A method for intelligently identifying forearm bone landmark points

Abstract

The application discloses a forearm bone landmark point intelligent identification method and belongs to the field of forearm bone landmark point identification. A forearm bone end axis is determined from a forearm bone three-dimensional model, a parallel plane system perpendicular to the forearm bone end axis is established, a plane with the largest profile area where each plane in the parallel plane system intersects the forearm bone three-dimensional model is selected as a forearm bone end characteristic plane, a forearm bone end local coordinate system is established, an initial forearm bone landmark point is determined by using the forearm bone end characteristic plane and the forearm bone three-dimensional model, local optimization is performed on the initial forearm bone landmark point in a preset neighborhood range, and a final identified forearm bone landmark point is determined. The application combines a statistical method and a geometric method, has good stability and individual specificity, and is suitable for automatic identification of bone landmark points of forearm images with certain bone deformities.

Description

Technical Field

[0001] This invention relates to the field of bony landmark recognition, and in particular to an intelligent recognition method for forearm bony landmarks. Background Technology

[0002] The identification of bony landmarks is widely used in many fields such as orthopedics, biomechanics, morphology, anthropometrics, epidemiology, anthropology, and forensic medicine.

[0003] Currently, bone landmarks are mostly identified manually, but manual identification has the following disadvantages: (1) it is difficult to provide reproducible results, and there are certain differences within and between observers; (2) manual identification requires relevant training for the identifyers, which consumes manpower and resources; (3) manual annotation reduces the degree of automation of medical image recognition.

[0004] The identification of bony landmarks can be divided into statistical methods and geometric methods. Statistical methods have lower requirements for the segmentation accuracy of the skeletal model, but require a certain size of labeled dataset to build a statistical shape model. Landmarks identified based on statistical methods have good robustness, but they can only infer the approximate location of the landmarks and have low accuracy in identifying individual variability. Geometric methods consider the curvature of the bone surface and other geometric information, and combine them with prior anatomical knowledge to build a mathematical model of bony landmarks. However, they are limited by the large differences in individual bone structure and may converge to a local optimum rather than the ideal optimum.

[0005] Therefore, the current automatic identification methods have the following problems:

[0006] (1) Studies on bone landmarks, especially long bone landmarks, have mainly focused on the lower limbs, with less research on the upper limbs;

[0007] (2) Statistical methods require a large amount of data, and the identification results are the average results obtained from statistics, with poor individual representation; while the identification using geometric methods has high reproducibility, but robustness is difficult to guarantee and may converge to other geometric feature points.

[0008] (3) For forearm images with certain bony deformities, simply using statistical and geometric methods may fail to identify them.

[0009] In summary, there is currently no method for identifying bony landmarks in forearm images that is suitable for the forearm bones, has a limited sample dataset, and is applicable to images with certain bony deformities. Summary of the Invention

[0010] The purpose of this invention is to provide an intelligent identification method for forearm bone landmarks, applicable to the automatic identification of bone landmarks in forearm images with certain bony deformities.

[0011] To achieve the above objectives, the present invention provides the following solution:

[0012] A method for intelligent recognition of forearm bone landmarks includes:

[0013] Forearm bones were extracted from forearm CT images to construct a three-dimensional model of the forearm bones; the forearm bones include the radius and ulna.

[0014] The axis of the forearm bone ends is determined in the three-dimensional model of the forearm bones; the forearm bone ends include the distal radius and the proximal ulna.

[0015] Establish a system of parallel surfaces perpendicular to the axis of the forearm bone end, and determine the surface containing the largest contour area where each surface in the parallel surface system intersects with the three-dimensional model of the forearm bone as the feature plane of the forearm bone end.

[0016] A local coordinate system for the forearm bone end is established with the center of the contour of the feature plane of the forearm bone end as the origin;

[0017] Based on the local coordinate system of the forearm bone ends, the initial bony landmarks of the forearm are determined using the feature plane of the forearm bone ends and the three-dimensional model of the forearm bone.

[0018] The initial forearm bony landmarks are locally optimized within a preset neighborhood to determine the final identified forearm bony landmarks.

[0019] Optionally, the axis of the forearm bone end is determined in the three-dimensional model of the forearm bone, specifically including:

[0020] When the forearm bone is the radius, the three-dimensional model of the forearm bone is determined to be the three-dimensional model of the radius;

[0021] A parallel plane system was constructed along the longitudinal axis of the forearm CT image in the distal 1 / 3 of the radius;

[0022] Determine the contour formed by the three-dimensional model of the radius and each plane in the parallel plane system, as well as the center of the contour;

[0023] The least squares method is used to fit the contour centers of each plane in the parallel plane system to a spatial straight line, which serves as the distal radius axis.

[0024] Optionally, establishing a local coordinate system for the forearm bone end, with the center of the forearm bone end feature plane as the origin, specifically includes:

[0025] The center point of the radial bone's characteristic planar contour is taken as the origin O of the local coordinate system of the distal radius. R ;

[0026] The intersection of the distal axis of the ulna and the characteristic plane of the radius is denoted as P. U ;

[0027] by The direction is the X-axis, the Z-axis is the axis of the distal radius, and the positive direction of the Z-axis is the direction from the proximal end to the distal end. The Y-axis is determined by the right-hand rule to establish a local coordinate system for the distal radius.

[0028] Optionally, the step of determining initial forearm bony landmarks based on the local coordinate system of the forearm bone ends, using the characteristic planes of the forearm bone ends and the three-dimensional model of the forearm bones, specifically includes:

[0029] A distal radius platform template is constructed; the distal radius platform template is marked with radial landmarks; the radial landmarks include the Lister's tubercle of the radius, the dorsal endpoint of the radial styloid process, the palmar endpoint of the radial styloid process, and the palmar endpoint of the radial sigmoid notch.

[0030] Normalize the characteristic planar contour of the radius to obtain the distal radial plateau;

[0031] Based on the contour of the distal radius plateau, using the local coordinate system of the distal radius, the point with the largest distance from the origin in the contour of the distal radius plateau in the palmar and ulnar regions is obtained as the initial palmar endpoint of the radial sigmoid notch.

[0032] The initial palmar end of the radial sigmoid notch is rotated to a fixed angle to obtain the rotated distal radial plateau contour.

[0033] The iterative nearest point method was used to register the rotated distal radius plateau contour with the distal radius plateau template contour;

[0034] Select the points in the registered distal radius plateau contour that are closest to the Lister tubercle of the radius and the Lister tubercle of the distal radius plateau template, the points in the registered distal radius plateau contour that are closest to the dorsal endpoint of the radial styloid process and the dorsal endpoint of the radial styloid process of the distal radius plateau template, the points in the registered distal radius plateau contour that are closest to the palmar endpoint of the radial styloid process and the palmar endpoint of the radial styloid process of the distal radius plateau template, and the points in the registered distal radius plateau contour that are closest to the palmar endpoint of the radial sigmoid notch and the palmar endpoint of the radial sigmoid notch of the distal radius plateau template, and record them as the initial radial landmark points.

[0035] Optionally, the construction of the distal radius platform template specifically includes:

[0036] Multiple forearm bone samples were selected, and radial landmarks were marked on each forearm bone sample.

[0037] Construct a three-dimensional model of the radius for each forearm bone sample, and determine the distal radius axis in each three-dimensional model;

[0038] Establish a system of parallel planes perpendicular to the axis of the distal radius, and determine the plane containing the largest contour area where each plane in the parallel plane system intersects with the three-dimensional model of the radius as the feature plane of the distal radius.

[0039] Determine the projection points of the radial landmarks onto the distal radial feature plane and the contour points of the distal radial feature plane;

[0040] Calculate the area of ​​each distal radius feature plane contour and unify all areas;

[0041] Rotate the palmar end of the radial sigmoid notch on each radial distal feature plane with a uniform area to the same angle to obtain multiple rotated radial distal feature planes.

[0042] Interpolate the contour segments on the rotated distal radius feature plane contour between the projection point of the palmar end of the radial sigmoid notch and the projection point of the Lister tubercle of the radius, between the projection point of the Lister tubercle of the radius and the projection point of the dorsal end of the radial styloid process, between the projection point of the dorsal end of the radial styloid process and the projection point of the palmar end of the radial styloid process, and between the projection point of the palmar end of the radial styloid process and the projection point of the palmar end of the radial sigmoid notch, so that a fixed number of contour points are interpolated on each contour segment.

[0043] Calculate the relationship between each contour point and O R The distance and angle of the line connecting them;

[0044] Based on the contour points of all forearm bone samples and O R The distance and angle of the connecting lines are used to determine the average contour of the distal radius feature plane, which serves as a template for the distal radius platform.

[0045] Optionally, the step of performing local optimization on the initial forearm bony landmarks within a preset neighborhood to determine the finally identified forearm bony landmarks specifically includes:

[0046] With the initial radial landmark location as the origin, the X-axis as the normal direction, the Y-axis as the tangent direction, and the Z-axis as the same as the radial shaft axis, a local coordinate system for the landmark is established.

[0047] Determine the pre-defined neighborhood range along the tangential and longitudinal directions on the local coordinate system of the marker point;

[0048] Let the origin O R Move a predetermined distance towards the proximal radius, and denote this as the base point. And select a value relative to the base point within the preset neighborhood range. The outermost neighboring point in the direction connecting the origin of the local coordinate system of the landmark is updated to the current radial landmark.

[0049] Replace the initial radial landmark with the current radial landmark and return to the step "Establish a local coordinate system for the landmark with the initial radial landmark as the origin, the X-axis as the normal direction, the Y-axis as the tangent direction, and the Z-axis as the same as the radial shaft axis", until convergence is achieved, and the locally optimal radial landmark is obtained.

[0050] Optionally, establishing a local coordinate system for the forearm bone end, with the center of the forearm bone end feature plane as the origin, specifically includes:

[0051] When the forearm bone is the ulna, the characteristic plane of the forearm bone end is determined to be the characteristic plane of the proximal ulna.

[0052] The origin is the point where the proximal axis of the ulna intersects with the characteristic plane of the proximal ulna.

[0053] With the proximal axis of the ulna as the Z-axis, the direction pointing towards the proximal end of the ulna is the positive direction of the Z-axis;

[0054] The X-axis is defined by the line connecting the intersection of the proximal radius axis and the characteristic plane of the proximal ulna and the origin, with the positive direction of the X-axis pointing towards the ulna.

[0055] The Y-axis and its positive direction are determined using the right-hand rule, thereby establishing a local coordinate system for the proximal ulna.

[0056] Optionally, the establishment of a local coordinate system for the forearm bone end, with the center of the forearm bone end feature plane as the origin, further includes:

[0057] The proximal ulna feature plane is fitted to an ellipse, and a parallel section system is established perpendicular to the minor axis of the ellipse;

[0058] Identify the points in the parallel section system that are located at the proximal end of the ulna along the axis of the proximal end of the ulna, forming a set of points;

[0059] The proximal ulnar crest is established from the set of points, and the plane containing the proximal ulnar crest is fitted to obtain the coronal plane of the proximal ulnar bone.

[0060] Calculate the angle between the proximal axis of the ulna and the coronal plane of the proximal ulna, and rotate the coronal plane of the proximal ulna according to the angle until it is basically coincident with the proximal axis of the ulna to obtain the corrected coronal plane of the proximal ulna.

[0061] Optionally, the step of determining initial forearm bony landmarks based on the local coordinate system of the forearm bone ends, using the characteristic planes of the forearm bone ends and the three-dimensional model of the forearm bones, specifically includes:

[0062] Determine the intersection contour of the 3D model of the ulna with the XOZ plane in the local coordinate system of the proximal ulna;

[0063] Based on the contour points on the intersecting contours, the first score of each contour point is calculated using the scoring function s1 = x + z, and the second score of each contour point is calculated using the scoring function s2 = -x + z.

[0064] The contour point with the highest first score is taken as the initial ulnar end point of the olecranon, and the contour point with the highest second score is taken as the initial radial end point of the olecranon.

[0065] Calculate the projection points of the contour points on the proximal ulna feature plane onto the local coordinate system of the proximal ulna, and determine the projection point that is farthest from the origin along the Y-axis of the local coordinate system of the proximal ulna. Take the contour point corresponding to the projection point that is farthest from the origin as the tip of the coronoid process of the ulna.

[0066] Establish the midline planes of the initial ulnar end point and the initial radial end point of the ulnar olecranon, and obtain the intersection contour of the ulnar three-dimensional model and the midline planes;

[0067] The arc segment in the region enclosed by the proximal ulnar feature plane and the posterior side of the corrected proximal ulnar coronal plane in the intersecting contour is defined proximally.

[0068] Calculate the curvature of each contour point in the arc segment, and take the contour point with the largest curvature as the initial olecranon tip of the ulna.

[0069] Optionally, the step of performing local optimization on the initial forearm bony landmarks within a preset neighborhood to determine the finally identified forearm bony landmarks specifically includes:

[0070] Move the origin of the local coordinate system at the proximal end of the ulna a predetermined distance in the negative X-axis direction, and denote it as the base point.

[0071] Establish from the base point A vector pointing to the current ulnar landmark is used, and the projection of each point within a preset neighborhood onto the vector is calculated to obtain the neighborhood point corresponding to the maximum projection value; the current ulnar landmark includes the ulnar end point of the initial olecranon and the radial end point of the initial olecranon.

[0072] Replace the current ulnar landmark with the neighborhood point corresponding to the maximum projection value, and return to step "Establish from base point". The vector pointing to the current ulnar landmark is calculated, and the projection of each point within a preset neighborhood onto the vector is obtained. The neighborhood point corresponding to the maximum projection value is obtained until convergence, thus obtaining the local optimum of the current ulnar landmark.

[0073] Local optimization is performed on the initial olecranon tip of the ulna to obtain the locally optimal olecranon tip of the ulna.

[0074] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0075] This invention discloses an intelligent method for recognizing forearm bony landmarks. First, the axis of the forearm bone ends is determined from a three-dimensional model of the forearm bones. A system of parallel surfaces perpendicular to this axis is established. The surface containing the largest intersection area between each surface of the parallel surface system and the three-dimensional model of the forearm bones is selected as the feature plane of the forearm bone ends. Then, a local coordinate system of the forearm bone ends is established. Initial forearm bony landmarks are determined using the feature plane, the three-dimensional model, and the local coordinate system. Finally, local optimization is performed on the initial forearm bony landmarks within a preset neighborhood to determine the final recognized forearm bony landmarks. This invention integrates statistical and geometric methods, exhibiting good stability and individual specificity. After identifying the approximate location of the landmarks based on anatomical knowledge, local optimization is used to refine the recognition results, improving accuracy. This method is suitable for the automatic recognition of bony landmarks in forearm images with certain bony deformities. Attached Figure Description

[0076] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0077] Figure 1 A schematic diagram of existing bony landmarks in the forearm; Figure 1 (a) in the diagram shows Lister's tubercle of the radius and the dorsal end of the radial styloid process. Figure 1 (b) in the diagram shows the palmar end point of the radial styloid process and the palmar end point of the radial sigmoid notch. Figure 1 (c) in the diagram shows the tip of the coronoid process of the ulna, the ulnar end of the olecranon of the ulna, and the radial end of the olecranon of the ulna. Figure 1 (d) in the diagram is a schematic diagram of the tip of the olecranon of the ulna;

[0078] Figure 2 A flowchart of the intelligent recognition method for forearm bone landmarks provided in an embodiment of the present invention;

[0079] Figure 3 This is a schematic diagram illustrating the principle of distal radius bone landmark recognition provided in Embodiment 1 of the present invention.

[0080] Figure 4 This is a schematic diagram of the local coordinate system of the distal radius provided in Embodiment 1 of the present invention;

[0081] Figure 5 This is a schematic diagram of a three-dimensional model of the left radius of a 2-year-old girl provided in Embodiment 1 of the present invention;

[0082] Figure 6This is a schematic diagram of the distal radius landmark of a 2-year-old girl provided in Embodiment 1 of the present invention;

[0083] Figure 7 This is a schematic diagram of the fitting of the distal radius axis of a 2-year-old girl according to Embodiment 1 of the present invention;

[0084] Figure 8 This is a diagram of the distal radius platform provided in Embodiment 1 of the present invention; Figure 8 (a) in the diagram is a schematic diagram of the distal radial plateau. Figure 8 (b) in the figure is a schematic diagram of the center of the distal radius plateau.

[0085] Figure 9 This is a schematic diagram of the distal radial plateau contour template of a 2-year-old girl provided in Embodiment 1 of the present invention;

[0086] Figure 10 This is a schematic diagram of the distal radius landmark of a 2-year-old girl based on a template, provided in Embodiment 1 of the present invention. Figure 10 (a) in the diagram is a schematic diagram of the dorsal landmark of the radius. Figure 10 (b) in the diagram is a schematic diagram of the palmar landmarks of the radius;

[0087] Figure 11 This is a schematic diagram of the local optimization base point setting of the radial landmark of a 2-year-old girl provided in Embodiment 1 of the present invention;

[0088] Figure 12 This is a schematic diagram of the local optimization of the radial landmark of a 2-year-old girl provided in Embodiment 1 of the present invention; Figure 12 (a) in the diagram is a schematic diagram of the local coordinate system of the marker point. Figure 12 (b) in the diagram is a neighborhood diagram;

[0089] Figure 13 This is a schematic diagram illustrating the principle of proximal ulna bone landmark recognition provided in Embodiment 2 of the present invention.

[0090] Figure 14 This is a schematic diagram of fitting the proximal ulna axis of a 2-year-old girl according to Embodiment 2 of the present invention;

[0091] Figure 15 This is a schematic diagram of the ulna of the forearm of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0092] Figure 16 This is a schematic diagram of the proximal coronal plane of the ulna of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0093] Figure 17 This is a schematic diagram of the proximal coronoid process of the ulna of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0094] Figure 18This is a schematic diagram of the local coordinate system of the proximal ulna of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0095] Figure 19 This is a schematic diagram of the ulnar end of the olecranon of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0096] Figure 20 This is a schematic diagram of the local optimization of the radial end of the olecranon of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0097] Figure 21 This is a schematic diagram of the olecranon tip of the ulna of a 2-year-old girl provided in Embodiment 2 of the present invention;

[0098] Figure 22 This is a schematic diagram of the proximal ulna landmark of a 2-year-old girl provided in Embodiment 2 of the present invention; Figure 22 (a) in the diagram is a schematic diagram of the tip of the olecranon of the ulna; Figure 22 (b) is a schematic diagram of the tip of the coronoid process of the ulna, the ulnar end of the olecranon of the ulna, and the radial end of the olecranon of the ulna. Detailed Implementation

[0099] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0100] The purpose of this invention is to provide an intelligent identification method for forearm bone landmarks, applicable to the automatic identification of bone landmarks in forearm images with certain bony deformities.

[0101] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0102] This invention utilizes template matching algorithms, local optimization algorithms, and other technologies to automate the identification of bony landmarks in the forearm, and can be used for establishing local coordinate systems of the radius and ulna and forearm bone registration.

[0103] Forearm bones: The bones of the human forearm include the ulna and radius. There are four key landmarks at the proximal end of the ulna and the distal end of the radius. The proximal end of the ulna is larger (proximal refers to the side closer to the upper arm), and the distal end of the radius is larger (distal refers to the side closer to the palm).

[0104] Forearm bony landmarks: The proximal ulna has four landmarks: the tip of the coronoid process, the tip of the olecranon, the ulnar end of the olecranon, and the radial end of the olecranon; the distal radius has four landmarks: Lister's tubercle, the dorsal end of the radial styloid process, the palmar end of the radial styloid process, and the palmar end of the radial sigmoid notch. Figure 1 As shown.

[0105] This invention provides an intelligent recognition method for forearm bone landmarks, such as... Figure 2 As shown, it includes:

[0106] Step S1: Extract the forearm bones from the forearm CT image and construct a three-dimensional model of the forearm bones; the forearm bones include the radius and ulna.

[0107] Step S2: Determine the axis of the forearm bone ends in the three-dimensional model of the forearm bones; the forearm bone ends include the distal radius and the proximal ulna.

[0108] Step S3: Establish a system of parallel surfaces perpendicular to the axis of the forearm bone end, and determine the surface with the largest contour area where each surface in the parallel surface system intersects with the three-dimensional model of the forearm bone as the feature plane of the forearm bone end.

[0109] Step S4: Establish a local coordinate system for the forearm bone end with the center of the contour of the feature plane of the forearm bone end as the origin.

[0110] Step S5: Based on the local coordinate system of the forearm bone ends, the initial forearm bony landmarks are determined using the feature plane of the forearm bone ends and the three-dimensional model of the forearm bones.

[0111] Step S6: Perform local optimization on the initial forearm bony landmarks within a preset neighborhood to determine the final identified forearm bony landmarks.

[0112] Since the forearm bones consist of the ulna and radius, there are four key landmarks at the proximal end of the ulna and the distal end of the radius. Therefore, the intelligent recognition method for forearm bone landmarks in this invention can be applied to the recognition of landmarks at both the proximal end of the ulna and the distal end of the radius.

[0113] Example 1

[0114] The intelligent recognition method for forearm bone landmarks provided by this invention is applied to the recognition of distal radius landmarks. The principle of distal radius bone landmark recognition is as follows: Figure 3 As shown. The process for identifying distal radius bone landmarks includes:

[0115] (1) Construct a distal radius platform template;

[0116] (2) Extract the radius of the forearm from the CT image of the forearm, construct a three-dimensional model of the radius, and determine the distal axis of the radius of the forearm in the three-dimensional model of the radius;

[0117] (3) Establish a parallel plane system perpendicular to the axis of the distal radius of the forearm, and determine the plane with the largest contour area where each plane in the parallel plane system intersects with the three-dimensional model of the radius as the distal radius platform;

[0118] (4) Establish a local coordinate system for the distal radius with the center of the contour of the distal radius platform as the origin;

[0119] (5) Based on the local coordinate system of the distal radius, the iterative nearest point method is used to register the contour of the distal radius platform and the contour of the distal radius platform template. The point in the registered distal radius platform contour that is closest to the landmark point of the distal radius platform template is recorded as the initial radial landmark point.

[0120] (6) Locally optimize the initial radial landmark within a preset neighborhood to determine the final distal radial bony landmark of the forearm.

[0121] The detailed process of identifying distal radius bone landmarks is as follows:

[0122] (I) Identification of bone landmarks at the distal radius

[0123] Step 1: Segment and extract the radius and ulna from the forearm CT image, and construct a 3D model R for the radius and a 3D surface model U for the ulna. Calculate the radius length L. R And take the distal 1 / 3 segment of the radius.

[0124] Step 2: In the distal 1 / 3 segment of the radius Construct a parallel surface system {π i The contour formed by the three-dimensional model of the radius and each plane is calculated. This contour is approximately elliptical, and its centroid c is calculated. i Using the least squares method, the contour centers of each surface in the parallel plane system are fitted to spatial straight lines to obtain the distal radius axis l. R =LS(c i ).

[0125] The method for establishing the parallel plane system is as follows: First, establish the parallel plane system along the longitudinal axis of the CT scan. Each plane is the same as the CT layer plane. Then, calculate the contour lines of the intersection of each plane with the radius model, extract and record the midpoints of the contour lines of each layer, fit the distal radius axis with these midpoints, use the direction of this axis as the normal vector of the new plane, and use the CT layer spacing as the spacing of the new plane system. Establish the parallel plane system in the distal 1 / 3 segment of the radius.

[0126] Step 3: Establish a perpendicular axis to the distal radius. R The system of parallel surfaces is used to calculate the area of ​​the contours where the 3D model intersects with each surface, and the surface π containing the largest contour area is obtained from this system. maxThis is denoted as the distal radial plateau.

[0127] Step 4: Establish the local coordinate system of the distal radius. Calculate the distal radius plateau. The center point of the contour serves as the origin O of the local coordinate system of the distal radius. R Using a method similar to step 2, the distal ulna axis l was obtained. U And calculate the intersection point of it with the plane containing the distal radius plateau, denoted as P. U ,by The direction is X R Axial direction, with the distal radius axis l R For Z R axial direction, Y R The axial direction is obtained using the right-hand rule, as shown in Table 1.

[0128] Table 1 Local coordinate system of distal radius

[0129]

[0130] Step 5: Construction of the distal radius platform template.

[0131] N forearm bone samples were selected, and four radial landmarks were marked by a professional physician as the gold standard (Global Truth, GT), denoted as... For i = 1, 2, 3, 4, the above steps are used to calculate the projections of the corresponding distal radius plateau contour points and GT landmarks onto the distal radius plateau. i = 1, 2, 3, 4.

[0132] Calculate the area of ​​the distal radial plateau contour and perform a similarity transformation to unify its area. For example, the similarity transformation can be performed by calculating the area of ​​the distal radial plateau based on its contour points and then unifying this area to a value of, for example, 200 mm. 2 That is, if the actual area is S, then the scaling factor is... Then, the palmar endpoint of the radial sigmoid notch is rotated to the same angle θ. The template is then standardized by contour segment processing; specifically, by… The distal radial plateau contour is divided into four parts: the palmar end of the sigmoid notch of the radius. ~Lister's tubercle of the radius Lister's tubercle of the radius ~dorsal end point of the radial styloid process dorsal side of the radial styloid process ~Palm end point of the radial styloid process and the palmar end point of the radial styloid process ~Palm end point of the radial sigmoid notch Each part is then interpolated to generate a fixed number of contour points. For example, all samples are processed using the palmar endpoints of the radial sigmoid notch. The radius Lister tubercle segment is interpolated into N1 small line segments. Based on the shape of the standardized segmentation, the meshes between samples are paired according to their node numbers. The relationship between each standardized contour point and O is calculated. R The distances and angles between the lines are calculated, and their mean and standard deviation are determined to obtain the average profile.

[0133] Step 6: Determine the far-end platform contour of the new sample {(x i ,y i Normalize according to its area And using the established coordinate system of the distal radius, the palmar and ulnar regions are obtained. The maximum distance is located at the palmar end of the radial sigmoid notch. Rotate it to a fixed angle θ. Then, outline the distal radial plateau of the new sample. and template outline The ICP algorithm was used for registration. After registration, the point closest to the marker point in the template was obtained from the distal radius plateau contour points of the new sample, and denoted as the radial marker point of the new sample. i = 1, 2, 3, 4.

[0134] Step 7: Local optimization of the radial landmark.

[0135] While an approximate radial landmark has been obtained based on the template, individual differences exist in human skeletons, and the bone landmark is located near the distal radial plateau rather than strictly on the distal radial plateau. Local optimization is performed near this point to obtain a more accurate radial landmark.

[0136] For template-based radial landmark recognition Establish a local coordinate system for marker points i = 1, 2, 3, 4, with the local coordinate directions being tangential, normal, and longitudinal. Based on this, define the extent of the neighborhood, such as... Figure 4 As shown. The neighborhood is divided into tangential and longitudinal ranges, forming an elliptical region. The specific size of the range is an adjustable parameter, generally set to 3-5 mm, while also considering factors such as patient age and bone size. The vectors from each point on the radius model to the current point are calculated, and then projected in the tangential and longitudinal directions to calculate the tangential and longitudinal distances from each point to the current point. Points meeting the requirements are considered within the neighborhood.

[0137] Set the origin O of the local coordinate system at the distal radius to... R Move a certain distance towards the proximal end, and denote it as the base point. Calculate the direction vector And in Within the neighborhood of a point, calculate the value of each point P. f Measurements far from the base point Will Move the point to Repeat this operation until convergence, obtaining locally optimized distal radius bony landmarks.

[0138] like Figure 5 As shown, using a three-dimensional model of the left radius of a 2-year-old girl as an example, the identification of bone landmarks at the distal end of the radius is illustrated.

[0139] There are four landmarks distal to it: Lister's tubercle of the radius, the dorsal endpoint of the radial styloid process, the ulnar endpoint of the radial styloid process, and the palmar endpoint of the radial sigmoid notch, such as... Figure 6 As shown.

[0140] First, a coordinate system for the distal radius is constructed. The longitudinal axis of the distal radius is fitted, as follows: Figure 7 As shown.

[0141] Establish a vertical plane system along the distal radius axis and calculate the area of ​​each section. Take the section with the largest area, such as... Figure 8 As shown by the dashed line in (a) of the diagram, this section represents the distal radius plateau.

[0142] The distal ulnar axis is fitted using a method that fits to the distal radius axis, and the intersection point with the plane containing the distal radial plateau is calculated. Figure 8 As shown at the left point in (b) of the diagram. Calculate the center of the distal radial plateau contour, as shown in the diagram. Figure 8 As shown at the right-hand point in (b) of the diagram. Further establish a local coordinate system for the distal radius.

[0143] Ten samples were taken, each with manually marked landmarks. The above operation was repeated to establish a distal radius platform, and the projection of the landmarks onto this platform was calculated. The contour area was calculated, and the area was corrected to 200 mm² using a scaling transformation. 2 The palmar endpoint of the radial sigmoid notch was corrected to -135° and adjusted as the starting position of the contour. The contour points of the samples were then standardized using interpolation. The number of nodes at the palmar endpoint of the radial sigmoid notch, the Lister's tubercle of the radius, the dorsal endpoint of the radial styloid process, and the palmar endpoint of the radial styloid process were uniformly set to 70, 40, 30, and 60 respectively. This ensures that on the radial distal plateau contour, points 70, 110, 140, and 0 are each four landmark points. Figure 9 As shown. Figure 9 The solid lines on the middle and outer sides represent the standardized contour of the distal radius plateau, while the solid lines on the inner side represent the original contour of the distal radius plateau.

[0144] The iterative nearest-point method is used to register the distal radial plateau template with the contour of the distal radial plateau of the new sample. After registration, landmark points matching the template are found in the contour of the new sample, such as... Figure 10 As shown.

[0145] like Figure 10 As shown, the distal radius landmarks are roughly identified, but the accuracy is poor. The reasons are: (1) The identified distal radius platform is the largest area. The distal platform may have accumulated errors due to the identification error of the distal radius shaft axis; (2) The distal radius landmarks are not strictly on the distal radius platform.

[0146] Local optimization of these landmarks is performed to establish base points, and the centroid of the distal radius plateau contour is determined. Figure 11 (The point located above the middle) moves a distance d proximally along the radial shaft axis, as shown. Figure 11 As shown.

[0147] Taking the dorsal endpoint of the radial styloid process as an example, establish its local coordinate system, such as... Figure 12 As shown in (a), the origin is the location of the marker, the X-axis is the normal direction, the Y-axis is the tangent direction, and the Z-axis is the same as the radial shaft axis. Based on this local coordinate system, the adjacent regions of the current marker are established, such as... Figure 12 As shown in (b) above. Connect the base point and the current marker point, and calculate the projection of the neighboring points along this vector direction. Find the outermost point of the neighboring points relative to this direction, update it as the current marker point, and repeat this step until convergence. This yields the locally optimal marker point, as shown below. Figure 12 As shown.

[0148] In the automatic identification of four landmarks at the distal radius, this invention proposes a method to obtain the approximate location of the landmarks by using statistical analysis of other samples to create templates and adopting a local optimization approach to incorporate individual specificity into the automatic correction process of the landmarks.

[0149] In the process of identifying bony landmarks at the distal radius, if there is radial angular deformity or moderate torsion deformity, the identification of the distal radial plateau is not affected due to the correction of the distal radial axis, and thus the effectiveness of identifying bony landmarks at the distal radius is not affected.

[0150] Example 2

[0151] The intelligent recognition method for forearm bone landmarks provided by this invention is applied to the recognition of proximal ulnar landmarks. The principle of proximal ulnar bone landmark recognition is as follows: Figure 13 As shown.

[0152] The process of identifying proximal ulnar landmarks is as follows:

[0153] Step 1: Forearm ulna segmentation and reconstruction

[0154] The radius and ulna of the forearm were segmented and obtained from forearm CT images, and three-dimensional models R of the radius and U of the ulna were constructed respectively. The length of the ulna was calculated, and the proximal 1 / 5 to 1 / 2 segment of the ulna was selected.

[0155] Step 2: Obtain the proximal axis of the ulna and the proximal axis of the radius.

[0156] In the ulna Construct a system of parallel surfaces and calculate the contours formed by the 3D model U of the ulna and each surface. Calculate the centers of these contours and use the least squares method to determine the centers c of the contours of each surface in the parallel surface system. i Fitting the line to a spatial straight line yields the proximal ulnar axis l. U =LS(c i The proximal radial segment was selected, and the proximal radial axis l was estimated using a similar method. R .

[0157] Step 3: Obtain the proximal ulna feature plane

[0158] Establish a perpendicular axis to the proximal ulna. U The system of parallel surfaces is used to calculate the area of ​​the contours where the 3D ulna model U intersects with each surface, and the surface with the largest contour area is obtained from this system. Fit the surface to an ellipse. Obtain the direction of the minor axis of the ellipse.

[0159] Step 4: Obtain the coronal plane of the proximal ulna

[0160] Establish a system of parallel sections perpendicular to the minor axis of the ellipse, and calculate the nearest point along the proximal ulnar axis in each section (the nearest point is determined along the proximal ulnar axis), forming a point set. Eliminate interference points far from the proximal ulnar end in this set, and use the remaining points to construct the proximal ulnar crest, fitting the plane containing the proximal ulnar crest. Calculate the angle between the proximal ulnar axis and the proximal ulnar coronal plane, and rotate the proximal ulnar coronal plane until it approximately coincides with the proximal ulnar axis, completing the correction of the proximal ulnar coronal plane. The proximal ulnar coronal plane is denoted as...

[0161] The method for identifying interference points is as follows: Based on the proximal axis of the ulna, find the nearest point on the ulnar model, and set a threshold parameter to indicate the degree to which the point is far from the nearest endpoint of the ulnar model. Then, calculate the longitudinal distance between each point in this set and the nearest endpoint; points exceeding the threshold parameter are identified as interference points.

[0162] Step 5: Obtain the local coordinate system of the proximal ulna.

[0163] Based on the proximal axis of the ulna plane with proximal ulna The intersection point is the origin, and the proximal axis of the ulna is also considered. The Z-axis (positive pointing proximally) is the axis of the proximal radius. plane with proximal ulna The line connecting the intersection point and the origin is the X-axis, with the ulnar side being positive. The Y-axis is determined according to the right-hand rule. This coordinate system is based on the right forearm bone, with the Y-axis pointing towards the anterior end as positive, as shown in Table 2.

[0164] Table 2 Local coordinate system of the proximal ulna

[0165]

[0166] Step 6: Identification and local optimization of the ulnar and radial endpoints of the olecranon.

[0167] Based on the local coordinate system of the proximal ulna, the contour of the intersection of the ulna model and the XOZ plane is obtained. Scoring functions are calculated at this contour point: s1 = x + z and s2 = -x + z. The point with the highest score is the ulnar end of the olecranon. and the radial end of the olecranon of the ulna The origin of the local coordinate system at the proximal end of the ulna is moved a certain distance in the negative X-axis direction, and this distance is denoted as the base point. by Let P be the normal vector for a point on the contour. f Calculate the corresponding scores s1 and s2. Take the score with the highest value as the locally optimized ulnar end point of the olecranon. and the radial end of the olecranon of the ulna The highest point of s1 corresponds to the ulnar end of the olecranon process of the ulna. The highest point of s2 corresponds to the radial end of the olecranon of the ulna.

[0168] Step 7: Identification of the tip of the coronoid process of the ulna

[0169] Calculate the contour points on the proximal ulna feature plane, and obtain the point that projects anteriorly onto the local coordinate system of the proximal ulna (in the local coordinate system of the proximal ulna, the Y-axis points anteriorly, so anteriorly can be understood as being anteriorly relative to the direction of the Y-axis vector), and obtain the tip of the coronoid process of the ulna.

[0170] Step 8: Identification and local optimization of the olecranon tip of the ulna

[0171] Establish the ulnar end of the olecranon. and the radial end of the olecranon of the ulna The mid-perpendicular plane is calculated, and the intersection contour of the ulna model and the mid-perpendicular plane is determined, locking the contour located on the ulna feature plane. Using the proximal contour and the arc segment of the posterior region of the proximal coronal plane of the ulna, calculate the curvature at each point, and take the point with the largest local curvature as the tip of the olecranon. Using the origin of the local coordinate system at the proximal end of the ulna as the base point, local optimization is performed on the tip of the olecranon of the ulna.

[0172] Example of ulnar landmark identification:

[0173] Figure 14 The image shows the ulna of the forearm of a 2-year-old girl. One-fifth to one-half of the proximal bone segment was taken, and the proximal axis of the ulna was fitted.

[0174] Establish a system of sections perpendicular to the axis of the proximal ulna, and find the section with the largest area, denoted as the proximal ulna feature plane. Perform ellipse fitting on the ulnar contour corresponding to this section, and determine its major and minor axes. Figure 15 In the diagram, the proximal ulnar feature plane intersects the outermost part of the ulna to form the outline of the corresponding proximal ulnar feature plane, and the dashed line represents the minor axis SA of the fitted ellipse.

[0175] Establish a cross-section system perpendicular to the minor axis, and calculate the point closest to the proximal end of the ulna in each cross-section. Exclude points farther from the proximal ulna, and record these as the proximal ulnar crest. Perform least-squares plane fitting to form the coronal plane of the proximal ulna. Correct the proximal ulnar coronal plane so that the proximal ulnar axis is parallel to this plane, as shown below. Figure 16 As shown.

[0176] In the proximal ulna feature plane contour points, find the point located anterior to the ulna and furthest from the proximal coronal plane of the ulna. This point is the coronal process of the ulna. Figure 17 As shown.

[0177] The coordinate system is defined with the projection of the coronoid process of the ulna onto the coronal plane of the proximal ulna as the origin, the axis of the proximal ulna as the Z-axis, the normal to the proximal ulnar crest as the X-axis, and the Y-axis determined by the right-hand rule. Figure 18 As shown.

[0178] Obtain the contour of the ulna model on the YOZ plane, as shown in the figure. Establish the scoring function f1 = -2z + y.

[0179] This allows for the identification of the ulnar end of the olecranon on this contour, such as... Figure 19 As shown, the identification method for the radial end of the olecranon of the ulna is similar, and the corresponding cost function is f2 = 2z + y.

[0180] After initially identifying the radial and ulnar endpoints of the olecranon, local optimization was performed, setting the neighborhood size to 4. Taking the radial endpoint of the olecranon as an example, points within its neighborhood are as follows: Figure 20 As shown.

[0181] Let d1 be the distance between the ulnar and radial ends of the olecranon of the ulna, and let p be the base point. b =p u -n x ·d1 / 3+n z ·d1 / 2, establish a vector from the base point to the current marker point, calculate the projection of each point in the neighborhood onto this vector, take the maximum value, and repeat the operation until convergence, to achieve local optimization of the radial end of the olecranon of the ulna. Local optimization of the ulnar end of the olecranon of the ulna is similar.

[0182] Construct the midline planes of the ulnar and radial endpoints of the olecranon process of the ulna, and calculate the contours where the ulnar surface intersects these planes, obtaining the contours as shown below. Figure 21 As shown. The optimal olecranon tip of the ulna is found using the evaluation function f3 = zx.

[0183] The finally identified proximal ulna landmarks are as follows: Figure 22 As shown, where, Figure 22 Point (a) in the diagram represents the olecranon tip of the ulna. Figure 22 In (b), the bottommost point is the coronoid process of the ulna, the point on the upper right is the ulnar end of the olecranon, and the point on the upper left is the radial end of the olecranon.

[0184] Example 3

[0185] When simultaneously identifying four key landmarks on the proximal ulna and distal radius, both Example 1 and Example 2 can be used to achieve accurate identification of all bony landmarks in the forearm. The process of identifying distal radius landmarks in Example 1 and identifying proximal ulna landmarks in Example 2 are detailed in Example 1 and Example 2 respectively, and will not be repeated here.

[0186] The method proposed in this invention integrates statistical and geometric methods, exhibiting good stability and individual specificity. Furthermore, this method is suitable for the automatic identification of bony landmarks in the forearm bones with certain deformities.

[0187] This invention can automatically identify bony landmarks of the forearm bones of the upper limb. In this process, the approximate location of the landmarks is first identified based on anatomical knowledge, and then the landmark identification results are optimized by local optimization to improve the accuracy of identification.

[0188] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0189] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent recognition of forearm bone landmarks, characterized in that, include: Forearm bones were extracted from forearm CT images to construct a three-dimensional model of the forearm bones; the forearm bones include the radius and ulna. The axis of the forearm bone ends is determined in the three-dimensional model of the forearm bones; the forearm bone ends include the distal radius and the proximal ulna. Establish a system of parallel surfaces perpendicular to the axis of the forearm bone end, and determine the surface containing the largest contour area where each surface in the parallel surface system intersects with the three-dimensional model of the forearm bone as the feature plane of the forearm bone end. A local coordinate system for the forearm bone end is established with the center of the contour of the feature plane of the forearm bone end as the origin; Based on the local coordinate system of the forearm bone ends, the initial bony landmarks of the forearm are determined using the feature plane of the forearm bone ends and the three-dimensional model of the forearm bone. Local optimization is performed on the initial forearm bony landmarks within a preset neighborhood to determine the final identified forearm bony landmarks; The method of determining initial forearm bony landmarks based on the local coordinate system of the forearm bone ends, using the characteristic planes of the forearm bone ends and the three-dimensional model of the forearm bone, specifically includes: A distal radius platform template is constructed; the distal radius platform template is marked with radial landmarks; the radial landmarks include Lister's tubercle of the radius, the dorsal endpoint of the radial styloid process, the palmar endpoint of the radial styloid process, and the palmar endpoint of the radial sigmoid notch. Normalize the characteristic planar contour of the radius to obtain the distal radial plateau; Based on the contour of the distal radius plateau, using the local coordinate system of the distal radius, the point with the largest distance from the origin in the contour of the distal radius plateau in the palmar and ulnar regions is obtained as the initial palmar endpoint of the radial sigmoid notch. The initial palmar end of the radial sigmoid notch is rotated to a fixed angle to obtain the rotated distal radial plateau contour; The iterative nearest point method was used to register the rotated distal radius plateau contour with the distal radius plateau template contour; Select the points in the registered distal radius plateau contour that are closest to the Lister tubercle of the radius and the Lister tubercle of the distal radius plateau template, the points in the registered distal radius plateau contour that are closest to the dorsal endpoint of the radial styloid process and the dorsal endpoint of the radial styloid process of the distal radius plateau template, the points in the registered distal radius plateau contour that are closest to the palmar endpoint of the radial styloid process and the palmar endpoint of the radial styloid process of the distal radius plateau template, and the points in the registered distal radius plateau contour that are closest to the palmar endpoint of the radial sigmoid notch and the palmar endpoint of the radial sigmoid notch of the distal radius plateau template, and record them as the initial radial landmark points.

2. The intelligent recognition method for forearm bone landmarks according to claim 1, characterized in that, Determining the axis of the forearm bone ends in the three-dimensional model of the forearm bone specifically includes: When the forearm bone is the radius, the three-dimensional model of the forearm bone is determined to be the three-dimensional model of the radius; A parallel plane system was constructed along the longitudinal axis of the forearm CT image in the distal third of the radius; Determine the contour formed by the three-dimensional model of the radius and each plane in the parallel plane system, as well as the center of the contour; The least squares method is used to fit the contour centers of each plane in the parallel plane system to a spatial straight line, which serves as the distal radius axis.

3. The intelligent recognition method for forearm bone landmarks according to claim 2, characterized in that, The establishment of a local coordinate system for the forearm bone end, with the center of the contour of the forearm bone end feature plane as the origin, specifically includes: The center point of the radial bone's characteristic planar contour is taken as the origin of the local coordinate system of the distal radius. ; The intersection of the distal axis of the ulna and the characteristic plane of the radius is denoted as... ; by The direction is the X-axis, the Z-axis is the axis of the distal radius, and the positive direction of the Z-axis is the direction from the proximal end to the distal end. The Y-axis is determined by the right-hand rule to establish a local coordinate system for the distal radius.

4. The intelligent recognition method for forearm bone landmarks according to claim 3, characterized in that, The construction of the distal radius platform template specifically includes: Multiple forearm bone samples were selected, and radial landmarks were marked on each forearm bone sample. Construct a three-dimensional model of the radius for each forearm bone sample, and determine the distal radius axis in each three-dimensional model; Establish a system of parallel planes perpendicular to the axis of the distal radius, and determine the plane containing the largest contour area where each plane in the parallel plane system intersects with the three-dimensional model of the radius as the feature plane of the distal radius. Determine the projection points of the radial landmarks onto the distal radial feature plane and the contour points of the distal radial feature plane; Calculate the area of ​​each distal radius feature plane contour and unify all areas; Rotate the palmar end of the radial sigmoid notch on each radial distal feature plane with a uniform area to the same angle to obtain multiple rotated radial distal feature planes. Interpolate the contour segments on the rotated distal radius feature plane contour between the projection point of the palmar end of the radial sigmoid notch and the projection point of the Lister tubercle of the radius, between the projection point of the Lister tubercle of the radius and the projection point of the dorsal end of the radial styloid process, between the projection point of the dorsal end of the radial styloid process and the projection point of the palmar end of the radial styloid process, and between the projection point of the palmar end of the radial styloid process and the projection point of the palmar end of the radial sigmoid notch, so that a fixed number of contour points are interpolated on each contour segment. Calculate each contour point and The distance and angle of the line connecting them; Based on the contour points of all forearm bone samples and The distance and angle of the connecting lines are used to determine the average contour of the distal radius feature plane, which serves as a template for the distal radius platform.

5. The intelligent recognition method for forearm bone landmarks according to claim 4, characterized in that, The step of performing local optimization on the initial forearm bony landmarks within a preset neighborhood to determine the final identified forearm bony landmarks specifically includes: With the initial radial landmark location as the origin, the X-axis as the normal direction, the Y-axis as the tangent direction, and the Z-axis as the same as the radial shaft axis, a local coordinate system for the landmark is established. Determine the pre-defined neighborhood range along the tangential and longitudinal directions on the local coordinate system of the marker point; The origin Move a predetermined distance towards the proximal radius, and denote the base point as the base point. And select a relative point from the preset neighborhood range. The outermost neighboring point in the direction connecting the origin of the local coordinate system of the landmark is updated to the current radial landmark. Replace the initial radial landmark with the current radial landmark and return to the step "Establish a local coordinate system for the landmark with the initial radial landmark as the origin, the X-axis as the normal direction, the Y-axis as the tangent direction, and the Z-axis as the same as the radial shaft axis" until convergence, and obtain the locally optimal radial landmark.

6. The intelligent recognition method for forearm bone landmarks according to claim 1 or 5, characterized in that, The establishment of a local coordinate system for the forearm bone end, with the center of the contour of the forearm bone end feature plane as the origin, specifically includes: When the forearm bone is the ulna, the characteristic plane of the forearm bone end is determined to be the characteristic plane of the proximal ulna. The origin is the point where the proximal axis of the ulna intersects with the characteristic plane of the proximal ulna. With the proximal axis of the ulna as the Z-axis, the direction pointing towards the proximal end of the ulna is the positive direction of the Z-axis; The X-axis is defined by the line connecting the intersection of the proximal radius axis and the characteristic plane of the proximal ulna and the origin, with the positive direction of the X-axis pointing towards the ulna. The Y-axis and its positive direction are determined using the right-hand rule, thereby establishing a local coordinate system for the proximal ulna.

7. The intelligent recognition method for forearm bone landmarks according to claim 6, characterized in that, The establishment of a local coordinate system for the forearm bone end, with the center of the contour of the forearm bone end feature plane as the origin, also includes: The proximal ulna feature plane is fitted to an ellipse, and a parallel section system is established perpendicular to the minor axis of the ellipse; Identify the points in the parallel section system that are located at the proximal end of the ulna along the axis of the proximal end of the ulna, forming a set of points; The proximal ulnar crest is established from the set of points, and the plane containing the proximal ulnar crest is fitted to obtain the coronal plane of the proximal ulnar bone. Calculate the angle between the proximal axis of the ulna and the coronal plane of the proximal ulna, and rotate the coronal plane of the proximal ulna according to the angle until it coincides with the proximal axis of the ulna to obtain the corrected coronal plane of the proximal ulna.

8. The intelligent recognition method for forearm bone landmarks according to claim 7, characterized in that, The method of determining initial forearm bony landmarks based on the local coordinate system of the forearm bone ends, using the characteristic planes of the forearm bone ends and the three-dimensional model of the forearm bone, specifically includes: Determine the intersection contour of the 3D model of the ulna with the XOZ plane in the local coordinate system of the proximal ulna; Based on the contour points on the intersecting contours, a scoring function is used. Calculate the first score for each contour point and use the scoring function. Calculate the second score for each contour point; The contour point with the highest first score is taken as the initial ulnar end point of the olecranon, and the contour point with the highest second score is taken as the initial radial end point of the olecranon. Calculate the projection points of the contour points on the proximal ulna feature plane onto the local coordinate system of the proximal ulna, and determine the projection point that is farthest from the origin along the Y-axis of the local coordinate system of the proximal ulna. Take the contour point corresponding to the projection point that is farthest from the origin as the tip of the coronoid process of the ulna. Establish the midline planes of the initial ulnar end point and the initial radial end point of the ulnar olecranon, and obtain the intersection contour of the ulnar three-dimensional model and the midline planes; The arc segment in the region enclosed by the proximal ulnar feature plane and the posterior side of the corrected proximal ulnar coronal plane in the intersecting contour is defined proximally. Calculate the curvature of each contour point in the arc segment, and take the contour point with the largest curvature as the initial olecranon tip of the ulna.

9. The intelligent recognition method for forearm bone landmarks according to claim 8, characterized in that, The step of performing local optimization on the initial forearm bony landmarks within a preset neighborhood to determine the final identified forearm bony landmarks specifically includes: Move the origin of the local coordinate system at the proximal end of the ulna a predetermined distance in the negative X-axis direction, and denote it as the base point. ; Establish from the base point A vector pointing to the current ulnar landmark is used, and the projection of each point within a preset neighborhood onto the vector is calculated to obtain the neighborhood point corresponding to the maximum projection value; the current ulnar landmark includes the ulnar end point of the initial olecranon and the radial end point of the initial olecranon. Replace the current ulnar landmark with the neighborhood point corresponding to the maximum projection value, and return to step "Establish from base point". The vector pointing to the current ulnar landmark is calculated, and the projection of each point within a preset neighborhood onto the vector is obtained. The neighborhood point corresponding to the maximum projection value is obtained until convergence, thus obtaining the local optimum of the current ulnar landmark. Local optimization is performed on the initial olecranon tip of the ulna to obtain the locally optimal olecranon tip of the ulna.

Images