional. The two-dimensional preoperative planning system is only an extension of the traditional operation of clinicians and two-dimensional digitization. It cannot perform accurate correction in three-dimensional space. It still requires clinicians to operate based on experience, which consumes a lot of clinicians' time and energy and the data obtained is not accurate enough. The clinical experience of doctors is also individual, resulting in poor accuracy and consistency. At the same time, the two-dimensional system cannot realize the personalized design of osteotomy guides and bone fixation plates, and cannot be integrated with the three-dimensional surgical navigation system.

[0004] However, due to the complexity of multiple parameters such as posture, position, and clinical constraints in three-dimensional space, preoperative planning based on 3D models of medical images is difficult to quickly generate preoperative planning solutions that meet clinical needs. At the same time, the preoperative planning solutions generated in this way need to be manually calculated and verified through repeated experiments. The development of clinically feasible solutions also requires close cooperation between surgeons with clinical knowledge and engineers with technical expertise, which is time-consuming and may involve multiple manual adjustments, resulting in unnecessary clinical costs. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and provide a 3D osteotomy and correction surgery preoperative intelligent planning method to achieve osteotomy, restoration alignment, bone fixation plate positioning and fixation screw positioning of deformed bones.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention provides a 3D osteotomy preoperative intelligent planning method, comprising the following steps:

[0008] S1. Generate STL models of the deformed bone and the contralateral healthy bone using CT data of the deformed bone and the contralateral healthy bone, and import them into the 3D osteotomy orthopedic preoperative intelligent planning system. Generate a mirror image model of the contralateral healthy bone by selecting a reference surface for coordinate symmetry. Use the mirror image model as a reconstruction target model, and establish a local coordinate system for the deformed bone model and the reconstruction target model.

[0009] S2, adjusting the initial postures of the deformed bone model and the reconstructed target model through the model space change module of the 3D osteotomy orthopedic preoperative intelligent planning system; and selecting landmark points at the distal joints of the deformed bone model and the reconstructed target model and searching for domain point sets;

[0010] S3, using the model registration module of the 3D osteotomy orthopedic preoperative intelligent planning system, using the search field point set in S2 to perform a primary registration based on landmark points, and selecting the point set of the deformed bone model and the distal 30% of the reconstructed target model for a secondary precise registration;

[0011] S4, based on the precise registration in S3, generating a deformity function image of the deformed bone model through the bone deformity automatic diagnosis module, and determining the deformity area and the cutting range of the deformed bone;

[0012] S5, cutting the deformed bone using the model cutting module according to the cutting range obtained in S4;

[0013] S6, using the parameter range of the osteotomy plane generated by the cutting range as the initial input of the optimization algorithm, and calculating the optimal preoperative planning parameters through the weight-based multi-stage multi-objective optimization algorithm in the osteotomy correction optimization module;

[0014] S7. Based on the optimization result of S6, preoperative planning is performed through the 3D osteotomy orthopedic preoperative intelligent planning system, the planning result is presented in 3D form, and the gap after osteotomy is filled through the bone tissue filling module.

[0015] Furthermore, the mirror image model of the contralateral healthy bone is generated by obtaining the symmetric point coordinates based on the point coordinates of the contralateral healthy bone in the coordinate system and taking the XOZ plane as the symmetry plane, and using the symmetric point coordinates to establish the geometric structure and topological structure of the mirror image model.

[0016] Furthermore, in S1, the method for constructing the local coordinate system is:

[0017] S101, establishing the Y axis of the local coordinate system based on the long axis of the bone, and rotating around the Y axis to correspond to the corrected transverse plane;

[0018] S102, based on the principle that the rotation around the Z axis corresponds to the correction of the sagittal plane, using the established Y axis, establish the Z axis of the local coordinate system;

[0019] S103, based on the principle of rotating around the X-axis corresponding to the corrected frontal plane, and using the vector product result of the established Y-axis and Z-axis, establish the X-axis of the local coordinate system.

[0020] Furthermore, in S2, the anatomical points of the distal joint are seven landmark points, and fifty points are searched within the 50mm area of ​​each anatomical point using the established topological structure relationship, using a breadth-first search algorithm, and finally generating the deformed bone model and the reconstructed target model to perform the point set to be aligned and the target point set.

[0021] Furthermore, the topological structure relationship of the deformed bone model is established through the existing point and surface unit data of the STL model according to the principle that all surface units composed of a point constitute all adjacent surface units of the point, all line units composed of a point constitute all adjacent line units of the point, line units are adjacent to one or two surface units, and surface units are adjacent to one, two or three surface units, to establish the three-dimensional model topological structure relationship in the 3D osteotomy pre-operative intelligent planning system.

[0022] Furthermore, in S3, the two precise registrations are both performed using the nearest iterative point method. The first precise registration is an initial rough registration by selecting anatomical points near the joint as landmark points; the second precise registration is a precise registration by selecting the roughly registered distal part point set.

[0023] Furthermore, in S4, the method for automatically diagnosing the deformed area of ​​the deformed bone includes:

[0024] S401, establishing a window function, and performing spatial discretization processing on the point set of the distally registered deformed model and the reconstructed target model along the Y axis of the local coordinate system of the deformed bone model;

[0025] S402, each discrete part includes a point set of the deformed bone model and the reconstructed target model, calculates and searches for a point with the smallest Euclidean distance between each point in the deformed bone model point set and the reconstructed target model point set, and calculates a root mean square error (RMSE);

[0026] S403, using the point on the Y-axis of the local coordinate system of the deformed bone model as the horizontal coordinate and the root mean square error as the vertical coordinate, a deformity function image is established, and the deformity area is determined by a deviation of 10% from the minimum root mean square error.

[0027] Furthermore, the window function is based on the local coordinate system as a reference, the X value and the Y value surround the deformed bone model and the healthy bone model, and the Y value is a discrete range.

[0028] Further, in S6,

[0029] The parameters in the optimization algorithm are transformed according to the clinical goals in osteotomy correction surgery, and all the parameters of the four clinical goals are encoded into the chromosomes of the multi-stage multi-objective optimization algorithm;

[0030] In the weight-based multi-stage multi-objective optimization algorithm, weight allocation is to change the calculation method of congestion in the ordinary multi-objective optimization algorithm. When calculating congestion, weight allocation is performed according to the importance of the current optimization goal, so that the optimized parameters are forced to remain in the ideal point space close to the smaller congestion.

[0031] The multi-stage in the weight-based multi-stage multi-objective optimization algorithm is to divide the four optimization objectives into two stages, use the optimization result of stage one to initialize stage two, and finally perform preoperative planning based on the parameters of the optimization result of stage two.

[0032] Furthermore, a 3D osteotomy and correction surgery preoperative intelligent planning system is provided, comprising:

[0033] Model space transformation module, used to adjust the spatial position of the skeleton model before intelligent planning;

[0034] Model registration module, which provides landmark-based, nearest iteration point ICP, four-point registration and fast point feature histogram registration methods, and is used to register the deformed bone model locally or as a whole with the reconstructed target model according to different situations;

[0035] The bone deformity automatic diagnosis module is used to automatically calculate the deformity range of the deformed bone model compared with the reconstructed target model and draw a deformity function image;

[0036] The model cutting module is used to cut the deformed bone within a rough range after alignment and automatic diagnosis.

[0037] The bone filling model generation module is used to generate a bone fragment gap filling model generated after osteotomy restoration to assist in the fixation of bone fragments.

[0038] The osteotomy correction optimization module is used to generate different proximal and distal potential bone fragments as input within the parameter range of the osteotomy plane, and use a weight-based multi-stage multi-objective optimization algorithm to calculate the optimal osteotomy cutting surface, restore the alignment transformation matrix, and the spatial position of the fixation plate and screws.

[0039] The beneficial effects of the present invention are as follows: the use of the model space transformation module can make the initial posture of the bone model more conducive to registration and restoration; the setting of the model registration module makes the local or overall reduction and alignment accuracy of the deformed model higher than the accuracy of the doctor's manual calculation of the reduction and the accuracy of other software-assisted reduction; the generation of the bone filling model can be performed by 3D printing bone grafts for bone filling, making the fixation after reduction more stable; the osteotomy correction optimization module is the core module of this system, through which the osteotomy plane, reduction alignment, spatial position of the fixation plate and screw orientation can be optimized on the basis of the initial osteotomy correction, and the solutions required for the clinical goals can be automatically generated, thereby improving the preoperative planning effect and accuracy of osteotomy correction, and greatly reducing the time required for preoperative planning, thereby realizing intelligent preoperative planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a general structural diagram of a 3D osteotomy preoperative intelligent planning method of the present invention;

[0041] Figure 2 is a block diagram of an automatic diagnosis module for bone deformity;

[0042] Figure 3 Block diagram of the multi-stage multi-objective optimization algorithm based on clinical goal weights. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] See also Figure 1 , a 3D osteotomy correction preoperative intelligent planning method, comprising the following steps:

[0045] S1. Generate STL models of the deformed bone and the contralateral healthy bone using CT data of the deformed bone and the contralateral healthy bone, and import them into the 3D osteotomy orthopedic preoperative intelligent planning system. Generate a mirror image model of the contralateral healthy bone by selecting a reference surface for coordinate symmetry. Use the mirror image model as a reconstruction target model, and establish a local coordinate system for the deformed bone model and the reconstruction target model.

[0046] S2, adjusting the initial postures of the deformed bone model and the reconstructed target model through the model space change module of the 3D osteotomy orthopedic preoperative intelligent planning system; and selecting landmark points at the distal joints of the deformed bone model and the reconstructed target model and searching for domain point sets;

[0047] S3, using the model registration module of the 3D osteotomy orthopedic preoperative intelligent planning system, using the search field point set in S2 to perform a primary registration based on landmark points, and selecting the point set of the deformed bone model and the distal 30% of the reconstructed target model for a secondary precise registration;

[0048] S4, based on the precise registration in S3, generating a deformity function image of the deformed bone model through the bone deformity automatic diagnosis module, and determining the deformity area and the cutting range of the deformed bone;

[0049] S5, cutting the deformed bone using the model cutting module according to the cutting range obtained in S4;

[0050] S6, using the parameter range of the osteotomy plane generated by the cutting range as the initial input of the optimization algorithm, and calculating the optimal preoperative planning parameters through the weight-based multi-stage multi-objective optimization algorithm in the osteotomy correction optimization module;

[0051] S7. Based on the optimization result of S6, preoperative planning is performed through the 3D osteotomy orthopedic preoperative intelligent planning system, the planning result is presented in 3D form, and the gap after osteotomy is filled through the bone tissue filling module.

[0052] The mirror image model of the contralateral healthy bone is generated by obtaining the symmetric point coordinates based on the point coordinates of the contralateral healthy bone in the coordinate system, taking the XOZ plane as the symmetry plane, and using the symmetric point coordinates to establish the geometric structure and topological structure of the mirror image model.

[0053] In S1, the method for constructing the local coordinate system is:

[0054] S101, establishing the Y axis of the local coordinate system based on the long axis of the bone, and rotating around the Y axis to correspond to the corrected transverse plane;

[0055] S102, based on the principle that the rotation around the Z axis corresponds to the correction of the sagittal plane, using the established Y axis, establish the Z axis of the local coordinate system;

[0056] S103, based on the principle of rotating around the X-axis corresponding to the corrected frontal plane, and using the vector product result of the established Y-axis and Z-axis, establish the X-axis of the local coordinate system.

[0057] In S2, the anatomical points of the distal joint are seven landmark points, and fifty points are searched within the 50mm area of ​​each anatomical point using the established topological structure relationship, using a breadth-first search algorithm, and finally generating the deformed bone model and the reconstructed target model to perform the point set to be aligned and the target point set.

[0058] The topological structure relationship of the deformed bone model is established through the existing point and surface unit data of the STL model, according to the principle that all surface units composed of a point constitute all adjacent surface units of the point, all line units composed of a point constitute all adjacent line units of the point, line units are adjacent to one or two surface units, and surface units are adjacent to one, two or three surface units, to establish the three-dimensional model topological structure relationship in the 3D osteotomy pre-operative intelligent planning system.

[0059] In S3, the two precise registrations are both performed using the nearest iterative point method. The first precise registration is the initial rough registration by selecting anatomical points near the joint as landmark points; the second precise registration is the precise registration by selecting the roughly registered distal part point set.

[0060] In S4, the method for automatically diagnosing the deformed area of ​​the deformed bone includes:

[0061] S401, establishing a window function, and performing spatial discretization processing on the point set of the distally registered deformed model and the reconstructed target model along the Y axis of the local coordinate system of the deformed bone model;

[0062] S402, each discrete part includes a point set of the deformed bone model and the reconstructed target model, calculates and searches for a point with the smallest Euclidean distance between each point in the deformed bone model point set and the reconstructed target model point set, and calculates a root mean square error (RMSE);

[0063] S403, using the point on the Y-axis of the local coordinate system of the deformed bone model as the horizontal coordinate and the root mean square error as the vertical coordinate, a deformity function image is established, and the deformity area is determined by a deviation of 10% from the minimum root mean square error.

[0064] The window function uses the local coordinate system as a reference, the X value and the Y value surround the deformed bone model and the healthy bone model, and the Y value is a discrete range.

[0065] In S6,

[0066] The parameters in the optimization algorithm are transformed according to the clinical goals in osteotomy correction surgery, and all the parameters of the four clinical goals are encoded into the chromosomes of the multi-stage multi-objective optimization algorithm;

[0067] In the weight-based multi-stage multi-objective optimization algorithm, weight allocation is to change the calculation method of congestion in the ordinary multi-objective optimization algorithm. When calculating congestion, weight allocation is performed according to the importance of the current optimization goal, so that the optimized parameters are forced to remain in the ideal point space close to the smaller congestion.

[0068] The multi-stage in the weight-based multi-stage multi-objective optimization algorithm is to divide the four optimization objectives into two stages, use the optimization result of stage one to initialize stage two, and finally perform preoperative planning based on the parameters of the optimization result of stage two.

[0069] A 3D osteotomy orthopedic preoperative intelligent planning system, used to implement a 3D osteotomy orthopedic preoperative intelligent planning method, and

[0070] Model space transformation module, used to adjust the spatial position of the skeleton model before intelligent planning;

[0071] Model registration module, which provides landmark-based, nearest iteration point ICP, four-point registration and fast point feature histogram registration methods, and is used to register the deformed bone model locally or as a whole with the reconstructed target model according to different situations;

[0072] See also Figure 2 , bone deformity automatic diagnosis module, used to automatically calculate the deformity range of the deformed bone model compared with the reconstructed target model and draw a deformity function image;

[0073] The model cutting module is used to cut the deformed bone within a rough range after alignment and automatic diagnosis.

[0074] The bone filling model generation module is used to generate a bone fragment gap filling model generated after osteotomy restoration to assist in the fixation of bone fragments.

[0075] The osteotomy correction optimization module is used to generate different proximal and distal potential bone fragments as input within the parameter range of the osteotomy plane, and use a weight-based multi-stage multi-objective optimization algorithm to calculate the optimal osteotomy cutting surface, restore the alignment transformation matrix, and the spatial position of the fixation plate and screws.

[0076] Embodiment 1

[0077] The healthy bone on the opposite side of the deformed bone is imported into the system and presented in the form of an STL model. The XOZ plane of the three-dimensional world coordinate system is set as the symmetry plane, and the symmetric points of all coordinate points of the healthy bone on the opposite side about the plane are obtained. Then, these points are used to establish the geometric structure and topological structure of the mirror model to complete the establishment of the reconstructed target model.

[0078] The deformed bone model and the reconstructed target model are simultaneously imported into the system, the distal end of the bone model is registered, the deformity function image is drawn, and the deformation area of ​​the deformed bone model is determined.

[0079] First, because the alignment of the joint area is more important than the alignment of the long bone area in osteotomy correction, according to the surgeon's advice, five clinical anatomical points were selected on the distal articular surface of the deformed bone model and the reconstructed target model, and two clinical anatomical points were selected in the long bone area around the articular surface, for a total of seven clinical anatomical points as registration landmark points.

[0080] The topological relationship of the STL model is established, including all the units that each point participates in constituting, all the lines that each point participates in constituting, two adjacent units on a line, and three units connected to each unit.

[0081] Using topological information, for each landmark point, a breadth-first search algorithm is used to search for 50 points within a 5mm area. If there are no 50 points within the 5mm neighborhood, all points within the neighborhood are searched. The search point set of the deformed bone model is used as the source point set, and the search point set of the reconstructed target model is used as the target point set. The source point set and the target point set are aligned using the nearest iteration point algorithm. Since the alignment result obtained based on the point set selected based on the landmark is only an initial alignment result based on the joint area, it cannot completely align the distal healthy part with the reconstructed target bone. Therefore, through experiments, 30% of the points at the distal end of the deformed bone model and the reconstructed target model are selected and re-aligned using the nearest iteration method based on the initial alignment, which can preliminarily align the healthy part at the distal end of the deformed bone with the reconstructed target bone.

[0082] The steps for selecting 30% of the point set are as follows:

[0083] Calculate the point P where the deformed point set has the smallest projection on the Y axis of the local coordinate system min and the largest point P max ;

[0084] Calculate the local coordinate system Y axis and P min and P max The point L with the smallest Euclidean distance min and L max ;

[0085] Calculate the distance point L min The length of max -L min ) point L(λ);

[0086] When λ is 0.3, the constraint conditions are calculated:

[0087] Constraint 1: less than the projection of L(0.3) on the Y axis of the local coordinate system;

[0088] Constraint 2: Less than L(0.3) and L min distance;

[0089] Search the projection on the Y axis of the local coordinate system in the deformed point set to satisfy the constraint condition 1 and to match L min The points whose distance satisfies constraint 2 constitute 30% of the point set of the distal end of the deformed bone.

[0090] After completing the registration using the distal 30% point set, the distal deformity area can be calculated using the registered model, as detailed below:

[0091] The calculation of the deformed area is achieved by setting a window function and moving the window function along the Y-axis of the local coordinate system of the deformed bone model to achieve the purpose of discretizing the deformed bone model and reconstructing the target model point set.

[0092] First, set the window function. The essence of the window function is a spatial range. Take the local coordinate system of the deformed bone as the reference coordinate system, set the x and z values ​​to be large enough to include the deformed bone model and the reconstructed target model. The y value of the window is the thickness of the window. Different spatial ranges can be controlled by selecting different window thicknesses, thereby controlling the deformity contour with different accuracies.

[0093] Each window includes the deformed bone model and the reconstructed target model point set in the same y value range. The point with the smallest Euclidean distance between each point in the deformed bone model point set and the reconstructed target model point set is found through traversal method, and their root mean square error RMSE is calculated.

[0094] Then, the RMSE is continuously calculated according to the window thickness from the farthest to the most proximal end of the deformed bone model, and the deformity contour function graph is drawn with the Y-axis of the local coordinate system of the deformed bone as the horizontal axis. The specific window thickness is determined according to the surgeon's advice.

[0095] Finally, the point with the minimum RMSE in the deformity contour map was found, and from this point, the position that deviated by 10% from this value was identified as the approximate starting point of the distal deformity.

[0096] After determining the deformed contour area, the range of a cutting plane can be roughly determined. However, the specific point where the cutting can make the restoration alignment and the fixing plate installation most reasonable requires the optimization of the cutting surface position. The optimization process in the present invention is achieved by a multi-stage multi-objective optimization algorithm based on weights, such as Figure 3 As shown, the following steps are included:

[0097] B1. In the multi-stage multi-objective optimization algorithm, a quantitative method is needed to judge the degree of conformity between the optimization results and the clinical goals. Therefore, a fitness function applicable to the algorithm is formulated for each clinical goal;

[0098] B2. According to the importance of different clinical goals in clinical osteotomy and correction surgery, a method for weighting clinical goals in the algorithm is developed;

[0099] B3. Execute a multi-stage multi-objective optimization algorithm based on clinical goal weights in stages. The first stage optimizes the three clinical goals of osteotomy plane, restoration of alignment, and spatial position of the fixation plate. The second stage optimizes the three goals of restoration of alignment, fixation plate position, and screw position based on the first stage. Finally, the best optimization parameters are selected based on the optimal solution for the spatial position of the fixation plate.

[0100] The detailed instructions are as follows:

[0101] First, before performing the optimization process, it is necessary to consider how to evaluate the optimization parameters. Here, the present invention develops fitness functions corresponding to four clinical objectives based on clinically relevant measurements, which are as follows:

[0102] Fitness function for repositioning and alignment: As explained in the invention content, the joint area needs to be more accurately aligned with the reconstruction target than other areas. The points on the joint surface are given a weight of 0.18, and the points outside the joint are given a weight of 0.05. The root mean square error of a total of 350 points within the seven landmark points of the deformed bone is calculated based on the weights, and the fitness function for repositioning and alignment is determined accordingly.

[0103] Fitness function of osteotomy plane: It is extremely difficult to directly control the osteotomy plane. In this invention, the osteotomy plane is implicitly controlled by controlling the bone protrusion. The bone protrusion means that after repositioning and alignment, a window function with a sufficiently large window thickness is used to enclose the cutting position at the osteotomy plane, and the average RMSE value of the distal fragment and the proximal fragment at this point with the reconstructed target bone is calculated respectively, and the fitness function of the osteotomy plane is formulated accordingly.

[0104] Fitness function of fixation plate orientation: First, the average bone model is generated based on the surgical approach and the intrinsic anatomy of the bone using the statistical shape model (SSM) method and the average fixation plate feasible area is determined. For a specific patient model, a model fitting registration algorithm is used to obtain a specific fixation plate feasible area.

[0105] Then, the point set on the fixation plate that should be in contact with the bone surface is determined, and the intrinsic coordinate system of the fixation plate is calculated by the principal component analysis method. The calculation method is as follows:

[0106] The covariance matrix is ​​constructed using the three-dimensional coordinates of the fixed plate in the world coordinate system, and the eigenvalues ​​and eigenvectors are calculated. Each eigenvector represents the direction of a coordinate axis of the intrinsic coordinate system. The positive direction of the Z axis is specified to point to the lower surface of the fixed plate. The origin of the coordinate system is determined by the mean of all points on the fixed plate.

[0107] The judgment condition is that the projection of the vector formed by a point on the fixation plate and the origin of the fixed coordinate system in the positive direction of the Z axis is less than 0.85 times the thickness of the fixation plate. The points that meet this condition are used as the point set that should contact the bone surface. This point set is used to align and contact the surface of the deformed bone fragment, and then the average distance between the fixation plate and the distal fragment and the proximal end is calculated respectively through the distance between the points. Finally, the average distance between the two is calculated, and the fitness function of the fixation plate orientation is defined accordingly.

[0108] Fitness function of screw orientation: Since inappropriate screw position and length will cause various complications in patients, the relationship between screw stability and bone density distribution is needed. According to this principle, we set the starting point and end point according to the two intersection points of the screw bone at the screw placement point, sample along the starting point and end point of the screw in three-dimensional space and calculate the bone density of different sampling points, and finally calculate the average value to define the fitness function of the screw orientation.

[0109] The above four fitness functions are represented by f1, f2, f3, and f4 respectively. After the four fitness functions are formulated, the main part of the multi-stage multi-objective algorithm based on clinical objective weights is implemented as follows:

[0110] First, the optimization parameters of the four clinical objectives are encoded into chromosomes in decimal code.

[0111] Ordinary multi-stage multi-objective genetic algorithms can only find uniform optimal solutions for equal-weight optimization objectives in the solution space. However, in actual clinical surgery, the importance of different clinical objectives is different, so a weight calculation model needs to be formulated. In the multi-stage multi-objective genetic algorithm, the crowding degree represents the importance of the solution in the same Pareto level solution set. The smaller the crowding degree, the more important the solution. So the weight distribution formula is Used for crowding calculation, the importance of different clinical goals is determined by weight-based crowding.

[0112] In the first stage, the reduction alignment (f1), osteotomy plane (f2), and fixation plate position (f3) were optimized, and the weights k1 = 0.2, k2 = 0.3, and k3 = 1 were assigned. The purpose of setting k3 to 1 is to force the algorithm to keep the solution set in the solution space close to the ideal point of f1 and f2 in the first stage.

[0113] At this stage, the chromosome 200 chromosomes are randomly generated within the parameter range of , and at each stage, all parameters (all objectives) are optimized, but only a subset of the fitness function is considered.

[0114] Iterate 200 times and finally take the 70 best solutions in the top Pareto column.

[0115] In the second stage, based on the 70 solutions in the first stage, 200 chromosomes are initialized within this parameter range, and the recovery alignment (f1), the fixed plate position (f3) and the screw position (f4) are optimized, with a weight k3 = 0.4. The other optimization objectives are not weighted. This is to ensure that the solution has an optimal fixed plate alignment position and avoid deteriorating the recovery alignment. After 200 iterations, only the 70 best solutions in the top Pareto column are saved and stored in XR 2 middle.

[0116] The optimized parameterized chromosome finally determined by the algorithm It is obtained by taking the best fixed plate alignment solution in the sorted solution set of the second stage, and the calculation formula is as follows:

[0117]

[0118] After the optimization is completed, the parameters for preoperative planning of osteotomy correction are obtained in the parameterized chromosome, which are the osteotomy plane for osteotomy: position parameter pl x ,pl y ,pl z , normal parameter Restore the aligned 4x4 homogeneous transformation matrix A f : Rotation matrix Translation Vector The 4x4 homogeneous transformation matrix of the fixation plate relative to the local coordinate system of the deformed bone: rotation matrix R p =(θ x ,θ y ,θ z ), the translation vector Screw position: a vector consisting of Euler angles of rotation around the coordinate axis i indicates the number of screws.

[0119] Through the above-mentioned preoperative planning parameters, the 3D model after osteotomy correction can be automatically generated in the 3D osteotomy correction preoperative intelligent planning system, and the doctor can use this result to perform clinical surgery.

[0120] The above-mentioned embodiments only express the implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be based on the attached claims.

Claims

1. A 3D osteotomy preoperative intelligent planning method, characterized in that: The following steps are involved: S1. Generate STL models of the deformed bone and the contralateral healthy bone using CT data of the deformed bone and the contralateral healthy bone, and import them into the 3D osteotomy orthopedic preoperative intelligent planning system. Generate a mirror image model of the contralateral healthy bone by selecting a reference surface for coordinate symmetry. Use the mirror image model as a reconstruction target model, and establish a local coordinate system for the deformed bone model and the reconstruction target model. S2, adjusting the initial postures of the deformed bone model and the reconstructed target model through the model space change module of the 3D osteotomy orthopedic preoperative intelligent planning system; and selecting landmark points at the distal joints of the deformed bone model and the reconstructed target model and searching for domain point sets; S3, using the model registration module of the 3D osteotomy orthopedic preoperative intelligent planning system, using the search field point set in S2 to perform a primary registration based on landmark points, and selecting the point set of the deformed bone model and the distal 30% of the reconstructed target model for a secondary precise registration; S4, based on the precise registration in S3, generating a deformity function image of the deformed bone model through the bone deformity automatic diagnosis module, and determining the deformity area and the cutting range of the deformed bone; S5, cutting the deformed bone using the model cutting module according to the cutting range obtained in S4; S6, using the parameter range of the osteotomy plane generated by the cutting range as the initial input of the optimization algorithm, and calculating the optimal preoperative planning parameters through the weight-based multi-stage multi-objective optimization algorithm in the osteotomy correction optimization module; S7. Based on the optimization result of S6, preoperative planning is performed through the 3D osteotomy orthopedic preoperative intelligent planning system, the planning result is presented in 3D form, and the gap after osteotomy is filled through the bone tissue filling module.

2. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: The mirror image model of the contralateral healthy bone is generated by obtaining the symmetric point coordinates based on the point coordinates of the contralateral healthy bone in the coordinate system, taking the XOZ plane as the symmetry plane, and using the symmetric point coordinates to establish the geometric structure and topological structure of the mirror image model.

3. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: In S1, the method for constructing the local coordinate system is: S101, establishing the Y axis of the local coordinate system based on the long axis of the bone, and rotating around the Y axis to correspond to the corrected transverse plane; S102, based on the principle that the rotation around the Z axis corresponds to the correction of the sagittal plane, using the established Y axis, establish the Z axis of the local coordinate system; S103, based on the principle of rotating around the X-axis corresponding to the corrected frontal plane, and using the vector product result of the established Y-axis and Z-axis, establish the X-axis of the local coordinate system.

4. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: In S2, the anatomical points of the distal joint are seven landmark points, and fifty points are searched within the 50mm area of ​​each anatomical point using the established topological structure relationship, using a breadth-first search algorithm, and finally generating the deformed bone model and the reconstructed target model to perform the point set to be aligned and the target point set.

5. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 4, characterized in that: The topological structure relationship of the deformed bone model is established through the existing point and surface unit data of the STL model, according to the principle that all surface units composed of a point constitute all adjacent surface units of the point, all line units composed of a point constitute all adjacent line units of the point, line units are adjacent to one or two surface units, and surface units are adjacent to one, two or three surface units, to establish the three-dimensional model topological structure relationship in the 3D osteotomy pre-operative intelligent planning system.

6. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: In S3, the two precise registrations are both performed using the nearest iterative point method, and the first precise registration is the initial rough registration by selecting anatomical points near the joint as landmark points; The second precise registration is to select the roughly registered distal part point set for precise registration.

7. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: In S4, the method for automatically diagnosing the deformed area of ​​the deformed bone includes: S401, establishing a window function, and performing spatial discretization processing on the point set of the distally registered deformed model and the reconstructed target model along the Y axis of the local coordinate system of the deformed bone model; S402, each discrete part includes a point set of the deformed bone model and the reconstructed target model, calculates and searches for a point with the smallest Euclidean distance between each point in the deformed bone model point set and the reconstructed target model point set, and calculates a root mean square error (RMSE); S403, using the point on the Y-axis of the local coordinate system of the deformed bone model as the horizontal coordinate and the root mean square error as the vertical coordinate, a deformity function image is established, and the deformity area is determined by a deviation of 10% from the minimum root mean square error.

8. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 7, characterized in that: The window function uses the local coordinate system as a reference, the X value and the Y value surround the deformed bone model and the healthy bone model, and the Y value is a discrete range.

9. The method for intelligent planning before 3D osteotomy and correction surgery according to claim 1, characterized in that: In S6, The parameters in the optimization algorithm are transformed according to the clinical goals in osteotomy correction surgery, and all the parameters of the four clinical goals are encoded into the chromosomes of the multi-stage multi-objective optimization algorithm; In the weight-based multi-stage multi-objective optimization algorithm, weight allocation is to change the calculation method of congestion in the ordinary multi-objective optimization algorithm. When calculating congestion, weight allocation is performed according to the importance of the current optimization goal, so that the optimized parameters are forced to remain in the ideal point space close to the smaller congestion. The multi-stage in the weight-based multi-stage multi-objective optimization algorithm is to divide the four optimization objectives into two stages, use the optimization result of stage one to initialize stage two, and finally perform preoperative planning based on the parameters of the optimization result of stage two.

10. A 3D osteotomy preoperative intelligent planning system, characterized in that: Used to implement the 3D osteotomy preoperative intelligent planning method as described in any one of claims 1 to 9, and Model space transformation module, used to adjust the spatial position of the skeleton model before intelligent planning; Model registration module, which provides landmark-based, nearest iteration point ICP, four-point registration and fast point feature histogram registration methods, and is used to register the deformed bone model locally or as a whole with the reconstructed target model according to different situations; The bone deformity automatic diagnosis module is used to automatically calculate the deformity range of the deformed bone model compared with the reconstructed target model and draw a deformity function image; Model cutting module, used to cut the deformed bone within a rough range after registration and automatic diagnosis; A bone filling model generation module is used to generate a bone fragment gap filling model generated after osteotomy restoration to assist in the fixation of bone fragments; The osteotomy correction optimization module is used to generate different proximal and distal potential bone fragments as input within the parameter range of the osteotomy plane, and use a weight-based multi-stage multi-objective optimization algorithm to calculate the optimal osteotomy cutting surface, restore the alignment transformation matrix, and the spatial position of the fixation plate and screws.

Images