Shoulder joint glenoid bone defect measurement method and system in three-dimensional CT image

Abstract

The invention provides a method for measuring shoulder joint glenoid bone defects in a three-dimensional CT image, the method is divided into five modules, namely a segmentation module, a registration module, a positioning module, a defect size measurement module and a defect proportion measurement module, and the method can realize arbitrary defect measurement containing shoulder joint glenoid CT data under a relatively high resolution ratio. The method does not depend on manual operation, and automatic shoulder joint glenoid bone defect measurement is achieved. According to the method provided by the invention, the proportion of the glenoid bone defect of the shoulder joint and the size of the defect bone block can be quantitatively calculated in a standard, automatic and relatively accurate manner, and the preoperative diagnosis efficiency of a doctor is greatly improved. The invention further provides a system for measuring the glenoid bone defect of the shoulder joint in the three-dimensional CT image. According to the system, the indexes of the shoulder joint glenoid bone defect can be automatically obtained from end to end, that is, the input is shoulder joint CT data of the uninjured side and the affected side of a patient, and the output is the indexes of the shoulder joint glenoid bone defect.

Description

Technical Field

[0001] The invention relates to a medical image processing technology, belonging to the technical field of computer application. Background Art

[0002] Bone defects of the glenoid cavity are a common traumatic injury usually caused by shoulder dislocation. During the dislocation, the humeral head may collide with the glenoid cavity in the front and lower part, resulting in a defect in the glenoid cavity bone. These injuries may cause shoulder instability and seriously affect the patient's quality of life.

[0003] The degree of glenoid bone defect is an important reference indicator for determining whether bone surgery is needed, and is also an important basis for selecting the size of bone grafts during bone surgery. Therefore, accurate measurement of the degree of glenoid bone defect is of great significance.

[0004] The degree of glenoid bone defect is usually assessed based on the results of imaging examinations. Doctors use medical imaging techniques such as computed tomography (CT) to scan the patient's shoulder joint, and then further analyze these medical images through experience and manual measurement to diagnose the patient and make treatment recommendations.

[0005] The measurement of the degree of glenoid bone defect based on the results of imaging examinations is often inseparable from some medical image processing software. In clinical practice, the degree of glenoid bone defect is generally judged by doctors based on their own experience, or measured manually by doctors using image processing software, but such a measurement paradigm has some inevitable defects: it takes a lot of time and energy for doctors to master and use these software proficiently; any of the current mainstream methods rely on manual measurement, which will cause certain subjective errors; any of the current mainstream methods measure the bone defect on the front of the glenoid, but the plane where the glenoid is located is not a plane in the standard sense, but a relative plane. The current measurement method that regards the glenoid surface as a plane completely ignores the information in three-dimensional space, such as the thickness (depth) of the glenoid defect; any of the current mainstream methods only considers the degree of glenoid bone defect (defect ratio) at most, but does not consider the amount of glenoid bone defect (length, width, depth, etc. of the defect site). Summary of the invention

[0006] Technical problem to be solved by the present invention: In view of the subjective and cumbersome problems of the existing manual method for measuring glenoid bone defects, the present invention proposes a method and system for measuring glenoid bone defects in three-dimensional CT images.

[0007] Technical solution: To solve the above technical problems, the present invention adopts the following technical solution:

[0008] A method for measuring glenoid bone defects in three-dimensional CT images is provided, which decomposes the task of glenoid bone defect measurement into five modules: segmentation module, registration module, positioning module, defect size measurement module and defect ratio measurement module. The goal of the segmentation module is to obtain the patient's healthy and affected shoulder joint images; the goal of the registration module is to align the affected shoulder joint with the healthy shoulder joint; the goal of the positioning module is to locate the healthy shoulder glenoid surface area, align the shoulder joint template image with a marked boundary box with the healthy shoulder joint image, so that the glenoid surface area of ​​the healthy shoulder joint is framed by the registered boundary box; the goal of the defect size measurement module is to measure the size of the affected shoulder glenoid surface defect; the goal of the defect ratio measurement module is to measure the ratio of the affected shoulder glenoid surface defect, wherein the defect ratio includes the defect area ratio and the defect diameter ratio.

[0009] The specific steps include:

[0010] Step 1: Split the module;

[0011] Step 1-1: Input the original CT image, which is referred to as I1 (including the healthy side or the affected side or both sides).

[0012] Step 1-2: Input I1 into the pre-trained shoulder joint segmentation model, and the segmentation result is called L1;

[0013] Step 1-3: Threshold segmentation of I1 to obtain bone tissue, which is called L2;

[0014] Step 1-4: Multiply L1 and L2 to get L3;

[0015] Step 1-5: Randomly select one (or more) coordinates from the non-zero value coordinates in L3, and use this (or more) coordinates as seed points to perform region growing segmentation on L2 to obtain L4;

[0016] Step 1-6: Perform morphological closing operation on L4 to fill the gaps and obtain L5;

[0017] Step 1-7: Multiply L5 and I1 to obtain I2 (I2 is the final image of the healthy side, the affected side, or both sides of the shoulder joint).

[0018] Step 2: Registration module;

[0019] Step 2-1: Input the healthy and affected shoulder joint images obtained by the segmentation module; the healthy shoulder joint image is used as a fixed image, called I f ; The image of the affected shoulder joint is taken as a moving image, called I m ;

[0020] Step 2-2: I m Mirror and get Im2 ;

[0021] Step 2-3: I m2 To I f Registration. Choose to use the grayscale-based registration algorithm for iteration, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measure is mutual information, and the optimization algorithm is gradient descent;

[0022] Step 2-4: Iterative optimization, the intermediate image in the registration process is denoted as I m3 ;

[0023] Step 2-5: After the iteration, the output image is recorded as I r (I r It is the image obtained after the image of the affected shoulder joint is registered).

[0024] Step 3: Positioning module;

[0025] Step 3-1: Input the shoulder joint template image with bounding box labels and the healthy shoulder joint image; the shoulder joint template image with bounding box labels is used as the moving image, called I b , the healthy shoulder joint image is used as a fixed image, called I f , the bounding box label is called B b ;

[0026] Step 3-2: I b To I f Perform registration. Select the grayscale-based registration algorithm for iterative optimization, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measurement method is mutual information, and the optimization method is gradient descent;

[0027] Step 3-3: Iteration is completed and the global registration transformation parameter T is obtained g , the global registration output image is I g ;

[0028] Step 3-4: T g Application in B b Go up and get B g ;

[0029] Step 3-5: B g Amplify and get B s ;

[0030] Step 3-6: Pass B s in I f and I g Get the ROI area, denoted as R f and R g ;

[0031] Step 3-7: R g To R f Perform registration. Select the grayscale-based registration algorithm for iterative optimization, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measurement method is mutual information, and the optimization method is gradient descent;

[0032] Step 3-8: Iteration is completed and the local transformation parameter T is obtained l ;

[0033] Step 3-9: T l Application in B g Go up and get B l (B l That is, the bounding box for locating the glenoid surface area of ​​the healthy shoulder).

[0034] Step 4: Defect size measurement module;

[0035] Step 4-1: Input the healthy shoulder joint image and the image of the affected shoulder joint after registration obtained by the registration module; the healthy shoulder joint image is referred to as I f The image of the affected shoulder joint after registration is called I r ;

[0036] Step 4-2: I f and I r Perform binarization operation to obtain L f and L r ;

[0037] Step 4-3: L f With L r Perform difference operation to obtain L s ;

[0038] Step 4-4: L s Perform morphological corrosion operation to obtain L s2 ;

[0039] Step 4-5: L s2 Perform connected domain analysis to obtain the maximum connected domain, denoted as L s3 ;

[0040] Step 4-6: In L s3 Randomly select one (or more) coordinates from the non-zero value coordinates of L and use this (or more) coordinates as seed points in L s Perform region growing segmentation on it and get L s4 (L s4 is the target bone block of the defect);

[0041] Step 4-7: Get L s4 The non-zero value coordinate data is denoted as D;

[0042] Step 4-8: Perform PCA operation on D to obtain the direction vector P1 of the longest side (the first principal component) and two direction vectors P2 and P3 (the second principal component and the third principal component) perpendicular to the longest side;

[0043] Step 4-9: Place L s4 Project to P1 and obtain the projection line segment S in the direction of the longest side;

[0044] Step 4-10: Place L s4 Project along P1 to a plane orthogonal to P1 to obtain the minimum circumscribed rectangle R of the projection area (the cube formed by the projection line segment S and the minimum circumscribed rectangle R is approximately regarded as the three-dimensional length, width, and depth of the bone defect on the glenoid surface of the affected shoulder, and is also approximately regarded as the size of the bone defect on the glenoid surface of the affected shoulder).

[0045] Step 5: Defect ratio measurement module;

[0046] Step 5-1: Input the bounding box of the healthy shoulder glenoid surface area obtained by the positioning module, the healthy shoulder joint image, and the image of the affected shoulder joint after registration obtained by the registration module; the bounding box of the healthy shoulder glenoid surface area is called B l , the image of the healthy shoulder joint is called I f The image of the affected shoulder joint after registration is called I r ;

[0047] Step 5-2: Find B l The face with the largest area is denoted as S, and the normal vector of the face is obtained, denoted as N;

[0048] Step 5-3: Get I f with I r In B l The non-zero coordinates in f and C r ;

[0049] Step 5-4: C f Project along the normal vector N, and the area falling on the plane S is recorded as A;

[0050] Step 5-5: Get the coordinate data on A, recorded as C a ;

[0051] Step 5-6: In C a A circle is fitted on the image, and the obtained circle is denoted as O, with the center as o, the radius as r, and the circumference as N;

[0052] Step 5-7: Get C f With C r The points falling within the fitting circle O are denoted as Cf2 and C r2 ;

[0053] Step 5-8: Calculate C f2 With C r2 The area represented is denoted by s f and r ;

[0054] Step 5-9: Calculate s r / s f , and get the proportion p s (p s That is, the defect area ratio of the healthy glenoid surface to the affected glenoid surface);

[0055] Step 5-10: Calculate C r2 The arc on the fitting circle N is denoted as A;

[0056] Step 5-11: Get the two endpoints of arc A, obtain line segment H, and calculate the length of line segment H, recorded as l;

[0057] Step 5-12: Calculate the distance from the center o to the line segment H, denoted as d;

[0058] Step 5-13: Calculate (rd) / 2r to get the ratio p d (p d That is, the ratio of the defect diameter between the healthy glenoid surface and the affected glenoid surface).

[0059] The present invention also proposes a glenoid bone defect measurement system in three-dimensional CT images, which realizes the task of glenoid bone defect measurement end-to-end and automation, with the input being the original CT images of the left and right shoulders and the output being the glenoid surface defect index of the affected side.

[0060] Compared with the prior art, the advantages of the present invention are as follows: 1. The present invention can provide quantitative shoulder glenoid surface defect indicators, such as defect ratio (%) and defect size (length, width, depth, mm); 2. The present invention can automatically measure indicators (defect ratio and size) without manual interactive operations; 3. The present invention can automatically position the shoulder glenoid surface in three-dimensional space; 4. The present invention provides a visualization effect in three-dimensional space to facilitate the display of defect ratio and size. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the entire process of the present invention;

[0062] Figure 2 It is a flow chart of the segmentation module;

[0063] Figure 3 Schematic diagram of segmentation results;

[0064] Figure 4 This is a schematic diagram of the registration module process;

[0065] Figure 5 This is a schematic diagram of the registration results;

[0066] Figure 6 This is a flow chart of the positioning module;

[0067] Figure 7 This is a schematic diagram of the positioning results;

[0068] Figure 8 This is a schematic diagram of the defect size measurement module;

[0069] Fig. 9 Schematic diagram of the defective bone block;

[0070] Fig.10 This is a schematic diagram of a defect ratio measurement module;

[0071] Fig.11 This is a schematic diagram of the glenoid surface circle fitting results. DETAILED DESCRIPTION

[0072] The specific implementation modes of the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0073] Example 1: See Figure 1-Figure 11 The present invention provides a method for measuring glenoid bone defect in three-dimensional CT images. Based on the original CT images of the left and right shoulders, the method uses five modules, namely, a segmentation module, a registration module, a positioning module, a defect size measurement module and a defect ratio measurement module, to quantitatively calculate the glenoid bone defect information on the affected side. The method specifically includes the following steps:

[0074] Step 1, segmentation module: segment the bilateral shoulder joint regions from the original CT image to obtain an image containing only the bilateral shoulder joints;

[0075] Step 2, registration module: register the bilateral shoulder joint images to obtain an image after the affected shoulder joint is registered with the healthy shoulder joint;

[0076] Step 3, positioning module: positioning the glenoid surface area of ​​the shoulder joint and obtaining a bounding box that frames the glenoid surface of the healthy side;

[0077] Step 4, defect size measurement module: measure the size of the glenoid defect to obtain the size of the bone defect on the affected glenoid surface;

[0078] Step 5, defect ratio measurement module: measure the ratio of the defect on the glenoid surface of the shoulder joint, and obtain the defect area ratio and diameter ratio of the glenoid surface of the affected side.

[0079] Among them, the shoulder joint segmentation module in step 1 obtains a more refined shoulder joint image after performing threshold segmentation, region growing segmentation, morphological operations and other operations on the basis of the segmentation results of the pre-trained shoulder joint segmentation model.

[0080] Among them, the shoulder joint registration module in step 2 first mirrors the affected shoulder joint image, and then spatially aligns the mirrored affected shoulder joint image with the healthy shoulder joint image.

[0081] The glenoid surface positioning module in step 3 uses the shoulder joint image with the marked bounding box as the template image, and transforms the bounding box on the template image to the healthy shoulder joint image through a two-stage global and local registration method;

[0082] Shoulder glenoid surface positioning module, the annotated bounding box is a three-dimensional cube, which is determined by the vertices of eight three-dimensional coordinates and is spatially tilted in the axial, coronal and sagittal planes.

[0083] Among them, the shoulder glenoid surface defect size measurement module in step 4 uses the difference between the healthy shoulder joint image and the affected shoulder joint image obtained after alignment, and extracts the defective bone block in the affected shoulder glenoid surface area through morphological operations, connected domain analysis and other operations; determines the point set of the defective bone block, and applies principal component analysis (PCA) to obtain the first three principal components of the point set; projects the point set to the first principal component to obtain the longest side of the defective bone block; projects along the first principal component onto the plane composed of the second and third principal components, and obtains the minimum circumscribed rectangle of the projection area; the longest side and the minimum circumscribed rectangle constitute the circumscribed cube of the defective bone block in the affected shoulder glenoid area, and the length, width and height of the cube represent the length, width and depth of the defective bone block.

[0084] Among them, the shoulder glenoid surface defect ratio measurement module in step 5 uses a bounding box that frames the healthy glenoid surface to determine the point set of the healthy shoulder joint and the affected shoulder joint within the bounding box; determines the surface with the largest area of ​​the bounding box as the projection surface, and the normal vector of the surface with the largest area of ​​the bounding box is the projection direction; projects the point set of the healthy shoulder joint falling within the bounding box onto the projection surface; fits a circle on the point set obtained by the projection; determines the point set of the healthy shoulder joint and the affected shoulder joint falling within the fitting circle, calculates the area enclosed by these points on the projection plane, and the ratio of the area of ​​the affected shoulder joint to the healthy shoulder joint is the defect area ratio of the affected shoulder glenoid surface; determines the intersection of the point of the affected shoulder joint falling within the fitting circle and the circumference of the fitting circle to obtain an arc, calculates the distance from the origin to the line segment between the two end points of the arc, and the ratio of the difference between the radius of the fitting circle and the distance to the diameter of the fitting circle is the defect diameter ratio of the affected shoulder glenoid surface.

[0085] In the shoulder glenoid surface defect ratio measurement module, the point set of the shoulder joint within the bounding box is a set of three-dimensional coordinate points with non-zero voxel values; the point set obtained by projecting the point set of the shoulder joint within the bounding box is still a set of three-dimensional coordinate points; the center of the fitted circle is a three-dimensional coordinate point; the intersection of the points of the affected shoulder joint falling within the fitted circle and the circumference of the fitted circle is a set of three-dimensional coordinate points.

[0086] Shoulder glenoid surface defect ratio measurement module, circle fitting is roughly described as finding a circle in a given point set so that the circle covers as many points as possible and has the smallest area possible. This problem is converted into an optimization problem, and the fitting problem is solved by using the idea of ​​solving optimization problems. An optimization objective function is defined, and the variable parameters of the objective function are the plane coordinate information about the circle: (x, y, r), where x and y represent the coordinates of the center of the circle, and r represents the radius of the circle; the value range of the objective function parameters is determined according to the point set; a two-stage optimization method is adopted, and a genetic algorithm is used to perform a global search within the value range to obtain a relatively optimal solution; the Nelder-Mead optimization algorithm is used to continue the local iterative search with the relatively optimal solution as the initial solution to obtain the optimal solution.

[0087] Example 2: A shoulder glenoid bone defect measurement system in a three-dimensional CT image is based on the original CT images of the left and right shoulders, and integrates five modules of segmentation, alignment, positioning, defect size measurement and defect ratio measurement to realize end-to-end and automatic calculation of the affected shoulder glenoid bone defect information. That is, by inputting the original CT images of the left and right shoulders, the defect index of the affected shoulder glenoid surface can be output without any manual processing.

[0088] Embodiment 3: A method for measuring a glenoid bone defect in a three-dimensional CT image, the method comprising the following steps:

[0089] Step 1: Split the module, such as Figure 2 Input: original CT images of the patient's bilateral shoulders. If both shoulders are in one image, one image is input; if both shoulders are in two images respectively, two images are input. Take inputting two images as an example; Output: images of the patient's healthy shoulder joint and affected shoulder joint.

[0090] Step 1.1: Input the original CT images of the healthy and affected shoulders of the patient, referred to as I h and I u ;

[0091] Step 1.2: I h Input the pre-trained shoulder joint segmentation model, and the segmentation result is called L1;

[0092] Step 1.3: I h Perform threshold segmentation to extract bone tissue, which is called L2;

[0093] Step 1.4: Multiply L1 and L2 to get L3;

[0094] Step 1.5: Randomly select one (or more) coordinates from the non-zero value coordinates in L3, and use this (or more) coordinates as seed points to perform region growing segmentation on L2 to obtain L4;

[0095] Step 1.6: Perform morphological closing operation on L4 to fill the gaps and obtain L5;

[0096] Step 1.7: Connect L5 and I h Multiply them and get I h2 (I h2 is the final image of the healthy shoulder joint), such as Figure 3 As shown;

[0097] Step 1.8: I u Repeat the above steps to get I u2 (I u2 This is the final image of the affected shoulder joint), such as Figure 3 shown.

[0098] Step 2: Registration module, such as Figure 4 Input: bilateral shoulder joint images of the patient; Output: the image of the affected shoulder joint after registration with the image of the healthy shoulder joint.

[0099] Step 2.1: Input the healthy and affected shoulder joint images obtained by the segmentation module; the healthy shoulder joint image is used as a fixed image, called I f ; The image of the affected shoulder joint is taken as a moving image, called I m ;

[0100] Step 2.2: I m Mirror and get I m2 ;

[0101] Step 2.3: I m2 To I f Registration. Choose to use the grayscale-based registration algorithm for iteration, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measure is mutual information, and the optimization algorithm is gradient descent;

[0102] Step 2.4: Iterative optimization, the intermediate image in the registration process is denoted as I m3 ;

[0103] Step 2.5: After the iteration, the output image is recorded as I r (I r is the image obtained after the image of the affected shoulder joint is registered), such as Figure 5 shown.

[0104] Step 3: Locate the module, such as Figure 6 Input: shoulder joint template image with bounding box labels and healthy shoulder joint image; Output: healthy shoulder joint image with bounding box labels.

[0105] Step 3.1: Input the shoulder joint template image with bounding box labels and the healthy shoulder joint image; the shoulder joint image with bounding box labels is used as the moving image, called I b , the healthy shoulder joint image is used as a fixed image, called I f , the bounding box label is called B b ;

[0106] Step 3.2: I b To I f Perform registration. Select the grayscale-based registration algorithm for iterative optimization, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measurement method is mutual information, and the optimization method is gradient descent;

[0107] Step 3.3: After iteration, the global registration transformation parameter T is obtained g , the global registration output image is I g ;

[0108] Step 3.4: T g Application in B b Go up and get B g ;

[0109] Step 3.5: B g Amplify and get B s ;

[0110] Step 3.6: Pass B s in I f and I g Get the ROI area, denoted as R f and R g ;

[0111] Step 3.7: R g To R f Perform registration. Select the grayscale-based registration algorithm for iterative optimization, the transformation method is rigid body transformation, the image interpolation method is linear interpolation, the similarity measurement method is mutual information, and the optimization method is gradient descent;

[0112] Step 3.8: Iteration is completed and the local transformation parameter T is obtained l , the global registration output image is I l ;

[0113] Step 3.9: T l Application in B gGo up and get B l (B l is the bounding box for locating the contralateral glenoid surface area, such as Figure 7 shown.

[0114] Step 4: Defect size measurement module, such as Figure 8 Input: healthy shoulder joint image and the image of the affected shoulder joint after registration obtained by the registration module; Output: the image of the bone defect on the glenoid surface of the affected shoulder and the size information of the bone defect on the glenoid surface of the affected shoulder.

[0115] Step 4.1: Input the healthy shoulder joint image and the image of the affected shoulder joint after registration obtained by the registration module; the healthy shoulder joint image is referred to as I f The image of the affected shoulder joint after registration is called I r ;

[0116] Step 4.2: I f and I r Perform binarization operation to obtain L f and L r ;

[0117] Step 4.3: L f With L r Perform difference operation to obtain L s ;

[0118] Step 4.4: L s Perform morphological corrosion operation to obtain L s2 ;

[0119] Step 4.5: L s2 Perform connected domain analysis to obtain the maximum connected domain, denoted as L s3 ;

[0120] Step 4.6: In L s3 Randomly select one (or more) coordinates from the non-zero value coordinates of L and use this (or more) coordinates as seed points in L s Perform region growing segmentation on it and get L s4 (L s4 is the target bone defect), such as Fig. 9 As shown;

[0121] Step 4.7: Get L s4 The non-zero value coordinate data is denoted as D;

[0122] Step 4.8: Perform PCA operation on D to obtain the direction vector P1 of the longest side (the first principal component) and two direction vectors P2 and P3 (the second principal component and the third principal component) perpendicular to the longest side;

[0123] Step 4.9: Ls4 Project to P1 and obtain the projection line segment S in the direction of the longest side;

[0124] Step 4.10: L s4 Project along P1 to a plane orthogonal to P1 to obtain the minimum circumscribed rectangle R of the projection area (the cube formed by the projection line segment S and the minimum circumscribed rectangle R is approximately regarded as the three-dimensional length, width, and depth of the bone defect on the glenoid surface of the affected shoulder, and is also approximately regarded as the size of the bone defect on the glenoid surface of the affected shoulder).

[0125] Step 5: Defect ratio measurement module, such as Fig.10 Input: the bounding box on the glenoid surface of the healthy shoulder, the image of the healthy shoulder joint and the image of the affected shoulder joint after registration obtained by the registration module; Output: the defect area ratio and defect diameter ratio of the glenoid surface of the affected shoulder.

[0126] Step 5.1: Input the bounding box of the healthy shoulder glenoid surface area obtained by the positioning module, the healthy shoulder joint image, and the image of the affected shoulder joint after registration obtained by the registration module; the bounding box of the healthy shoulder glenoid surface area is called B l , the image of the healthy shoulder joint is called I f The image of the affected shoulder joint after registration is called I r ;

[0127] Step 5.2: Find B l The face with the largest area is denoted as S, and the normal vector of the face is obtained, denoted as N;

[0128] Step 5.3: Get I f with I r In B l The non-zero coordinates in f and C r ;

[0129] Step 5.4: C f Project along the normal vector N, and the area falling on the plane S is recorded as A;

[0130] Step 5.5: Get the coordinate data on A, recorded as C a ;

[0131] Step 5.6: In C a The circle is fitted on the surface, and the obtained circle is denoted as O, with the center as o, the radius as r, and the circumference as N. Fig.11 As shown;

[0132] Step 5.7: Get C f With C r The points falling within the fitting circle O are denoted as C f2 and Cr2 ;

[0133] Step 5.8: Calculate C f2 With C r2 The area represented is denoted by s f and r ;

[0134] Step 5.9: Calculate s r / s f , and get the proportion p s (p s That is, the defect area ratio of the healthy glenoid surface to the affected glenoid surface);

[0135] Step 5.10: Calculate C r2 The arc on the fitting circle N is denoted as A;

[0136] Step 5.11: Get the two endpoints of arc A, obtain line segment H, and calculate the length of line segment H, denoted as l;

[0137] Step 5.12: Calculate the distance from the center of the circle o to the line segment H, denoted as d;

[0138] Step 5.13: Calculate (rd) / 2r to get the ratio p d (p d That is, the ratio of the defect diameter between the healthy glenoid surface and the affected glenoid surface).

[0139] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. The above embodiments are only preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Within the knowledge of ordinary technicians in this field, several improvements or equivalent substitutions can be made without departing from the principle of the present invention, and these improvements and substitutions all belong to the protection scope of the present invention.

Claims

1. A method for measuring glenoid bone defect in a three-dimensional CT image, characterized in that: Based on the original CT images of the left and right shoulders, a method for quantitatively calculating the glenoid bone defect information of the affected shoulder is implemented by using five modules, namely, a segmentation module, a registration module, a positioning module, a defect size measurement module and a defect ratio measurement module. Specifically, the method comprises the following steps: Step 1, segmentation module: segment the bilateral shoulder joint regions from the original CT image to obtain an image containing only the bilateral shoulder joints; Step 2, registration module: register the bilateral shoulder joint images to obtain an image after the affected shoulder joint is registered with the healthy shoulder joint; Step 3, positioning module: positioning the glenoid surface area of ​​the shoulder joint and obtaining a bounding box that frames the glenoid surface of the healthy side; Step 4, defect size measurement module: measure the size of the glenoid defect to obtain the size of the bone defect on the affected glenoid surface; Step 5, defect ratio measurement module: measure the ratio of the defect on the glenoid surface of the shoulder joint, and obtain the defect area ratio and diameter ratio of the glenoid surface of the affected side.

2. The method for measuring glenoid bone defect in a three-dimensional CT image according to claim 1, characterized in that: The shoulder joint segmentation module in step 1 performs threshold segmentation, region growing segmentation, and morphological operations on the segmentation results of the pre-trained shoulder joint segmentation model to obtain a more refined shoulder joint image.

3. The method for measuring glenoid bone defect in a three-dimensional CT image according to claim 1, characterized in that: The shoulder joint registration module in step 2 first mirrors the affected shoulder joint image, and then spatially aligns the mirrored affected shoulder joint image with the healthy shoulder joint image.

4. The method for measuring glenoid bone defect in a three-dimensional CT image according to claim 1, characterized in that: The glenoid surface positioning module in step 3 uses the shoulder joint image with a marked bounding box as a template image, and transforms the bounding box on the template image to the healthy shoulder joint image through a global and local two-stage registration method.

5. The method for measuring glenoid bone defect in three-dimensional CT images according to claim 1, characterized in that: The glenoid surface positioning module in step 3 uses a three-dimensional cube as the bounding box, which is determined by eight vertices of three-dimensional coordinates and is spatially tilted in the axial, coronal and sagittal planes.

6. The method for measuring glenoid bone defect in a three-dimensional CT image according to claim 1, characterized in that: The shoulder glenoid surface defect size measurement module in step 4 uses the healthy shoulder joint image and the registered affected shoulder joint image to perform a difference, and extracts the defective bone block in the affected shoulder glenoid surface area through morphological operations, connected domain analysis and other operations; determines the point set of the defective bone block, and applies principal component analysis (PCA) to obtain the first three principal components of the point set; projects the point set to the first principal component to obtain the longest side of the defective bone block; projects along the first principal component onto the plane composed of the second and third principal components, and obtains the minimum circumscribed rectangle of the projection area; the longest side and the minimum circumscribed rectangle form a circumscribed cube of the defective bone block in the affected shoulder glenoid area, and the length, width and height of the cube represent the length, width and depth of the defective bone block.

7. The method for measuring glenoid bone defect in three-dimensional CT images according to claim 1, characterized in that: The shoulder glenoid defect ratio measurement module in step 5 uses a bounding box that frames the healthy glenoid surface to determine the point set of the healthy shoulder joint and the affected shoulder joint within the bounding box; determines the surface with the largest area of ​​the bounding box as the projection surface, and the normal vector of the surface with the largest area of ​​the bounding box is the projection direction; projects the point set of the healthy shoulder joint falling within the bounding box onto the projection surface; performs circle fitting on the point set obtained by the projection; determines the point set of the healthy shoulder joint and the affected shoulder joint falling within the fitting circle, calculates the area enclosed by these points on the projection plane, and the ratio of the area of ​​the affected shoulder joint to the healthy shoulder joint is the defect area ratio of the affected shoulder glenoid surface; determines the intersection of the point of the affected shoulder joint falling within the fitting circle and the circumference of the fitting circle to obtain an arc, calculates the distance from the origin to the line segment between the two end points of the arc, and the ratio of the difference between the radius of the fitting circle and the distance to the diameter of the fitting circle is the defect diameter ratio of the affected shoulder glenoid surface.

8. The method for measuring glenoid bone defect in three-dimensional CT images according to claim 1, characterized in that: In the shoulder glenoid surface defect proportion measurement module in step 5, the point set of the shoulder joint in the bounding box is a set of three-dimensional coordinate points with non-zero voxel values; the point set obtained by projecting the point set of the shoulder joint in the bounding box is still a set of three-dimensional coordinate points; the center of the fitted circle is a three-dimensional coordinate point; the intersection of the points of the affected shoulder joint falling in the fitting circle and the circumference of the fitting circle is a set of three-dimensional coordinate points.

9. The method for measuring glenoid bone defect in three-dimensional CT images according to claim 1, characterized in that: In the shoulder glenoid surface defect proportion measurement module in step 5, the circle fitting is roughly described as finding a circle in a given point set so that the circle covers as many points as possible and has as small an area as possible. The problem is converted into an optimization problem, and the fitting problem is solved by using the idea of ​​solving optimization problems. An optimization objective function is defined, and the variable parameters of the objective function are the plane coordinate information about the circle: (x, y, r), where x and y represent the coordinates of the center of the circle, and r represents the radius of the circle; the value range of the objective function parameters is determined according to the point set; a two-stage optimization method is adopted, and a genetic algorithm is used to perform a global search in the value range to obtain a relatively optimal solution; the Nelder-Mead optimization algorithm is used to continue the local iterative search with the relatively optimal solution as the initial solution to obtain the optimal solution.

10. A system for measuring glenoid bone defects in three-dimensional CT images, used to implement a method for measuring glenoid bone defects in three-dimensional CT images as claimed in any one of claims 1 to 9, characterized in that Based on the original CT images of the left and right shoulders, the five modules of segmentation, registration, positioning, defect size measurement and defect ratio measurement are integrated to realize end-to-end and automatic calculation of the glenoid bone defect information of the affected shoulder. That is, by inputting the original CT images of the left and right shoulders, the defect indicators of the glenoid surface of the affected shoulder can be output without any manual processing.

Images