Shoulder joint skin point cloud data driven deltoid muscle outer surface modeling method
By collecting shoulder joint skin point cloud data and deltoid MRI, combined with improved clustering algorithms and feature extraction, the accuracy of muscle morphology modeling in the existing technology is solved, high-precision muscle external surface reconstruction is achieved, and the diagnostic accuracy of chronic muscle injury is improved.
Patent Information
- Application Number
- CN202510462900.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-14
- Publication Date
- 2025-07-08
AI Technical Summary
The prior art is difficult to reconstruct subcutaneous muscle morphology with high accuracy, especially in the diagnosis of chronic muscle injury. There is a lack of objective modeling methods, which leads to difficulty in diagnosis and strong subjectivity.
By collecting shoulder joint skin point cloud data, combining deltoid MRI, the mapping relationship between surface skin and muscle is constructed and the shape of the outer surface of the muscle is reconstructed.
High-precision muscle surface modeling is achieved, objective diagnostic assistance is provided, and the diagnostic accuracy and motor evaluation ability of chronic muscle injury are improved.
Smart Images
Figure CN120279189A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of human muscle modeling, and particularly relates to a method for modeling the outer surface of the deltoid muscle driven by shoulder joint skin point cloud data. Background Art
[0002] Muscle injury is a common type of physiological injury, mainly divided into acute injury and chronic injury. Due to changes in living habits and working methods, the number of people with chronic injuries shows an increasing trend year by year. The proportion of acute injuries in hospital outpatient clinics is less than 5%, and chronic cumulative injury is the most common form of muscle injury. For muscle injuries with obvious tissue changes, accurate diagnosis can generally be made through medical examinations such as muscle enzyme determination, muscle ultrasound, muscle biopsy, and muscle magnetic resonance imaging. However, for chronic muscle injuries with less obvious tissue changes, it is difficult to make a definite diagnosis using the above examination methods because the injury is not obvious, and these examinations take a long time and cost a lot. Currently, the main diagnostic methods for chronic muscle injuries are still palpation and observation, which are highly subjective and require a high level of professional quality of doctors. The muscle morphology is affected by its state, and different morphologies represent inconsistent appearances and structures of the muscle. Therefore, muscle injury is often accompanied by changes in the surface morphology, and these morphological changes can reflect the state of an individual's muscle. By analyzing the muscle surface morphology, the situation of muscle injury can be effectively evaluated. However, the changes in muscle morphology are diverse and complex, and its morphological modeling is crucial for injury diagnosis and motion assessment, and at the same time, it also poses higher requirements for the human muscle morphology model.
[0003] With the development of medical image information visualization technology, the construction of human muscle morphology models has been widely used in fields such as tissue numerical simulation analysis. Currently, combining point cloud data with medical imaging technology has become a common method for obtaining the morphology of muscle tissue.
[0004] By collecting surface skin point cloud data, the morphology of the subcutaneous muscle can be estimated, thereby realizing the digital representation of the complex surface of the muscle. The muscle morphology model based on skin point cloud data can be used as an auxiliary means for doctors to diagnose chronic muscle injuries. The structure of the muscle system is complex, with muscles intertwined and interconnected with each other. Each muscle has its unique shape, size, and position, and their connection methods and movement methods are also different. Therefore, in the modeling process, accurately grasping muscle characteristics and expressing them with high precision are key problems that need to be solved urgently. Surface skin point cloud data can be obtained by non-contact high-precision instruments. To infer the surface of the subcutaneous muscle using these data, it is necessary to clarify the data mapping relationship between the two. The establishment of this data mapping relationship is also crucial for accurately reconstructing the morphology of the subcutaneous muscle. Summary of the Invention
[0005] The technical problem solved by the present invention is to provide a method for modeling the outer surface of the deltoid muscle driven by shoulder joint skin point cloud data. The basic solution provided by the present invention: A method for modeling the outer surface of the deltoid muscle driven by shoulder joint skin point cloud data, including the following steps: S1. Collect the point cloud data of the shoulder joint skin surface and the deltoid muscle MRI. S2. Process the original data of the collected shoulder joint skin surface point cloud data, and identify the abnormal point cloud data of the processed shoulder joint skin surface point cloud data. S3. Use the preprocessed shoulder joint skin surface point cloud data to construct a feature dimension describing the shape change of the skin surface, and identify and remove the redundant points in the target area of the point cloud through an improved clustering algorithm, and divide the corresponding surface skin contour of the deltoid muscle with the shoulder joint skin surface point cloud data. S4. Determine the subcutaneous tissue thickness according to the shoulder joint skin surface point cloud data and the deltoid muscle outer MRI, and construct the mapping relationship between the deltoid muscle and the surface skin according to the subcutaneous tissue thickness. S5. Reconstruct the shape of the outer surface of the muscle according to the mapping relationship between the surface skin and the outer surface of the deltoid muscle.
[0006] Further, the S2 includes the following steps: S21. Adopt the KD-Tree algorithm to calculate the distance between the data points and the adjacent points, identify the noise points in the sparse area of the shoulder joint skin surface point cloud data, and after identifying the noise points, remove the points marked as noise from the shoulder joint skin surface point cloud data to complete the noise reduction of the point cloud data. S22. Use the voxel filtering method to downsample the shoulder joint skin surface point cloud data.
[0007] Further, the S2 also includes the following steps: S23. Determine the feature dimension of the collected shoulder joint skin surface point cloud data through curvature features, normal vector features, and point density features, specifically including the following steps: S231. Select the points for which the curvature needs to be calculated in the collected shoulder joint skin surface point cloud data , and delimit a neighborhood centered on :
[0008] where , represents the points included in the neighborhood, and by taking the derivative of each of them respectively, we can obtain:
[0009] S232. Obtain the parameters describing the surface through the surface equation:
[0010] S233. Calculate the neighborhood information of each point based on the processed point cloud coordinate dataset, and determine the normal vector of the point cloud at each point accordingly, thereby obtaining a set of normal vectors through processing. :
[0011] S234. Then downsample the point cloud data containing normal vector information to obtain a new point cloud and the corresponding set of normal vectors :
[0012]
[0013] S235. Calculate the neighborhood information of each point in the new set of point clouds Determine the included angle between the normal vectors of each point in the neighborhood and the central point, calculate the standard deviation, average value, and median of the included angles within the neighborhood, and obtain the initial eigenvalue dataset of all sample points ; S236. Perform a third resampling on the eigenvalue dataset to reduce the dataset density and obtain the resampled eigenvalue dataset .
[0014] Furthermore, step S2 further includes the following steps: S24. Use the principal component analysis method for data dimensionality reduction, which specifically includes the following steps: S241. Standardize each attribute in the eigenvalue dataset through the following formula:
[0015] S242. Calculate the covariance matrix of the dataset to reflect the correlation degree between various features in the dataset:
[0016] S243. Solve the eigenvalues and eigenvectors of the covariance matrix through the following formula to find the main components that can explain the data variability to the greatest extent:
[0017] S244. Select the main eigenvectors with the largest eigenvalues and project the original data into the new space defined by these eigenvectors to complete the dimensionality reduction.
[0018] Furthermore, step S2 further includes the following steps: S25. Select four performance index evaluation points, namely running time, normal vector direction consistency index, average angle deviation of normal vectors, and normal vector entropy value, to evaluate the abnormal data of the point cloud. The specific steps are as follows: S251. Calculate the normal vector direction consistency index through the following formula:
[0019] where N is the number of points in the point cloud, is the normal vector of the nth point, is the standard deviation function is the point and is the dot product of; S252. Calculate the average angle deviation of normal vectors through the following formula:
[0020] where , is the arithmetic mean of all normal vectors, represents the norm of the ith vector; S253. Calculate the normal vector entropy using the following formula:
[0021] where is the number of times the normal vector appears in a specific interval.
[0022] Furthermore, the said S3 includes the following steps: S311. Initialize the model parameters, mark all data points as visited, set the feature radius eps = 0.2, the minimum number of neighboring points min_Pit = 16, and set the neighborhood threshold near_threshold = 0.5; S312. Input the feature data set , including coordinates and feature temperature; S313. Select an initial point for core judgment, traverse each point , if is visited, check whether there are at least min_Pit points in its neighborhood. When not satisfied, mark as noise. When satisfied, start forming a new cluster with as the core point; S314. For the marked core points, find all the points in their neighborhoods; for each neighborhood point , calculate its proximity threshold difference from the core point. If the difference is less than near_threshold = 0.5, and itself meets the condition of becoming a core point, then continue to expand the cluster. If If the core point condition is not met, but the feature dimension difference is less than the threshold, it will be marked as a boundary point; S315. Repeat the above steps until all points are visited; if a point is neither a core point nor a boundary point, it is identified as a noise point; S316. Output the clustering set F , output each cluster identifier to form the final clustering set.
[0023] Furthermore, step S3 further includes the following steps: S32. Obtain anatomical knowledge For the feature data set perform correction, which specifically includes the following steps: S321. Initialize the model parameters, read the feature data set F and read the anatomical knowledge set K; S322. Traverse each contour feature point , and apply K for refinement and correction; S323. For each feature data , refine its feature region, for each point p in the contour , find its nearest key point and key line , and calculate the distances and ; if is less than the set threshold dp_threshold, then adjust the point p towards the key point , and the adjustment amount is (p, , ); if is less than the preset threshold dl_threshold, then adjust the point p towards the key line , and the adjustment amount is (p, , ); otherwise, do not adjust the point p; S324. Find the nearest key point of point p , calculate the distance d(p, ) between p and all key points , select the minimum distance d(p, ), and the corresponding key point is ; S325. Find the nearest key line of point p , calculate the distance d(p, ) between p and all key lines , ), key line represented by two endpoints li_start and li_end, select the minimum distance d(p, , ), and the corresponding critical line is ; S326. Calculate the distance from point p to the critical line l: Let the starting point of l be l_start and the ending point be l_end, calculate the vector and the vector , calculate the length of and the unit vector , calculate , to obtain the closest point of p on l, calculate the distance ; S327. Adjust point p according to the key point and the critical line . Let the adjustment coefficient alpha = 0.5. When , then . If , then ; S328. Output the clustering set H, output each cluster identifier, and form the final clustering set.
[0024] Furthermore, the S4 includes the following steps: S41. Determine the thickness set D between the skin surface and the muscle according to the shoulder joint skin surface point cloud data and the deltoid muscle MRI, which specifically includes the following steps: S411. Obtain a set of shoulder joint skin surface point cloud data , the corresponding unit normal vector is , and the deltoid muscle MRI is . For the i-th point , , are respectively , , , where is the data expression in stl format. Each represents a small triangular patch. The points on each small triangular patch plane satisfy:
[0025] where n is the unit normal vector of the triangle
[0026] S412. The ray equation of the i-th point is
[0027] where o is the origin, is the direction vector and t is the elongation length; S413. Substitute the ray equation into the triangular plane equation to obtain , if then the ray is parallel to the plane or the ray lies within the plane. When t and the intersection point p are , use barycentric coordinates to determine whether the point p actually lies within the boundary of the triangle:
[0028]
[0029]
[0030]
[0031]
[0032] S413. If α, β, and γ are all non - negative, then P actually lies within the boundary of the triangle. Repeat the above steps and traverse all the points in turn to find all the points mapped from the skin surface to the muscle surface, and then the thickness between the skin surface and the muscle ; S42. Combine the thickness value with the point coordinates and the normal vector direction to obtain the mapping relationship between the surface skin and the outer surface of the deltoid muscle .
[0033] Furthermore, step S5 also includes the following steps: S51. According to the mapping relationship between the surface skin and the outer surface of the deltoid muscle, map the point cloud data of the shoulder joint skin surface to the corresponding muscle points:
[0034] where represents the point on the outer surface of the muscle, represents the skin point, represents the mapping relationship; S52. Use the mapping relationship to map all skin points to the corresponding muscle points to generate the point cloud data of the outer surface of the muscle. Step S5 also includes the following steps: S51. According to the mapping relationship between the surface skin and the outer surface of the deltoid muscle, map the point cloud data of the shoulder joint skin surface to the corresponding muscle points:
[0035] where represents the point on the outer surface of the muscle, represents the skin point, represents the mapping relationship; S52. Use the mapping relationship to map all skin points to the corresponding muscle points to generate the point cloud data of the outer surface of the muscle.
[0036] The principle and advantages of the present invention are as follows: In view of the limitations and advantages of the point cloud data acquisition method, a surface skin - muscle surface data acquisition platform is built to complete the acquisition of the surface skin and deltoid muscle surface data of the shoulder joint, construct a pre - processing process for the surface shapes of various data, obtain the morphological data of the surface skin - deltoid muscle of the shoulder joint, and complete the surface modeling of the deltoid muscle and the shoulder joint skin.
[0037] A method for denoising curved surface point clouds based on an improved clustering algorithm is constructed to extract the characteristics of curved surface point cloud data. Using the density - based clustering algorithm, clusters of surface shapes are found in the point cloud spatial data. An improved density - based clustering algorithm is used to identify abnormal data in the surface point cloud and remove the abnormal points in the surface point cloud data. The surface shape features are calculated using the point cloud data information and screened to identify the morphological features in the surface point cloud, forming a method for dividing the corresponding surface skin contour of the deltoid muscle.
[0038] The collected MRI data is used to reconstruct the surface morphological models of the skin and the deltoid muscle. Based on the proposed method for dividing the surface skin contour, combined with the collected magnetic resonance muscle surface data, the distribution of the subcutaneous tissue thickness between the surface skin and the deltoid muscle is calculated. Thus, a method for reconstructing the muscle surface morphology of the deltoid muscle surface is formed from the surface skin point cloud data, the surface skin contour division, and the subcutaneous tissue thickness estimation. Brief Description of the Drawings
[0039] Figure 1 It is a schematic diagram of an embodiment of the present invention. Detailed Description of the Embodiment
[0040] The following is a further detailed description through specific embodiments: The embodiment is basically as shown in the attached Figure 1 figure: A method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data includes the following steps: S1. Collect the shoulder joint skin surface point cloud data and the deltoid muscle MRI; S2. Process the original data of the collected shoulder joint skin surface point cloud data, and identify the abnormal point cloud data in the processed shoulder joint skin surface point cloud data; S3. Use the processed shoulder joint skin surface point cloud data to construct a feature dimension describing the change of the skin surface shape, and identify and remove the redundant points in the target area of the point cloud through an improved clustering algorithm, and divide the corresponding surface skin contour of the deltoid muscle with the shoulder joint skin surface point cloud data; S4. Determine the thickness of the subcutaneous tissue based on the point cloud data of the shoulder joint skin surface and the MRI type of the deltoid muscle, and construct the mapping relationship between the deltoid muscle and the surface skin according to the thickness of the subcutaneous tissue; S5. Reconstruct the shape of the outer surface of the muscle according to the mapping relationship between the surface skin and the outer surface of the deltoid muscle.
[0041] Furthermore, the S2 includes the following steps: S21. Adopt the KD-Tree algorithm to calculate the distance between data points and adjacent points, identify the noise points in the sparse area of the point cloud data of the shoulder joint skin surface. After identifying the noise points, remove the points marked as noise from the point cloud data of the shoulder joint skin surface to complete the noise reduction of the point cloud data; S22. Downsample the point cloud data of the shoulder joint skin surface by using the voxel filtering method.
[0042] Furthermore, the S2 also includes the following steps: S23. Determine the characteristic dimensions for collecting the point cloud data of the shoulder joint skin surface through curvature features, normal vector features, and point density features, which specifically include the following steps: S231. Select the points for which the curvature needs to be calculated in the point cloud data of the shoulder joint skin surface , and delimit a neighborhood centered on :
[0043] where represents the points included in the neighborhood. Taking the derivative of each of them respectively, we can obtain:
[0044] S232. Obtain the parameters describing the surface through the surface equation:
[0045] S233. Based on the processed point cloud coordinate data set, calculate the neighborhood information of each point, and determine the normal vector of the point cloud at each point accordingly, thereby obtaining the normal vector set :
[0046] S234. Then downsample the point cloud data containing normal vector information to obtain a new point cloud and the corresponding normal vector set :
[0047]
[0048] S235. Calculate the new point cloud set For each point in [[ID=]], determine the neighborhood information of each point, calculate the angle between the normal vector of each point in the neighborhood and the center point, and calculate the standard deviation, average value, and median of the angles in the neighborhood to obtain the initial eigenvalue dataset of all sample points ; S236. Perform the third resampling on the eigenvalue dataset to reduce the dataset density and obtain the downsampled eigenvalue dataset .
[0049] Furthermore, step S2 further includes the following steps: S24. Use the principal component analysis method for data dimensionality reduction, which specifically includes the following steps: S241. Standardize each attribute in the eigenvalue dataset through the following formula:
[0050] S242. Calculate the covariance matrix of the dataset to reflect the correlation degree between various features in the dataset:
[0051] S243. Solve the eigenvalues and eigenvectors of the covariance matrix through the following formula to find the main components that can explain the data variability to the greatest extent:
[0052] S244. Select the main eigenvectors with the largest eigenvalues and project the original data into the new space defined by these eigenvectors to complete the dimensionality reduction.
[0053] Furthermore, step S2 further includes the following steps: S25. Select four performance indicators, namely running time, normal vector direction consistency index, normal vector average angle deviation, and normal vector entropy value, to evaluate the point cloud abnormal data, which specifically includes the following steps: S251. Calculate the normal vector direction consistency index through the following formula:
[0054] where N is the number of points in the point cloud, is the normal vector of the nth point, is the standard deviation function is the point and is the dot product of; S252. Calculate the normal vector average angle deviation through the following formula:
[0055] where , is the arithmetic mean of all normal vectors, represents the norm of the i-th vector; S253 calculates the normal vector entropy using the following formula:
[0056] where is the number of times the normal vector appears within a specific interval.
[0057] Further, the said S3 includes the following steps: S311. Initialize the model parameters, mark all data points as visited, set the feature radius eps = 0.2, the minimum number of neighboring points min_Pit = 16, and set the proximity threshold near_threshold = 0.5; S312. Input the feature dataset , including coordinates and feature temperature; S313. Select an initial point for core judgment, traverse each point , if is visited, check whether there are at least min_Pit points in its neighborhood. When not satisfied, mark as noise. When satisfied, start forming a new cluster with as the core point; S314. For the marked core points, find all the points in their neighborhoods; for each neighborhood point , calculate the difference between its proximity threshold and the core point. If the difference is less than near_threshold = 0.5, and itself meets the condition of becoming a core point, continue to expand the cluster. If does not meet the core point condition, but the feature dimension difference is less than the threshold, mark it as a boundary point; S315. Repeat the above steps until all points are visited; if a point is neither a core point nor a boundary point, it is identified as a noise point; F S316. Output the clustering set
[0058] Further, the said S3 also includes the following steps: S32. Obtain anatomical knowledge Revise the feature dataset , which specifically includes the following steps: S321. Initialize the model parameters, read the feature dataset F and the anatomical knowledge set K; S322. Traverse each contour feature point , and apply K for refinement and correction; S323. For each piece of feature data , refine its feature region. For each point p in the contour , find its nearest key point and key line , and calculate the distances to the key point and the key line and ; if is less than the set threshold, dp_threshold, then adjust point p towards the key point , with the adjustment amount being (p, , ); if is less than the preset threshold dl_threshold, then adjust point p towards the key line , with the adjustment amount being (p, , ); otherwise, do not adjust point p; S324. Find the nearest key point of point p, calculate the distance d(p, ) between p and all key points , select the minimum distance d(p, ), and the corresponding key point is ; S325. Find the nearest key line of point p, calculate the distance d(p, ) between p and all key lines , ). The key line is represented by two endpoints li_start and li_end. Select the minimum distance d(p, , ), and the corresponding key line is ; S326. Calculate the distance from point p to the key line l: Let the starting point of l be l_start and the ending point be l_end. Calculate the vectors and vector , calculate the length of and the unit vector , calculate , obtain the nearest point of p on l, and calculate the distance ; S327. Adjust point p according to the key point and the key line . Let the adjustment coefficient alpha = 0.5. When , then , if , then ; S328. Output the clustering set H, output each cluster identifier, and form the final clustering set.
[0059] Further, the S4 includes the following steps: S41. Determine the thickness set D between the skin surface and the muscle according to the shoulder joint skin surface point cloud data and the deltoid muscle MRI, which specifically includes the following steps: S411. Obtain a set of shoulder joint skin surface point cloud data , and the corresponding unit normal vector is , and the deltoid muscle MRI is , where for the i-th point , , are respectively , , , where is the data expression in stl format, and each represents a small triangular patch, and the points on each small triangular patch plane satisfy:
[0060] where n is the unit normal vector of the triangle
[0061] S412. The ray equation of the i-th point is
[0062] where o is the origin, is the direction vector, and t is the elongation length; S413. Substitute the ray equation into the triangular plane equation to obtain , if then the ray is parallel to the plane or the ray is in the plane. When t and the intersection point p are , use the barycentric coordinate to judge and confirm whether the point p is actually located inside the boundary of the triangle:
[0063]
[0064]
[0065]
[0066]
[0067] S413. If α, β, and γ are all non - negative, then P is actually located inside the boundary of the triangle. Repeat the above steps. After traversing all the points in sequence, all the points mapped from the skin surface to the muscle surface can be obtained, and then the thickness between the skin surface and the muscle ; S42. Combine the thickness value with the point coordinates and the normal vector direction to obtain the mapping relationship between the surface skin and the outer surface of the deltoid muscle 。
[0068] Furthermore, step S5 also includes the following steps: S51. According to the mapping relationship between the surface skin and the outer surface of the deltoid muscle, map the point cloud data of the skin surface of the shoulder joint to the corresponding muscle points:
[0069] where represents the points on the outer surface of the muscle, represents the skin points, represents the mapping relationship; S52. Use the mapping relationship to map all skin points to the corresponding muscle points, and generate the point cloud data of the outer surface of the muscle. Step S5 also includes the following steps: S51. According to the mapping relationship between the surface skin and the outer surface of the deltoid muscle, map the point cloud data of the skin surface of the shoulder joint to the corresponding muscle points:
[0070] where represents the points on the outer surface of the muscle, represents the skin points, represents the mapping relationship; S52. Use the mapping relationship to map all skin points to the corresponding muscle points, and generate the point cloud data of the outer surface of the muscle.
[0071] The principle and advantages of the present invention are as follows: Aiming at the limitations and advantages of the point cloud data acquisition method, a surface skin - muscle surface data acquisition platform is built to complete the acquisition of the surface skin data and deltoid muscle surface data of the shoulder joint, construct a pre - processing process for the surface shapes of various data, obtain the surface skin - deltoid muscle morphological data of the shoulder joint, and complete the modeling of the deltoid muscle and the skin surface of the shoulder joint.
[0072] Construct a surface point cloud noise reduction method based on an improved clustering algorithm to extract the characteristics of the surface point cloud data. Use the density - based clustering algorithm to discover the clustering of the surface shape in the point cloud spatial data. Adopt an improved density - based clustering algorithm to identify the abnormal data of the surface point cloud and remove the abnormal points in the surface point cloud data. Calculate and screen the surface shape characteristics using the point cloud data information, identify the morphological characteristics in the surface point cloud, and form a method for dividing the corresponding surface skin contour of the deltoid muscle.
[0073] Using the collected MRI data, reconstruct the surface morphological models of the skin and deltoid muscle. Based on the proposed method for dividing the surface skin contour, combined with the collected magnetic resonance muscle surface data, calculate the distribution of the subcutaneous tissue thickness between the surface skin and the deltoid muscle. Thus, a method for reconstructing the muscle surface morphology on the deltoid muscle surface is formed, which includes surface skin point cloud data, surface skin contour division, and subcutaneous tissue thickness estimation.
[0074] The above are only embodiments of the present invention. Specific structures and common knowledge such as characteristics in the solution are not described in detail here. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention belongs before the filing date or the priority date, can know all the existing technologies in this field, and have the ability to apply conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application, combine their own abilities to improve and implement this solution. Some typical well-known structures or well-known methods should not become obstacles for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, which will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be based on the content of its claims, and the specific implementation manners described in the specification can be used to interpret the content of the claims.
Claims
1. A method for modeling the outer surface of the deltoid muscle driven by the skin point cloud data of the shoulder joint, characterized in that: It includes the following steps: S1. Collect the point cloud data of the shoulder joint skin surface and the MRI of the deltoid muscle; S2. Process the original data of the collected point cloud data of the shoulder joint skin surface, and identify the abnormal point cloud data of the processed point cloud data of the shoulder joint skin surface; S3. Use the processed point cloud data of the shoulder joint skin surface to construct the feature dimensions describing the shape change of the skin surface, identify and remove the redundant points in the target area of the point cloud through an improved clustering algorithm, and divide the corresponding surface skin contour of the deltoid muscle with the point cloud data of the shoulder joint skin surface; S4. Determine the subcutaneous tissue thickness according to the point cloud data of the shoulder joint skin surface and the external MRI of the deltoid muscle, and construct the mapping relationship between the deltoid muscle and the surface skin according to the subcutaneous tissue thickness; S5. Reconstruct the shape of the outer surface of the muscle according to the mapping relationship between the surface skin and the outer surface of the deltoid muscle.
2. The modeling method of the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 1, wherein: The S2 includes the following steps: S21. Use the KD-Tree algorithm to calculate the distance between the data points and the adjacent points, identify the noise points in the sparse area of the point cloud data of the shoulder joint skin surface. After identifying the noise points, remove the points marked as noise from the point cloud data of the shoulder joint skin surface to complete the noise reduction of the point cloud data; S22. Use the voxel filtering method to downsample the point cloud data of the shoulder joint skin surface.
3. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 2, wherein: The S2 also includes the following steps: S23. Determine the feature dimensions of the collected point cloud data of the shoulder joint skin surface through curvature features, normal vector features, and point density features. The specific steps are as follows: S231. Select the points for which the curvature needs to be calculated from the point cloud data on the skin surface of the shoulder joint , and as the center to delimit a neighborhood: Among them , indicating the points included in the neighborhood, and taking the derivative of each of them respectively, we can obtain: S232. Obtain the parameters describing the surface through the surface equation: S233. Calculate the neighborhood information of each point based on the processed point cloud coordinate data set, and determine the normal vector of the point cloud at each point accordingly, thereby obtaining a normal vector set through processing : S234. Then, downsample the point cloud data containing normal vector information to obtain a new point cloud and the corresponding normal vector set : S235. Calculate the new point cloud set For each point in it, determine the neighborhood information of each point, calculate the angle between the normal vector of each point in the neighborhood and the center point, calculate the standard deviation, average value, and median of the angles in the neighborhood, and obtain the initial eigenvalue data set of all sample points ; S236. The eigenvalue dataset was resampled for the third time to reduce the dataset density and obtain the downsampled eigenvalue dataset. .
4. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 3, wherein: The S2 also includes the following steps: S24. Use the principal component analysis method for data dimensionality reduction. The specific steps are as follows: S241. Standardize each attribute in the eigenvalue dataset through the following formula: S242. Calculate the covariance matrix of the dataset to reflect the correlation degree between each feature in the dataset: S243. Solve the eigenvalues and eigenvectors of the covariance matrix through the following formula to find the main components that can explain the data variability to the greatest extent: S244. Select the main eigenvectors with the largest eigenvalues, and project the original data into the new space defined by these eigenvectors to complete the dimensionality reduction.
5. The modeling method of the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 4, wherein: The S2 also includes the following steps: S25. Select four performance indicators, namely running time, normal vector direction consistency index, normal vector average angle deviation, and normal vector entropy value, to evaluate the abnormal point cloud data. The specific steps are as follows: S251. Calculate the normal vector direction consistency index through the following formula: where N is the number of points in the point cloud, is the normal vector of the n-th point, is the standard deviation function is the point and is the dot product of; S252. Calculate the normal vector average angle deviation through the following formula: wherein , is the arithmetic mean of all normal vectors, represents the norm of the i-th vector; S253. Calculate the normal vector entropy using the following formula: Among them is the number of times the normal vector appears within a specific interval.
6. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 5, wherein: The S3 includes the following steps: S311. Initialize the model parameters, mark all data points as unvisited, set the feature radius eps = 0.2, the minimum number of adjacent points min_Pit = 16, and set the adjacent threshold near_threshold = 0.5; S312. Input feature data set , including coordinates and characteristic temperature; S313. Select an initial point for core judgment and traverse each point , if is visited, check whether there are at least min_Pit points in its neighborhood. When not satisfied, mark as noise. When satisfied, start forming a new cluster with it as the core point; S314. For the marked core points, find all the points in the neighborhood; for each neighborhood point , calculate its proximity threshold difference from the core point. If the difference is less than near_threshold = 0.5, and itself meets the condition of becoming a core point, then continue to expand the cluster. If it does not meet the core point condition, but the feature dimension difference is less than the threshold, mark it as a border point; S315. Repeat the above steps until all points are visited; if a point is neither a core point nor a boundary point, it is identified as a noise point; S316. Output the clustering set F , output each cluster identifier to form the final clustering set.
7. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 6, wherein: The said S3 further includes the following steps: S32. Obtain anatomical knowledge For the feature data set Perform correction, which specifically includes the following steps: S321. Initialize the model parameters, and read the feature data set F and the anatomical knowledge set K; S322. Traverse each contour feature point , apply K for refinement and correction; S323. For each piece of feature data , refine its feature region. For each point p in the contour , find its nearest key point and key line , and calculate the distances to the key point and the key line ; if is less than the set threshold dp_threshold, then adjust point p towards the key point , with the adjustment amount being (p, , ); if is less than the preset threshold dl_threshold, then adjust point p towards the key line , with the adjustment amount being (p, , ); otherwise, do not adjust point p; S324. Find the nearest key point of point p , calculate the distance d(p, between p and all key points ), select the minimum distance d(p, ), and the corresponding key point is ; S325. Find the nearest critical line of point p , calculate the distance d(p, from p to all critical lines , ), where the critical line is represented by two endpoints li_start and li_end, and select the minimum distance d(p, , ). The corresponding critical line is ; S326. Calculate the distance from point p to the critical line l: Let the starting point of l be l_start and the ending point be l_end, and calculate the vector and the vector . Calculate the length of and the unit vector . Calculate to obtain the closest point of p on l, and calculate the distance ; S327. Adjust point p according to the key points and the key lines . Set the adjustment coefficient alpha = 0.
5. When , then . If , then ; S328. Output the clustering set H, and output each cluster identifier to form the final clustering set.
8. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 7, wherein: The said S4 includes the following steps: S41. Determine the thickness set D between the skin surface and the muscle according to the shoulder joint skin surface point cloud data and the deltoid muscle MRI, which specifically includes the following steps: S411. Obtain a set of point cloud data on the skin surface of the shoulder joint , and the corresponding unit normal vector is , the deltoid muscle MRI is , where for the i-th point , , are respectively , , , where is the data expression in stl format, and each represents a small triangular patch, and the points on each small triangular patch plane satisfy: where n is the unit normal vector of the triangle S412. The ray equation of the i-th point is where o is the origin, is the direction vector, and t is the elongation length; S413. Substitute the ray equation into the triangular plane equation to obtain , if then the ray is parallel to the plane or the ray lies within the plane. When t and the intersection point p are , use barycentric coordinates to determine whether the point p actually lies within the boundaries of the triangle: S413. If α, β, and γ are all non - negative, then P is actually located inside the boundary of the triangle. Repeat the above steps. After traversing all the points in sequence, all the points mapped from the skin surface to the muscle surface can be obtained, and then the thickness between the skin surface and the muscle ; S42. Combine the thickness value with the point coordinates and the normal vector direction to obtain the mapping relationship between the surface skin and the outer surface of the deltoid muscle .
9. The method for modeling the outer surface of the deltoid muscle driven by the shoulder joint skin point cloud data according to claim 8, wherein: The said S5 further includes the following steps: S51. Map the point cloud data of the shoulder joint skin surface to the corresponding muscle points according to the mapping relationship between the surface skin and the outer surface of the deltoid muscle: wherein represents a point on the outer surface of the muscle, represents a skin point, represents a mapping relationship; S52. Use the mapping relationship to map all skin points to the corresponding muscle points to generate the point cloud data of the outer surface of the muscle.