A detection method for back spine points based on multi-scale constraints

Abstract

The present application relates to the technical field of scoliosis, and in particular proposes a method for detecting back spine points based on multi-scale constraints. The method includes obtaining three-dimensional contour line data of a human back cross-section, performing polynomial fitting based on the three-dimensional contour line data of the human back cross-section to obtain a fitted contour line; calculating the curvature, slope, and asymmetric function value based on the fitted contour line to extract back spine points; performing multi-scale constraints based on the back spine points to obtain continuous back spine points; and performing non-uniform rational B-spline fitting based on the continuous back spine points to obtain a back spine line. This method solves the problem of poor continuity in detecting back spine points, and uses three methods to constrain the values ​​of back spine points, thereby ensuring the accuracy of back spine point detection. Human intervention is reduced in the process of back spine point extraction, and the entire process is radiation-free, so it is suitable for detecting the midline of the adolescent spine.

Description

Technical Field

[0001] The present application relates to the technical field of scoliosis, and in particular proposes a detection method for back spine points based on multi-scale constraints. Background Art

[0002] Scoliosis is a three-dimensional spinal deformity characterized by lateral curvature of one or more segments of the spine in the coronal plane, deviating from the midline and accompanied by vertebral rotation. Scoliosis can cause body deformities and, in severe cases, even lead to cardiopulmonary dysfunction. With the increasing prevalence of adolescent idiopathic scoliosis in recent years, early screening and timely intervention can reduce its severity. Therefore, my country has included scoliosis in its student monitoring and annual health checkups.

[0003] Traditional scoliosis screening methods include general examinations, Adams tests, and scoliometers. These manual screening methods suffer from low efficiency and accuracy. Imaging tests include X-rays, EOS, and Moiré fringe detection. X-rays and EOS detection of the spinal line expose adolescents to unnecessary radiation exposure. Using asymmetric surface curvature functions in Moiré fringe images can create false spinal points, resulting in discontinuous extraction of dorsal spinal points and low accuracy of the dorsal spinal midline. Summary of the Invention

[0004] In view of the above problems, a back spine point detection method based on multi-scale constraints is proposed to solve the problem.

[0005] A back spine point detection method based on multi-scale constraints, the method comprising:

[0006] Acquiring three-dimensional contour line data of a human back cross section, and performing polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line;

[0007] Calculating curvature, slope, and asymmetry function values ​​based on the fitted contour line, and extracting back spine points;

[0008] Performing multi-scale constraints on the back spine points to obtain continuous back spine points;

[0009] The midline of the dorsal spine is obtained by performing non-uniform rational B-spline fitting on the continuous dorsal spine points.

[0010] Preferably, the acquiring of the three-dimensional contour line data of the human back cross section and performing polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line include:

[0011] The three-dimensional contour line data of the human back section is collected by using an automated scoliosis screening system, wherein the three-dimensional contour line data of the human back section is hundreds of parallel and equidistant three-dimensional contour line data of the human back section;

[0012] Projecting the three-dimensional contour line data of the human back cross section onto a two-dimensional plane to obtain a contour line of the two-dimensional plane;

[0013] According to the contour line of the two-dimensional plane, the contour line of the two-dimensional plane is fitted using a polynomial fitting technique to obtain a fitted contour line.

[0014] Preferably, calculating the curvature, slope, and asymmetry function value based on the fitted contour line and extracting the back spine point includes:

[0015] Calculating the curvature of each point on the fitted contour line, and calculating the extreme value of the curvature on the fitted contour line using a differential method;

[0016] Calculating the centroid coordinates of the three-dimensional contour line data of the human back cross section using a centroid formula;

[0017] According to the centroid coordinates, extract the extreme values ​​around the centroid coordinates to obtain a back spine point set E;

[0018] Calculate the slope of each point on the fitted contour line;

[0019] Obtaining a back spine point set F according to a variation characteristic of the slope on the fitting contour line;

[0020] Sampling points are obtained based on the fitting contour line, with a sampling interval of 0.1 mm;

[0021] Calculating the maximum principal curvature, the minimum principal curvature, and the direction of the principal curvature of the sampling point, and substituting the maximum principal curvature, the minimum principal curvature, and the direction of the principal curvature into the asymmetric function to obtain a relative minimum value;

[0022] According to the centroid coordinates, extract the relative minimum value around the centroid coordinates to obtain a back spine point set Q;

[0023] The mean of the back spine point set E, the back spine point set F, and the back spine point set Q on the fitting contour line is calculated, and the point on each fitting contour line closest to the mean is extracted as the back spine point.

[0024] Preferably, multi-scale constraints are applied to the back spine points to obtain continuous back spine points, including:

[0025] Calculate the average value of two adjacent points among the back spine points;

[0026] Extracting the point on the fitted contour line closest to the average value as the latest spine point, traversing the contour line multiple times until the back spine point is no longer updated, and obtaining a back spine point set P;

[0027] According to the back spine point set P, the average value of the back spine points extracted at multiple scales is obtained, and the contour line is traversed multiple times until the back spine points are no longer updated, thereby obtaining a continuous back spine point set. .

[0028] Preferably, performing non-uniform rational B-spline fitting on the continuous back spine points to obtain the back spine midline includes:

[0029] According to the dorsal spine points , the parameter value is obtained by using the accumulated chord length parameter method to obtain the node vector;

[0030] Calculating a cubic B-spline basis function based on the node vector;

[0031] Substituting the result calculated by the cubic B-spline basis function into the weighted control matrix, and solving the linear equations using the tangent vector boundary condition to obtain the control vertex;

[0032] Fitting is performed based on the control vertices to obtain the dorsal spine midline.

[0033] The beneficial effects of the technical solution of the present invention are as follows: the present application proposes a method for detecting back spine points based on multi-scale constraints, the method comprising obtaining three-dimensional contour line data of a human back cross section, performing polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line; calculating the curvature, slope, and asymmetric function value based on the fitted contour line to extract the back spine points; performing multi-scale constraints based on the back spine points to obtain continuous back spine points; and performing non-uniform rational B-spline fitting based on the continuous back spine points to obtain a back spine line. The above scheme solves the problem of poor continuity in detecting back spine points, and utilizes three methods to constrain the value pairs of back spine points, thereby ensuring the accuracy of back spine point detection, reducing human intervention in the process of back spine point extraction, and the entire process is radiation-free, so it is suitable for detecting the midline of the spine of adolescents. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solution of this application, the following is a brief introduction to the drawings required for the description of this application. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 A schematic diagram of a multi-scale constraint-based back spine point detection method provided in an example of the present application is shown;

[0036] Figure 2 A schematic diagram of three-dimensional contour line data of a human back provided in an example of the present application is shown;

[0037] Figure 3 The dorsal spine midline obtained in an example of the present application is shown. Implementation Method

[0038] In order to more clearly illustrate the implementation method of the present invention, the technical solution of the present invention is further described in detail below in conjunction with the drawings and embodiments. The described embodiment is only one embodiment of the present application, rather than all embodiments.

[0039] In view of the shortcomings of existing scoliosis detection methods, such as low detection efficiency, low detection accuracy, and radiation, and the appearance of false spinal points when using the surface curvature asymmetric function in the Moire fringe image, which makes the extracted back spinal points discontinuous and leads to low accuracy of the back spine midline, this application extracts a detection method for back spinal points based on multi-scale constraints, so as to achieve the purpose of continuous, high-precision, radiation-free and high-efficiency detection of spinal points.

[0040] Please refer to Figure 1 , shows a flowchart of a method for detecting back spine points based on multi-scale constraints provided by an embodiment of the present application. The algorithm includes the following steps:

[0041] S1. Acquire three-dimensional contour line data of a human back cross section, and perform polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line;

[0042] S2. Calculating curvature, slope, and asymmetry function values ​​based on the fitted contour line, and extracting back spine points;

[0043] S3, performing multi-scale constraints based on the back spine points to obtain continuous back spine points;

[0044] S4. Perform non-uniform rational B-spline fitting based on the continuous back spine points to obtain the back spine midline.

[0045] The above-mentioned automatic detection algorithm for back rotation angle is further described through an embodiment.

[0046] As described in step S1, three-dimensional contour line data of the human back cross section is obtained, and a polynomial fitting technique is performed on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line.

[0047] In one embodiment of the present invention, the specific process of "performing polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line" in step S1 can be further explained in combination with the following description.

[0048] As described in the following steps:

[0049] The three-dimensional contour line data of the human back section is collected by using an automated scoliosis screening system, wherein the three-dimensional contour line data of the human back section is hundreds of parallel and equidistant three-dimensional contour line data of the human back section;

[0050] It should be noted that the automated scoliosis screening system used to collect the three-dimensional contour line data of the human back section requires the subject to be in the same posture as the Adams flexion test, that is, the subject stands with his knees and feet together, arms straight and palms together, and the torso slowly bends forward about 90 degrees. The laser line in the automated scoliosis screening system scans the subject, and hundreds of parallel and equidistant three-dimensional contour line data of the human back section can be obtained. Figure 2 As shown;

[0051] Projecting the three-dimensional contour line data of the human back cross section onto a two-dimensional plane to obtain a contour line of the two-dimensional plane;

[0052] According to the contour line of the two-dimensional plane, the contour line of the two-dimensional plane is fitted using a polynomial fitting technique to obtain a fitted contour line, wherein the polynomial fitting calculation formula is as follows:

[0053]

[0054] Where n is the polynomial order, are the coefficients of the polynomial, For univariate input.

[0055] As described in step S2, the curvature, slope, and asymmetry function value are calculated based on the fitted contour line, and the back spine points are extracted.

[0056] In one embodiment of the present invention, the specific process of "extracting back spine points" in step S2 can be further explained in combination with the following description.

[0057] As described in the following steps:

[0058] Calculate the curvature of each point on the fitted contour line, and use the differential method to calculate the extreme value of the curvature on the fitted contour line, wherein the calculation formula of the curvature is as follows:

[0059]

[0060] in, is a polynomial function, is the first-order derivative of the polynomial, is the second derivative of the polynomial;

[0061] The centroid coordinates of the three-dimensional contour line data of the human back cross section are calculated using the centroid formula, wherein the centroid coordinate calculation formula is as follows:

[0062]

[0063] Where n is the number of point clouds;

[0064] According to the centroid coordinates, extract the extreme values ​​around the centroid coordinates to obtain a back spine point set E;

[0065] Calculate the slope of each point on the fitted contour line;

[0066] Obtaining a back spine point set F according to a spine variation feature of the slope on the fitting contour line;

[0067] It should be noted that, when the spinal change characteristics of the fitted contour line are obvious, the characteristic points are extracted as the back spine points; when the spinal change characteristics of the fitted contour line are not obvious, the point on the contour line to be determined in the fitted contour line where the slope difference between the slope of the previous or next contour line is the smallest is used as the back spine point of the contour line to be determined;

[0068] Sampling points are obtained based on the fitting contour line, with a sampling interval of 0.1 mm;

[0069] The maximum principal curvature, minimum principal curvature, and direction of the principal curvature of the sampling point are calculated, and the maximum principal curvature, minimum principal curvature, and direction of the principal curvature are substituted into the asymmetric function to obtain a relative minimum value. The asymmetric function calculation formula is as follows:

[0070]

[0071] Among them, l and r are the mirror points of point p, , , , L is the length of the two mirror points, is the maximum principal curvature, is the minimum principal curvature, and are the principal curvature directions of the two mirror points;

[0072] According to the centroid coordinates, extract the relative minimum value around the centroid coordinates to obtain a back spine point set Q;

[0073] The mean of the back spine point set E, the back spine point set F, and the back spine point set Q on the fitting contour line is calculated, and the point on each fitting contour line closest to the mean is extracted as the back spine point.

[0074] As described in step S3, multi-scale constraints are performed based on the back spine points to obtain continuous back spine points.

[0075] In one embodiment of the present invention, the specific process of "obtaining continuous back spine points" in step S3 can be further explained in combination with the following description.

[0076] As described in the following steps:

[0077] Calculate the average value of two adjacent points among the back spine points;

[0078] Extracting the point on the fitted contour line closest to the average value as the latest spine point, traversing the contour line multiple times until the back spine point is no longer updated, and obtaining a back spine point set P;

[0079] According to the back spine point set P, the average value of the back spine points extracted at multiple scales is obtained, and the contour line is traversed multiple times until the back spine points are no longer updated, thereby obtaining a continuous back spine point set. .

[0080] As described in step S4, non-uniform rational B-spline fitting is performed based on the continuous back spine points to obtain the back spine midline.

[0081] In one embodiment of the present invention, the specific process of "performing non-uniform rational B-spline fitting on the back spine points to obtain the back spine midline" in step S3 can be further explained in conjunction with the following description.

[0082] As described in the following steps:

[0083] According to the dorsal spine points , the parameter value is obtained by using the accumulated chord length parameter method to obtain the node vector ;

[0084] According to the node vector, the cubic B-spline basis function is calculated, wherein the cubic B-spline basis function calculation formula is as follows:

[0085]

[0086] in, is the cubic canonical B-spline basis function;

[0087] Substituting the result calculated by the cubic B-spline basis function into the weighted control matrix, and solving the linear equations using the tangent vector boundary condition to obtain the control vertex;

[0088] According to the control vertices, the midline of the back spine is obtained as follows: Figure 3 shown.

[0089] The above is a detailed introduction to the multi-scale constrained back spine point detection method provided by this application. The description of the above embodiments is only used to help understand the method of this application and its core idea. For technical personnel in this field, various other combinations that do not deviate from the essence of this application can be made based on the technologies disclosed in this application. These combinations are still within the scope of protection of this application.

Claims

1. A back spine point detection method based on multi-scale constraints, characterized in that: The method comprises: Acquiring three-dimensional contour line data of a human back cross section, and performing polynomial fitting based on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line; Calculating the curvature, slope, and asymmetry function value based on the fitted contour line and extracting the back spine points includes the following steps: Step 1: Calculate the curvature of each point on the fitted contour line, and calculate the extreme value of the curvature on the fitted contour line using a differential method; Step 2: Calculate the centroid coordinates of the three-dimensional contour line data of the human back cross section using a centroid formula; Step 3: Based on the centroid coordinates, extract the extreme values ​​around the centroid coordinates to obtain a back spine point set E; Step 4: Calculate the slope of each point on the fitted contour line; Step 5: Obtain a back spine point set F based on the variation characteristics of the slope on the fitted contour line; Step 6: Sampling is performed based on the fitted contour line to obtain sampling points, with a sampling interval of 0.1 mm; Step 7: Calculate the maximum principal curvature, minimum principal curvature, and direction of the principal curvature of the sampling point, and substitute the maximum principal curvature, minimum principal curvature, and direction of the principal curvature into the asymmetric function to obtain a relative minimum value; Step 8: Calculate the centroid coordinates of the three-dimensional contour line data of the human back cross section using a centroid formula; Step 9: Based on the centroid coordinates, extract the relative minimum value around the centroid coordinates to obtain the back spine point set Q; Step 10: Calculating the mean of the back spine point set E, the back spine point set F, and the back spine point set Q on the fitted contour line, and extracting the point on each fitted contour line closest to the mean as the back spine point; Performing multi-scale constraints on the back spine points to obtain continuous back spine points; The midline of the dorsal spine is obtained by performing non-uniform rational B-spline fitting on the continuous dorsal spine points.

2. The method according to claim 1, characterized in that The method of obtaining three-dimensional contour line data of a human back cross section and performing a polynomial fitting technique on the three-dimensional contour line data of the human back cross section to obtain a fitted contour line comprises: The three-dimensional contour line data of the human back section is collected by using an automated scoliosis screening system, wherein the three-dimensional contour line data of the human back section is hundreds of parallel and equidistant three-dimensional contour line data of the human back section; Projecting the three-dimensional contour line data of the human back cross section onto a two-dimensional plane to obtain a contour line of the two-dimensional plane; According to the contour line of the two-dimensional plane, the contour line of the two-dimensional plane is fitted using a polynomial fitting technique to obtain a fitted contour line.

3. The method according to claim 1, characterized in that The multi-scale constraint is performed on the back spine points to obtain continuous back spine points, including: Calculate the average value of two adjacent points among the back spine points; Extracting the point on the fitted contour line closest to the average value as the latest spine point, traversing the contour line multiple times until the back spine point is no longer updated, and obtaining a back spine point set P; According to the back spine point set P, the average value of the multi-scale extracted back spine points is obtained, and the contour line is traversed multiple times until the back spine points are no longer updated, and a continuous back spine point set P is obtained. ' .

4. The method according to claim 1, wherein The method of performing non-uniform rational B-spline fitting on the continuous back spine points to obtain the back spine midline includes: According to the back spine point P ' , the parameter value is obtained by using the accumulated chord length parameter method to obtain the node vector; Calculating a cubic B-spline basis function based on the node vector; Substituting the result calculated by the cubic B-spline basis function into the weighted control matrix, and solving the linear equations using the tangent vector boundary condition to obtain the control vertex; Fitting is performed based on the control vertices to obtain the dorsal spine midline.

Images