Three-dimensional skeleton model adaptive optimization method based on anatomical features and electronic device

Abstract

The invention discloses a skeleton three-dimensional model adaptive optimization method based on anatomical features and an electronic device, relates to the technical field of medical images, and solves the technical problem that in the prior art, a complex skeleton curved surface cannot be comprehensively captured by using pure linear combination on all sample models, so that the misjudgment risk is increased. The method comprises the following steps: acquiring CT data of a skeleton to establish a skeleton three-dimensional model; identifying and extracting preset key anatomical feature points in the three-dimensional skeleton model; obtaining an initial error value; optimizing the skeleton three-dimensional model based on the key anatomical features; converting the optimized three-dimensional skeleton model into target image data, and calculating an average error value of contour lines in the target image data; performing secondary optimization on the skeleton three-dimensional model according to the size relationship between the average error value and the initial error value to obtain a smooth secondary skeleton three-dimensional model with the minimum mesh number within the initial error value; according to the method, the visualization effect of the skeleton model during surgical navigation is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the field of medical imaging and relates to a three-dimensional skeleton extraction and modeling technology, specifically a method and an electronic device for adaptively optimizing a three-dimensional skeleton model based on anatomical features. Background Art

[0002] In the current context, the sample models obtained through linear combination limit the diversity of bone three-dimensional model construction; if the sample library is not comprehensive enough or the morphological differences between samples are large, the linear combination may not be able to capture complex shape changes, resulting in an inaccurate model. Especially for the complex geometric structures of bones, such as articular surfaces or bone ridges, simple linear combinations may not be sufficient to express their morphological characteristics.

[0003] The prior art (invention patent application number CN116452755B) discloses a skeleton model construction method, system, medium and equipment, which includes: obtaining a sample model and marking sample feature points; selecting target feature points and regional points on a real skeleton; transforming the target feature points and regional points, and aligning the target feature points with the sample feature points; formulating a strategy, assigning an influencing factor to the sample model, linearly combining all sample models, constructing an initial model, and determining the initial feature points corresponding to the target feature points in the initial model; adjusting the strategy according to the distance between the initial feature points and the target feature points, selecting the strategy corresponding to the minimum distance as the optimal strategy, and determining the optimal initial model; determining the matching points corresponding to each regional point in the optimal initial model, calculating the transformation relationship, and transforming all points of the optimal initial model according to the transformation relationship to obtain the target skeleton model;

[0004] The existing technology cannot fully capture complex bone surfaces by using simple linear combinations of all sample models, thereby increasing the error rate of joint surface curvature and increasing the risk of misjudgment.

[0005] The present invention provides a method and an electronic device for adaptively optimizing a three-dimensional skeleton model based on anatomical features to solve the above technical problems. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an adaptive optimization method and electronic device for a three-dimensional bone model based on anatomical features, which is used to solve the technical problem that the prior art cannot fully capture complex bone surfaces by using a simple linear combination of all sample models, thereby increasing the error rate of the articular surface curvature and increasing the risk of misjudgment.

[0007] To achieve the above object, a first aspect of the present invention provides a method for adaptively optimizing a skeletal three-dimensional model based on anatomical features, comprising:

[0008] S100: Acquire CT data of the bone to establish a three-dimensional bone model; identify and extract preset key anatomical feature points in the three-dimensional bone model;

[0009] S200: Optimize the 3D bone model based on key anatomical features;

[0010] S300: Obtain the initial error value of the skeleton three-dimensional model; convert the optimized skeleton three-dimensional model into target image data, and calculate the average error value of the contour line in the target image data; perform secondary optimization on the skeleton three-dimensional model according to the relationship between the average error value and the initial error value, and obtain a secondary skeleton three-dimensional model with the least number of meshes and the smoothest quality within the initial error value.

[0011] Preferably, the step of acquiring CT data of bones and establishing a three-dimensional bone model comprises:

[0012] S110: Obtain bone imaging data using computed tomography technology, with preset tomography thickness;

[0013] S121: Retrieve several cells obtained by the medical image segmentation algorithm, and obtain the cell vertex state code based on the positional relationship between the cell vertex and the bone area:

[0014] S122: Obtain the intersection point by linear interpolation technology according to the cell vertex state code;

[0015] S123: Determine the edge line to be drawn at the intersection point according to a predefined MarchingSquares algorithm lookup table, and use the edge line drawn at the intersection point as the contour line;

[0016] S130: Reconstructing the segmented contour lines using a marching cube algorithm in an isosurface three-dimensional reconstruction algorithm to obtain a three-dimensional skeleton model.

[0017] It should be noted that the value range of the tomographic thickness is (0, 1 mm).

[0018] The present invention sets the tomographic thickness within the range of (0, 1 mm), which means that an extremely thin image layer can be obtained, thereby retaining more detailed information, which helps to more accurately restore the fine structure of the bone in subsequent three-dimensional reconstruction.

[0019] Preferably, the optimizing the three-dimensional skeleton model based on key anatomical features comprises:

[0020] Retrieving the skeletal 3D model, adjusting the mesh structure of the skeletal 3D model based on a triangulation re-meshing algorithm, and generating a skeletal 3D model with a consistent mesh size;

[0021] Retrieve key anatomical feature points, calculate the positions of the closest projection points of the key anatomical feature points on the re-meshed 3D bone model, and record the triangular facets corresponding to the closest projection points;

[0022] Determine whether to add control points based on the positional relationship between the nearest projection point and the triangle corresponding to the nearest projection point; if yes, add control points to the corresponding triangle; if no, use the key anatomical feature points as the vertices of the triangle;

[0023] The triangular facets are subjected to secondary triangulation to readjust the mesh structure of the skeleton three-dimensional model, and a smoothing algorithm is used to smooth the readjusted mesh structure of the skeleton three-dimensional model to obtain an optimized skeleton three-dimensional model.

[0024] The present invention can generate a 3D skeleton model with consistent mesh size through a remeshing algorithm based on triangulation, which helps to ensure that the model has similar accuracy and resolution in different areas, facilitating subsequent analysis and processing.

[0025] Preferably, the calculating of the closest projection point position of the key anatomical feature point on the re-gridded three-dimensional skeleton model includes:

[0026] The position of the i-th key anatomical feature point is recorded as x i , through the formula Calculate the polygon mesh surface S with x i The Euclidean distance to the nearest projection point position y * ; In the formula, y represents x i At the projection point of each triangle; the value range of i is a positive integer;

[0027] By the formula d(x)=sign[n(y * )·(x i -y * )]·||x i -y * ||Calculate the symbol distance; where n(y * ) is y * The surface normal vector at

[0028] Determine whether the sign of d(x) is positive; if it is, then the i-th key anatomical feature point x i On the side pointed by the normal vector; otherwise, the i-th key anatomical feature point x i Located inside the 3D model of the skeleton.

[0029] It should be noted that the i-th key anatomical feature point x i The side pointed by the normal vector is the i-th key anatomical feature point x i Outside the 3D model of the skeleton

[0030] Preferably, the step of determining whether to add a control point based on the positional relationship between the nearest projection point and the triangle corresponding to the nearest projection point includes:

[0031] Determine whether the nearest projection point of the key anatomical feature point is within the triangle;

[0032] If yes, then add key anatomical feature points as control points;

[0033] If not, determine whether the nearest projection point of the key anatomical feature point is on the boundary line of the triangle; if yes, add the key anatomical feature point as a control point; if not, mark it as not necessary to add the key anatomical feature point as a control point.

[0034] The present invention can further optimize the shape and details of the model by determining whether to add control points based on the positional relationship between the nearest projection point and the corresponding triangular facet; at the same time, adding control points can enhance the local control ability of the model, making the three-dimensional bone model closer to the real anatomical structure.

[0035] Preferably, the smoothing process of the skeletal three-dimensional model having the re-adjusted mesh structure using a smoothing algorithm comprises:

[0036] Retrieve the position v of the kth vertex in the 3D skeleton model after re-adjusting the mesh structure k and the position v of the jth vertex j ;

[0037] By formula The smoothed position of the k-th vertex is calculated, where N(k) represents the set of adjacent vertices of vertex k; N represents the number of adjacent vertices, λ represents the smoothing weight, and λ∈[0,1].

[0038] The present invention uses a smoothing algorithm to smooth the skeletal three-dimensional model with a readjusted mesh structure, thereby eliminating jagged edges or uneven areas that may be generated by re-meshing, which is beneficial to improving the visual quality of the model and making it more suitable for subsequent rendering and visualization operations.

[0039] Preferably, converting the optimized skeletal three-dimensional model into target image data comprises:

[0040] S311: Preset the resolution (Nx, Ny, Nz) and voxel spacing (Δx, Δy, Δz) of the target image data;

[0041] By formula p ijk =(x min +c1·Δ x ,y min +c2·Δy , z min +c3·Δ z ) calculates the coordinates of the voxel center point; where c1∈[0, N x -1],c2∈[0,N y -1],c3∈[0,N z -1]; c1 represents the pixel number in the x-axis direction, c2 represents the pixel number in the y-axis direction, and c3 represents the pixel number in the z-axis direction.

[0042] S312: Acquire a plurality of polygons formed by the intersection of contour lines in the three-dimensional model of the skeleton;

[0043] By formula Calculate the mask;

[0044] S313: By formula Calculate and obtain target image data.

[0045] Preferably, the calculating the average error value of the contour lines in the target image data includes:

[0046] S321: Retrieve a number of contour lines and calculate the difference between adjacent contour lines corresponding to each layer in the target image data;

[0047] S322: Calculate the average of the calculated error values of all layers to obtain the average error value of the optimized skeletal three-dimensional model.

[0048] The adjacent contour lines corresponding to each layer are recorded as A and B respectively;

[0049] By formula The average error value is calculated, where a represents the ath point on contour line A and b represents the bth point on contour line B.

[0050] The present invention calculates the error value by using the average symmetric surface distance, calculates the average closest distance from A to B and from B to A, and then takes the average of the two, which is beneficial to measuring the average point-to-point distance between the two contour lines and reflecting the overall degree of fit.

[0051] Preferably, the secondary optimization of the skeleton three-dimensional model according to the magnitude relationship between the average error value and the initial error value comprises:

[0052] Retrieve the initial error value and determine whether the average error value is greater than the initial error value. If so, expand the normal direction of each vertex in the mesh of the skeleton 3D model by half of the average error value based on the average error value; if not, simplify the mesh of the skeleton 3D model according to a preset ratio.

[0053] Retrieve the mesh-transformed skeletal three-dimensional model, and jump to steps S311-S322 to recalculate the quadratic average error value; determine whether the quadratic average error value is less than or equal to the initial error value; if so, perform a quadratic mesh simplification on the skeletal three-dimensional model according to a preset ratio; if not, based on the quadratic average error value, further expand the normal direction of each vertex in the mesh of the skeletal three-dimensional model by half of the quadratic average error value;

[0054] Until the error value of the contour line is within the initial error value, the number of meshes is minimal and the secondary skeleton three-dimensional model is smooth.

[0055] To achieve the above-mentioned purpose, the second aspect of the present invention provides an electronic device for adaptive optimization of bone three-dimensional models based on anatomical features, comprising: a memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute the method for adaptive optimization of bone three-dimensional models based on anatomical features provided in the first aspect by executing the executable instructions.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The present invention can greatly improve the visualization effect of the bone model during surgical navigation by minimizing the data volume of the three-dimensional bone model and maintaining an optimal smoothing effect while keeping the anatomical features unchanged. By obtaining the CT data of the bone and identifying key anatomical feature points, it can ensure that the model is highly consistent with the real bone in terms of anatomical structure. The acquisition of initial error values and optimization based on key anatomical features help to reduce the difference between the model and the real bone. The secondary optimization process aims to reduce the number of grids while maintaining the smoothness and accuracy of the model, which helps to reduce the computational cost of model processing and improve rendering speed and interactive performance.

[0058] 2. The present invention uses a model with contour line errors within the initial error value and the minimum number of grids, which is more suitable for various application scenarios, such as medical education, surgical planning, clinical diagnosis, etc.; the CT-MRI registration error is reduced by target image data conversion; the model complexity is reduced by 40-70% through an adaptive grid simplification algorithm; the sawtooth effect is reduced while maintaining the continuity of the model surface through intelligent smoothing processing, and the smooth model surface is beneficial to reduce visual noise and interference, and improve the readability and ease of use of the model; the present invention maximizes the realization of accurate and real-time visualization of bone changes during intraoperative navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0060] Figure 1 Schematic diagram of the process flow for model optimization of the present invention;

[0061] Figure 2 Schematic diagram of the specific steps for the initial optimization of the skeletal three-dimensional model of the present invention;

[0062] Figure 3 A model diagram of the three-dimensional skeleton model of the present invention;

[0063] Figure 4 Schematic diagram of the positional relationship between the closest projection point and the triangle facet where it is located in the present invention;

[0064] Figure 5 Schematic diagram of the specific steps of secondary optimization of the skeletal three-dimensional model of the present invention. DETAILED DESCRIPTION

[0065] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] See also Figure 1 The first embodiment of the present invention provides a method for adaptively optimizing a three-dimensional skeleton model based on anatomical features, comprising:

[0067] S100: Acquire CT data of the bone to establish a three-dimensional bone model; identify and extract preset key anatomical feature points in the three-dimensional bone model;

[0068] S200: Optimize the 3D bone model based on key anatomical features;

[0069] S300: Obtain the initial error value of the skeleton three-dimensional model; convert the optimized skeleton three-dimensional model into target image data, and calculate the average error value of the contour line in the target image data; perform secondary optimization on the skeleton three-dimensional model according to the relationship between the average error value and the initial error value, and obtain a secondary skeleton three-dimensional model with the least number of meshes and the smoothest quality within the initial error value.

[0070] See also Figure 2-Figure 4 , the specific steps for the initial optimization of the skeleton 3D model:

[0071] S110: Obtain bone imaging data using computed tomography technology, with preset tomography thickness;

[0072] S121: Retrieve several cells obtained by the medical image segmentation algorithm, and obtain the cell vertex state code based on the positional relationship between the cell vertex and the bone area:

[0073] S122: Obtain the intersection point by linear interpolation technology according to the cell vertex state code;

[0074] S123: Determine the edge line to be drawn at the intersection point according to a predefined MarchingSquares algorithm lookup table, and use the edge line drawn at the intersection point as the contour line;

[0075] S130: reconstructing the segmented contour lines using a marching cube algorithm in an isosurface 3D reconstruction algorithm to obtain a skeletal 3D model;

[0076] Identify and extract key anatomical feature points preset in the 3D bone model;

[0077] For example, the DICOM3.0 data of the human femur bone is obtained by computer tomography technology and recorded as image_Bone, and the thickness of the tomography is set to 1mm;

[0078] Each slice layer corresponding to the bone image data image_Bone is a binary image. binary Extract the vertex coordinates of the contour line (contour line). The algorithm is as follows:

[0079] S121: Get cell vertex state code: call several 2×2 pixel cells and generate a 4-bit binary code based on whether the vertex is within the skeleton area:

[0080] Among them, v0, v1, v2, and v3 are the four vertices of the cell;

[0081] S122: Linear interpolation calculation of intersection points: For edges that cross the boundary (set the vertex state to between 0 and 1), use linear interpolation to calculate the intersection coordinates:

[0082] Intersection point = (1-α)·p a +α·p b

[0083] Among them, v a , v b are the two endpoints of the edge, p a , p bare vertex coordinates, α is the interpolation weight, and α∈[0,1]; T represents the minimum pixel value in the bone region of the binary image, I(v a ) represents endpoint v a The pixel value, I(v b ) represents endpoint v b Pixel value of

[0084] S123: Connect the contour segments according to the predefined MarchingSquares algorithm lookup table, and connect the intersection points into contour lines;

[0085] The segmented contour lines are reconstructed by the marching cube algorithm (MarchingCubes) in the isosurface 3D reconstruction algorithm to obtain the skeletal 3D model model_Bone, as shown in the following example: Figure 3 (1) shown;

[0086] Extract key anatomical feature points: the rotation center of the femoral head, the greater trochanter, the lowest point of the medial distal femoral condyle, the highest point of the lateral distal femoral condyle, the highest point of the lateral femoral condyle, the highest point of the medial femoral condyle, the highest point of the medial posterior femoral condyle, and the highest point of the lateral posterior femoral condyle;

[0087] The initial error value of the bone 3D model is set to 1 / 10 of the tomographic thickness, that is, the initial error value is 0.1 mm. The initial error value can also be customized by the user according to the user's accuracy requirements.

[0088] The mesh structure of the skeleton 3D model model_Bone is adjusted based on the Delaunay triangulation remeshing algorithm to generate a skeleton 3D model with consistent mesh size, such as Figure 3 (2) shown;

[0089] The positional relationship between key anatomical feature points and the 3D bone model model_Bone is obtained through signed distance;

[0090] The position of the i-th key anatomical feature point is recorded as x i , through the formula Calculate the polygon mesh surface S with x i The Euclidean distance to the nearest projection point position y * ; In the formula, y represents x i At the projection point of each triangle; the value range of i is a positive integer;

[0091] Add the extracted key anatomical feature points keyPoints to the 3D skeleton model remeshModel_Bone, that is, add keyPoints to the corresponding triangles and re-triangulate them to obtain a new 3D skeleton model keyModel_Bone;

[0092] There are three situations in triangulation:

[0093] (1) If the nearest projection point of the key anatomical feature point is within the triangle, then add the key anatomical feature point as a control point and re-triangulate, such as Figure 4 (1)

[0094] (2) If the nearest projection point of the key anatomical feature point is near the vertex of the triangle, the key anatomical feature point is taken as the vertex, such as Figure 4 (2) shown;

[0095] (3) If the nearest projection point of the key anatomical feature point is on the edge of the triangle, then add the key anatomical feature point as a control point and re-triangulate, such as Figure 4 (3) shown;

[0096] The triangular facets are subjected to secondary triangulation to readjust the mesh structure of the skeleton three-dimensional model, and a smoothing algorithm is used to smooth the readjusted mesh structure of the skeleton three-dimensional model to obtain an optimized skeleton three-dimensional model.

[0097] See also Figure 3 , the specific steps of secondary optimization of the skeleton 3D model:

[0098] S311: Preset the resolution (Nx, Ny, Nz) and voxel spacing (Δx, Δy, Δz) of the target image data;

[0099] By formula p ijk =(x min +c1·Δ x ,y min +c2·Δ y , z min +c3·Δ z ) calculates the coordinates of the voxel center point; where c1∈[0, N x -1],c2∈[0,N y -1],c3∈[0,N z -1]; c1 represents the pixel number in the x-axis direction, c2 represents the pixel number in the y-axis direction, and c3 represents the pixel number in the z-axis direction.

[0100] S312: Acquire a plurality of polygons formed by the intersection of contour lines in the three-dimensional model of the skeleton;

[0101] By formula

[0102] S313: By formula Calculate and obtain target image data.

[0103] Preferably, the calculating the average error value of the contour lines in the target image data includes:

[0104] S321: Retrieve a number of contour lines and calculate the difference between adjacent contour lines corresponding to each layer in the target image data;

[0105] S322: Calculate the average of the calculated error values of all layers to obtain the average error value of the optimized skeletal three-dimensional model.

[0106] The adjacent contour lines corresponding to each layer are recorded as A and B respectively;

[0107] By formula The average error value is calculated, where a represents the ath point on contour line A and b represents the bth point on contour line B.

[0108] Retrieve the initial error value and determine whether the average error value is greater than the initial error value; if so, based on the average error value, expand the normal direction of each vertex in the mesh of the skeleton 3D model outward by half of the average error value; if not, simplify the mesh of the skeleton 3D model according to a preset ratio;

[0109] Retrieve the mesh-transformed skeletal three-dimensional model, and jump to steps S311-S322 to recalculate the quadratic average error value; determine whether the quadratic average error value is less than or equal to the initial error value; if so, perform a quadratic mesh simplification on the skeletal three-dimensional model according to a preset ratio; if not, based on the quadratic average error value, further expand the normal direction of each vertex in the mesh of the skeletal three-dimensional model by half of the quadratic average error value;

[0110] Until the error value of the contour line is within the initial error value, the number of meshes is minimal and the secondary skeleton three-dimensional model is smooth.

[0111] For example, the optimized three-dimensional bone model is converted into target image data, and the average error value of the contour line in the target image data is calculated to be 0.06 mm;

[0112] Since the average error value of 0.06mm is less than the initial error value of 0.1mm, the skeleton three-dimensional model is mesh simplified according to a preset ratio of 0.8 (the number of triangles after simplification is 80% of the previous number); the skeleton three-dimensional model after mesh expansion is smoothed using the Laplace smoothing algorithm, and the quadratic average error value is calculated again to be 0.11mm. 0.11mm is greater than 0.1mm, so based on the original error value of 0.11mm, the normal direction of each vertex in the mesh of the skeleton three-dimensional model is expanded outward by half of the error value; since the error value has exceeded the initial error value, the skeleton three-dimensional model before mesh simplification is selected, that is, the skeleton three-dimensional model corresponding to the average error value of 0.06mm.

[0113] It should be noted that if the error value is still within the acceptable error value range, the above process is continued to simplify the mesh; if the error value is not within the acceptable range after simplification, the previous result is a smooth skeletal 3D model with the least number of meshes.

[0114] An embodiment of the second aspect of the present invention provides an electronic device for adaptive optimization of a three-dimensional bone model based on anatomical features, comprising: a memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute the method for adaptive optimization of a three-dimensional bone model based on anatomical features provided in the first aspect by executing the executable instructions.

[0115] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0116] The working principle of the present invention is as follows: CT data of the bone is obtained to establish a three-dimensional bone model; key anatomical feature points preset in the three-dimensional bone model are identified and extracted; the three-dimensional bone model is optimized based on the key anatomical features; the optimized three-dimensional bone model is converted into target image data, and the average error value of the contour line in the target image data is calculated; the three-dimensional bone model is secondary optimized according to the relationship between the average error value and the initial error value, so as to obtain a secondary three-dimensional bone model with the least number of grids and the smoothest quality within the initial error value.

[0117] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for adaptive optimization of a three-dimensional skeleton model based on anatomical features, characterized in that: include: S100: Acquire CT data of bones and establish a three-dimensional bone model; Identify and extract key anatomical feature points preset in the 3D bone model; S200: Optimize the 3D bone model based on key anatomical features; S300: Obtaining an initial error value of the skeletal three-dimensional model; The optimized skeletal three-dimensional model is converted into target image data, and the average error value of the contour line in the target image data is calculated; the skeletal three-dimensional model is secondary optimized according to the relationship between the average error value and the initial error value, and a secondary skeletal three-dimensional model with the least number of meshes and the smoothest quality is obtained within the initial error value.

2. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 1, wherein: The step of acquiring CT data of the skeleton and establishing a three-dimensional skeleton model comprises: S110: Obtain bone imaging data using computed tomography technology, with preset tomography thickness; S120: calling the medical image segmentation algorithm of the nnUNet deep learning network to segment the bone image data according to a preset tomographic thickness to obtain a plurality of contour lines; S130: Reconstructing the segmented contour lines using a marching cube algorithm in an isosurface three-dimensional reconstruction algorithm to obtain a three-dimensional skeleton model.

3. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 1, wherein: The optimization of the three-dimensional bone model based on key anatomical features includes: Retrieving the skeletal 3D model, adjusting the mesh structure of the skeletal 3D model based on a triangulation re-meshing algorithm, and generating a skeletal 3D model with a consistent mesh size; Retrieve key anatomical feature points, calculate the positions of the closest projection points of the key anatomical feature points on the re-meshed 3D bone model, and record the triangular facets corresponding to the closest projection points; Determine whether to add control points based on the positional relationship between the nearest projection point and the triangle corresponding to the nearest projection point; if yes, add control points to the corresponding triangle; if no, use the key anatomical feature points as the vertices of the triangle; The triangular facets are subjected to secondary triangulation to readjust the mesh structure of the skeleton three-dimensional model, and a smoothing algorithm is used to smooth the readjusted mesh structure of the skeleton three-dimensional model to obtain an optimized skeleton three-dimensional model.

4. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 3, wherein: The calculating of the closest projection point position of the key anatomical feature point on the re-grid-divided three-dimensional skeleton model includes: The position of the i-th key anatomical feature point is recorded as x i , through the formula Calculate the polygon mesh surface S with x i The Euclidean distance to the nearest projection point position y * ; Where y represents x i At the projection point of each triangle; the value range of i is a positive integer; By the formula d(x)=sign[n(y * )·(x i -y * )]·||x i -y * ||Calculate the symbol distance; where n(y * ) is y * The surface normal vector at Determine whether the sign of d(x) is positive; if it is, then the i-th key anatomical feature point x i On the side pointed by the normal vector; otherwise, the i-th key anatomical feature point x i Located inside the 3D model of the skeleton.

5. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 3, wherein: The step of determining whether to add a control point based on the positional relationship between the nearest projection point and the triangle corresponding to the nearest projection point includes: Determine whether the nearest projection point of the key anatomical feature point is within the triangle; If yes, then add key anatomical feature points as control points; If not, determine whether the nearest projection point of the key anatomical feature point is on the boundary line of the triangle; if yes, add the key anatomical feature point as a control point; if not, mark it as not necessary to add the key anatomical feature point as a control point.

6. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 3, wherein: The smoothing algorithm is used to smooth the skeletal three-dimensional model of the re-adjusted mesh structure, including: Retrieve the position v of the kth vertex in the 3D skeleton model after re-adjusting the mesh structure k and the position v of the jth vertex j ; Through the formula The smoothed position of the k-th vertex is calculated, where N(k) represents the set of adjacent vertices of vertex k; N represents the number of adjacent vertices, λ represents the smoothing weight, and λ∈[0,1].

7. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 1, wherein: The step of converting the optimized three-dimensional skeleton model into target image data includes: S311: Preset the resolution (Nx, Ny, Nz) and voxel spacing (Δx, Δy, Δz) of the target image data; By formula p ijk =(x min +c1·Δ x ,y min +c2·Δ y , z min +c3·Δ z ) calculates the coordinates of the voxel center point; where c1∈[0, N x -1],c2∈[0,N y -1],c3∈[0,N z -1]; c1 represents the pixel number in the x-axis direction, c2 represents the pixel number in the y-axis direction, and c3 represents the pixel number in the z-axis direction. S312: obtaining a plurality of polygons formed by the intersection of contour lines in the three-dimensional model of the skeleton, determining the positional relationship between the voxel center point and the nearest polygon by using a ray intersection method, and obtaining a mask based on the positional relationship; S313: Obtain target image data by applying the mask to the skeletal three-dimensional model.

8. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 1, wherein: The calculating the average error value of the contour line in the target image data includes: S321: Retrieve a number of contour lines and calculate the difference between adjacent contour lines corresponding to each layer in the target image data; S322: Calculate the average of the calculated error values of all layers to obtain the average error value of the optimized skeletal three-dimensional model.

9. The method for adaptively optimizing a skeletal three-dimensional model based on anatomical features according to claim 8, wherein: The secondary optimization of the three-dimensional skeleton model according to the magnitude relationship between the average error value and the initial error value includes: Retrieve the initial error value and determine whether the average error value is greater than the initial error value. If so, expand the normal direction of each vertex in the mesh of the skeleton 3D model by half of the average error value based on the average error value; if not, simplify the mesh of the skeleton 3D model according to a preset ratio; Retrieve the mesh-transformed skeletal three-dimensional model, and jump to steps S311-S322 to recalculate the quadratic average error value; determine whether the quadratic average error value is less than or equal to the initial error value; if so, perform a quadratic mesh simplification on the skeletal three-dimensional model according to a preset ratio; if not, based on the quadratic average error value, further expand the normal direction of each vertex in the mesh of the skeletal three-dimensional model by half of the quadratic average error value; Until the error value of the contour line is within the initial error value, the number of meshes is minimal and the secondary skeleton three-dimensional model is smooth.

10. An electronic device for adaptively optimizing a three-dimensional bone model based on anatomical features, characterized in that: include: A memory and a processor, wherein the memory stores executable instructions of the processor; wherein the processor is configured to execute the anatomical feature-based three-dimensional bone model adaptive optimization method according to any one of claims 1 to 9 by executing the executable instructions.

Images