A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm
By screening and filtering the spine edge in the diagnosis of abdominal aortic aneurysm, the interference distortion coefficient and degree of influence are calculated, the spine deformation detection error under the influence of image quality and noise is solved, and detection accuracy and data clarity are improved.
Patent Information
- Application Number
- CN202411204478.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In machine vision-assisted abdominal aortic aneurysm diagnosis, factors such as image quality and noise lead to spinal deformation detection errors, affecting detection accuracy.
By obtaining the spinal grayscale images of patients and healthy humans, dividing the abdominal aorta and spinal regions, screening the spinal edges, calculating the interference distortion coefficient and the degree of influence, performing image filtering and edge detection, matching the spinal edge images to identify deformations.
Improves the accuracy and morphological evaluation of the spine edge, provides clear spinal image data, supports spinal deformation recognition, and reduces detection errors.
Smart Images

Figure CN118735912B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to a spinal deformation recognition method for auxiliary diagnosis of abdominal aortic aneurysm. Background Art
[0002] In the auxiliary diagnosis of abdominal aortic aneurysm, spinal deformity can be used as one of the important identification features; when the abdominal aortic aneurysm grows and expands laterally, it may put pressure on the spine and cause scoliosis, which may cause abnormal spinal curvature and symptoms such as shoulder asymmetry and pelvic inequality; therefore, the patient's spinal deformity can be analyzed for auxiliary diagnosis of abdominal aortic aneurysm.
[0003] In machine vision-assisted diagnosis of abdominal aortic aneurysm, accurate detection of patients' spinal deformities is crucial. However, factors such as image quality and noise may pose challenges to the detection algorithm. When body tissue causes image blur, artifacts or partial occlusion, machine vision systems often have difficulty accurately identifying spinal deformities, which may lead to detection errors and affect the accuracy of spinal deformation identification. Summary of the invention
[0004] The present invention provides a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm to solve the existing problems.
[0005] The present invention provides a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm using the following technical solution:
[0006] An embodiment of the present invention provides a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm, the method comprising the following steps:
[0007] Acquire a grayscale image of the patient's spine including the patient's abdominal aorta and an image of the edge of a healthy human spine;
[0008] The abdominal aorta region and the spinal column region are segmented from the patient's spinal column grayscale image;
[0009] According to the distribution of the edges in the spinal region and the distance from each edge to the abdominal aorta region, a number of spinal edges are selected from all the edges; multiple rectangles are fitted to all the spinal edges to obtain a number of vertebral rectangles, the fitting error of each vertebral rectangle and the fitting error of each spinal edge;
[0010] The interference distortion coefficient of the spine is obtained according to the distance between the pixel points on the boundary of the abdominal aorta area, the number of vertebral rectangles and the fitting error of each vertebral rectangle; the degree of influence of the spine is obtained according to the distance between the edges of the spine, the fitting error of each spine edge and the gradient amplitude of all pixel points on the edges of the spine;
[0011] According to the interference distortion coefficient and the degree of influence of the spine, the filtering weight of the grayscale image of the patient's spine is obtained; according to the filtering weight, the grayscale image of the patient's spine is filtered to obtain a filtered image; the edge detection of the filtered image is performed to obtain the edge image of the patient's spine, and the edge image of the patient's spine is matched with the edge image of the healthy human spine to obtain the patient's spine deformation detection result.
[0012] Furthermore, the step of dividing the abdominal aorta region and the spinal column region from the patient's spinal column grayscale image includes the following specific steps:
[0013] Obtaining a segmentation threshold of the patient's spine grayscale image, and recording the region formed by all pixels with grayscale values greater than the segmentation threshold in the patient's spine grayscale image as the abdominal aorta region;
[0014] According to the preset first cluster number, a clustering operation is performed on the area except the abdominal aorta area in the patient's spine grayscale image to obtain a plurality of cluster areas;
[0015] Use the Canny edge detection algorithm to perform operations on each cluster area to obtain several edge lines;
[0016] Use edge tracking to obtain the bifurcation points on all edge lines, and use the bifurcation points as segmentation points to segment each edge line into several edges;
[0017] Perform straight line fitting on all edges respectively to obtain the fitting error of each edge;
[0018] The mean of the fitting errors of all edges in each cluster region is calculated, and the cluster region corresponding to the smallest mean of the fitting errors of all edges is recorded as the spine region.
[0019] Furthermore, the method of selecting a plurality of spinal edges from all the edges according to the distribution of the edges in the spinal region and the distance between each edge and the abdominal aorta region includes the following specific steps:
[0020] Perform straight line fitting on all edges respectively to obtain the fitting straight line corresponding to each edge;
[0021] Obtain the distances between all the connecting lines of any two pixels in the abdominal aorta region, and record the connecting line of the two pixels corresponding to the maximum value of the connecting line distances as the extended straight line segment of the abdominal aorta region;
[0022] The minimum angle formed by the fitting straight line of each edge of the spine region and the straight line extending through the abdominal aorta region is recorded as the direction coefficient of each edge of the spine region;
[0023] According to the direction coefficients of all the edges of the spine region and the shortest distances between all the edges of the spine region and the extended straight line segments of the abdominal aorta region, the collinearity coefficients of any two edges of the spine region are obtained;
[0024] Among all the edges in the spine area, all the edges with direction coefficients less than the preset direction coefficient threshold are recorded as suspected edges on both sides of the spine;
[0025] According to the collinearity coefficients of all the suspected edges on both sides of the spine, several spine edges are screened out from all the suspected edges on both sides of the spine.
[0026] Furthermore, the specific calculation formula for obtaining the collinearity coefficient of any two edges of the spine region based on the direction coefficients of all edges of the spine region and the shortest distance between all edges of the spine region and the extended straight line segments of the abdominal aorta region is:
[0027]
[0028] In the formula, The first Article and Collinearity coefficient of strip edges; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first Direction coefficient of the strip edge; The first Direction coefficient of the strip edge; is the absolute value function.
[0029] Furthermore, the method of selecting a plurality of spine edges from all the suspected spine edges according to the collinearity coefficients of all the suspected spine edges comprises the following specific steps:
[0030] In the spine region, the inverse of the collinearity coefficient of any two suspected spine edges is taken as the clustering distance, and according to the preset number of second clusters, all the suspected spine edges are clustered to obtain a number of clusters;
[0031] The mean of the collinearity coefficients of all two suspected spine edges in each cluster is recorded as the spine edge possibility of each cluster;
[0032] The first two clusters with the highest spine edge probability are recorded as the left spine edge cluster and the right spine edge cluster respectively;
[0033] In the spine region, convex hull detection is performed on the edges on both sides of the suspected spine in the left spine edge cluster and the right spine edge cluster to obtain a convex hull region;
[0034] The edges on both sides of the suspected spine that are not in the left spine edge cluster and the right spine edge cluster within the convex hull region are recorded as suspected vertebral edges;
[0035] Multiple rectangles are fitted to all edges within the convex hull region to obtain the fitting error of each suspected vertebral edge;
[0036] According to the direction coefficient and fitting error of each suspected vertebral edge, the vertebral edge possibility of each suspected vertebral edge is obtained;
[0037] The suspected vertebral edge whose vertebral edge possibility is greater than a preset vertebral edge possibility threshold is recorded as a vertebral edge;
[0038] The suspected two side edges of the spine and the vertebral edges in the left spine edge cluster and the right spine edge cluster are recorded as spine edges.
[0039] Furthermore, the specific calculation formula for obtaining the vertebral edge possibility of each suspected vertebral edge according to the direction coefficient and fitting error of each suspected vertebral edge is:
[0040]
[0041] In the formula, Indicates Possibility of vertebral margins of suspected vertebral margins; Indicates Direction coefficient of the suspected vertebral edge; is the preset angle threshold; is the absolute value function; Indicates The fitting error of the suspected vertebral edge; is an exponential function with a natural constant as its base.
[0042] Furthermore, the interference distortion coefficient of the spine is obtained according to the distance between the pixel points on the boundary of the abdominal aorta region, the number of vertebral rectangles and the fitting error of each vertebral rectangle, and the specific steps include the following:
[0043] The shortest distance from each vertebral rectangle to the abdominal aorta region corresponds to the pixel point on the border of the abdominal aorta region, which is recorded as the main pixel point. On the border of the abdominal aorta region, starting from the main pixel point, the normal line of the main pixel point is traversed to the inside of the abdominal aorta region pixel by pixel point, and the other pixel points on the first traversed border of the abdominal aorta region are recorded as sub-pixel points. The distance from the main pixel point to the sub-pixel point is recorded as the width of the abdominal aorta closest to each vertebral rectangle.
[0044] The interference distortion coefficient of the spine is obtained according to the distance of each vertebral rectangle to the nearest abdominal aorta width, the number of vertebral rectangles and the fitting error of each vertebral rectangle.
[0045] Furthermore, the specific calculation formula for obtaining the interference distortion coefficient of the spine according to the width of the abdominal aorta closest to each vertebral rectangle, the number of vertebral rectangles, and the fitting error of each vertebral rectangle is:
[0046]
[0047] In the formula, represents the interference distortion coefficient of the spine; Represents the number of vertebral rectangles; Indicates The fitting error of the vertebral rectangle; Indicates The width of the abdominal aorta closest to the vertebral rectangle; is the absolute value function; is a hyperbolic function.
[0048] Furthermore, the specific calculation formula for the degree of influence of the spine is obtained according to the distance between the spine edges, the fitting error of each spine edge and the gradient amplitude of all pixel points on the spine edge:
[0049]
[0050] In the formula, Indicates the degree to which the spine is affected; represents the number of spine edges; Indicates The minimum value of the closest distance between the edge of a spine and all other spine edges; Indicates The fitting error of the spine edge; Indicates The variance of the gradient amplitude of all pixels on the edge of the spine; is a hyperbolic function.
[0051] Furthermore, the filtering weight of the patient's spine grayscale image is obtained according to the interference distortion coefficient and the affected degree of the spine, and the specific steps include the following:
[0052] The product of the interference distortion coefficient of the spine and the affected degree of the spine is recorded as the filtering weight of the patient's spine grayscale image.
[0053] The beneficial effects of the technical solution of the present invention are:
[0054] In an embodiment of the present invention, a grayscale image of the patient's spine including the patient's abdominal aorta and an image of the edge of the healthy human spine are obtained; the abdominal aorta region and the spinal region are divided from the grayscale image of the patient's spine to facilitate subsequent analysis; based on the distribution of edges in the spinal region and the distance from each edge to the abdominal aorta region, a number of spinal edges are screened from all edges, and the spinal edges are accurately obtained to improve the accuracy of the subsequent vertebral rectangles; multiple rectangles are fitted to all spinal edges to obtain a number of vertebral rectangles, the fitting error of each vertebral rectangle and the fitting error of each spinal edge, and the shape and position information of the vertebrae are obtained; based on the distance between pixel points on the boundary of the abdominal aorta region, The number of vertebral rectangles and the fitting error of each vertebral rectangle are used to obtain the interference distortion coefficient of the spine; the degree of influence of the spine is obtained according to the distance between the edges of the spine, the fitting error of each edge of the spine, and the gradient amplitude of all pixels on the edge of the spine, and the change in the morphology of the spine and the degree of influence of external interference are evaluated; the filtering weight of the grayscale image of the patient's spine is obtained according to the interference distortion coefficient and the degree of influence of the spine; the grayscale image of the patient's spine is filtered according to the filtering weight to obtain a filtered image; edge detection is performed on the filtered image to obtain the edge image of the patient's spine, and the edge image of the patient's spine is matched with the edge image of the healthy human spine to obtain the patient's spine deformation detection result. So far, the present invention obtains the abdominal aorta area through the abdominal aorta characteristics, screens the spine edge according to the distribution characteristics between the spine and the abdominal aorta, and finally obtains a clear spine image by filtering according to the degree to which the edge of the spine is affected, which provides good data support for subsequent spinal deformation identification. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0056] Figure 1 The present invention is a flowchart of the steps of a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm.
[0057] Figure 2 An example patient spine grayscale image provided for an embodiment of the present invention. DETAILED DESCRIPTION
[0058] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0059] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0060] A specific scheme of a spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm provided by the present invention is described in detail below with reference to the accompanying drawings.
[0061] See also Figure 1 , which shows a flowchart of a method for identifying spinal deformity for auxiliary diagnosis of abdominal aortic aneurysm provided by an embodiment of the present invention, the method comprising the following steps:
[0062] Step S001: Acquire a grayscale image of the patient's spine including the patient's abdominal aorta and an image of the edge of a healthy human spine.
[0063] Using a CT scan to acquire an image of the patient's spine including the patient's abdominal aorta;
[0064] The patient's spine image is grayed out to obtain a grayscale image of the patient's spine. The example image is as follows: Figure 2 As shown;
[0065] The healthy human spine image is collected by CT scanning, and the edge detection is performed to obtain the healthy human spine edge image.
[0066] Step S002: Segment the abdominal aorta region and the spinal column region from the patient's spinal column grayscale image.
[0067] For the grayscale images of the patient's spine, there may be interference factors such as blurred noise caused by patient movement, etc., and the human body tissue is relatively complex, which will affect the observation of the spine. In order to subsequently analyze the deformation effect between the abdominal aortic aneurysm and the spine, it is necessary to first obtain the abdominal aorta area. During CT scanning, contrast agents are used to enhance the development of blood vessels, thereby making the outline and internal structure of the abdominal aortic aneurysm more clearly displayed, so the abdominal aorta is displayed as a highlighted strip area in the image.
[0068] The segmentation threshold of the patient's spine grayscale image is obtained using a threshold segmentation algorithm. In the patient's spine grayscale image, the area composed of all pixels with grayscale values greater than the segmentation threshold is recorded as the abdominal aorta area, wherein the threshold segmentation algorithm is a well-known technology.
[0069] In order to screen out the edge of the spinal region, feature screening is required based on the spinal features. The spine is composed of several vertebrae, and the vertebrae are approximately cylindrical. In the patient's spinal grayscale image, the spine is located on the same side of the abdominal aorta.
[0070] It should be noted that if there are multiple maximum values, any one is selected as an example. The abdominal aorta region in the patient's spine grayscale image vertically runs through the image, dividing the image into left and right parts. Therefore, the preset number of first clusters in this embodiment is 2.
[0071] According to the preset number of first clusters, a K-means clustering algorithm is used to perform a clustering operation on the area except the abdominal aorta area in the patient's spine grayscale image to obtain a plurality of cluster areas.
[0072] Using the Canny edge detection algorithm to calculate each cluster area respectively to obtain a number of edge lines, wherein the Canny edge detection is a well-known technology;
[0073] It should be noted that the several edge lines include the edge of the spine and the edge of body tissue.
[0074] Edge tracking is used to obtain the bifurcation points on all edge lines, and each edge line is divided into several edges using the bifurcation points as segmentation points.
[0075] Using the least square method, straight line fitting is performed on all edges respectively to obtain a fitting straight line corresponding to each edge and a fitting error of each edge, wherein straight line fitting is a well-known technology;
[0076] The mean of the fitting errors of all edges in each cluster region is calculated, and the cluster region corresponding to the smallest mean of the fitting errors of all edges is recorded as the spine region.
[0077] Step S003: Filter out a number of spinal edges from all the edges based on the distribution of edges in the spinal region and the distance from each edge to the abdominal aorta region; fit multiple rectangles on all spinal edges to obtain a number of vertebral rectangles, the fitting error of each vertebral rectangle, and the fitting error of each spinal edge.
[0078] Since the spinal column region is composed of several vertebrae, and the vertebrae are approximately cylindrical, and the size difference between adjacent vertebrae is not large, and the spine is approximately located on a straight line as a whole when viewed from the side, but there is some physiological curvature; therefore, for the fitted straight lines of each edge, since the abdominal aorta extends along the spinal column, the fitted straight line of the spinal column edge is parallel to the extended straight line segment of the abdominal aorta region, and the distance between the fitted straight line of the spinal column edge and the extended straight line segment of the abdominal aorta region is close.
[0079] Obtain the distances between all the connecting lines of any two pixels in the abdominal aorta region, and record the connecting line of the two pixels corresponding to the maximum value of the connecting line distances as the extended straight line segment of the abdominal aorta region;
[0080] The minimum angle formed by the fitting straight line of each edge of the spine region and the straight line extending through the abdominal aorta region is recorded as the direction coefficient of each edge of the spine region;
[0081] Obtain the shortest distance between each edge of the spinal region and the extended straight line segment of the abdominal aorta region;
[0082] According to the direction coefficients of all edges in the spine region and the shortest distance between all edges in the spine region and the extended straight line segments in the abdominal aorta region, the collinearity coefficients of any two edges in the spine region are obtained:
[0083]
[0084] In the formula, The first Article and Collinearity coefficient of strip edges; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first Direction coefficient of the strip edge; The first Direction coefficient of the strip edge; is the absolute value function.
[0085] It should be noted that the first Article and The smaller the difference between the shortest distances of the edge of the spine and the extended straight line segment of the abdominal aorta area, the greater the collinearity coefficient; Article and The smaller the difference in the directional coefficients of the edges, the greater the degree of parallelism of the two edges, that is, the greater the collinearity coefficient; if the denominator is zero, set the denominator to one to ensure that the formula is valid.
[0086] In this embodiment, the preset directivity threshold is 8 degrees, which is described as an example. In other implementations, it can be set to other values, which are not limited in this embodiment;
[0087] Among all the edges in the spine area, all the edges with direction coefficients less than the preset direction coefficient threshold are recorded as suspected edges on both sides of the spine;
[0088] It should be noted that in order to ensure the accuracy of the detection of the edges on both sides of the spine, it is necessary to further identify the edges on both sides of the spine based on the collinearity coefficient.
[0089] The preset number of second clusters in this embodiment is 5. In the spine region, the reciprocal of the collinearity coefficient of any two suspected spine edges is used as the clustering distance. According to the preset number of second clusters, the K-means clustering algorithm is used to perform clustering operations on all suspected spine edges to obtain a number of clusters.
[0090] The mean of the collinearity coefficients of all two suspected spine edges in each cluster is recorded as the spine edge possibility of each cluster;
[0091] The first two clusters with the greatest spine edge probability are recorded as the left spine edge cluster and the right spine edge cluster, respectively.
[0092] In the spine region, convex hull detection is performed on the edges on both sides of the suspected spine in the left spine edge cluster and the right spine edge cluster to obtain a convex hull region, wherein convex hull detection is a well-known technology;
[0093] It should be noted that the edge of the vertebra is between the left and right edges of the spine, so the convex hull area includes the edge of the vertebra.
[0094] The edges on both sides of the suspected spine that are not in the left spine edge cluster and the right spine edge cluster within the convex hull region are recorded as suspected vertebral edges;
[0095] The least square method is used to fit multiple rectangles to all edges in the convex hull region to obtain the fitting error of each suspected vertebral edge, wherein the least square method of fitting rectangles is a well-known technique.
[0096] The preset angle threshold in this embodiment is 90 degrees, which is used as an example for description. In other implementations, it can be set to other values, which are not limited in this embodiment;
[0097] According to the direction coefficient and fitting error of each suspected vertebral edge, the vertebral edge possibility of each suspected vertebral edge is obtained:
[0098]
[0099] In the formula, Indicates Possibility of vertebral margins of suspected vertebral margins; Indicates Direction coefficient of the suspected vertebral edge; is the preset angle threshold; is the absolute value function; Indicates The fitting error of the suspected vertebral edge; is an exponential function with a natural constant as the base. To present the inverse proportional relationship and normalization processing, the implementer can set the inverse proportional function and normalization function according to the actual situation.
[0100] It should be noted that Shidi The difference between the direction coefficient of the suspected vertebral edge and 90 degrees, the smaller the difference, the The more perpendicular the straight line segment extending from the suspected vertebral edge to the abdominal aorta area is, the greater the possibility that it belongs to the vertebral edge. The smaller the fitting error of the suspected vertebral edge, the better the The closer the suspected vertebral edge is, the more likely it is a vertebral edge.
[0101] In this embodiment, the vertebral edge possibility threshold is preset to 0.8, which is used as an example for description. In other implementations, it can be set to other values, which are not limited in this embodiment;
[0102] A suspected vertebral edge whose vertebral edge likelihood is greater than a preset vertebral edge likelihood threshold is recorded as a vertebral edge.
[0103] The suspected two side edges of the spine and the vertebral edges in the left spine edge cluster and the right spine edge cluster are recorded as spine edges.
[0104] Using the least square method, multiple rectangles are fitted to all spine edges to obtain several vertebral rectangles, the fitting error of each vertebral rectangle and the fitting error of each spine edge. The least square method of rectangle fitting is a well-known technology.
[0105] Step S004: The interference distortion coefficient of the spine is obtained according to the distance between the pixel points on the boundary of the abdominal aorta area, the number of vertebral rectangles and the fitting error of each vertebral rectangle; the degree of influence of the spine is obtained according to the distance between the edges of the spine, the fitting error of each edge of the spine and the gradient amplitude of all pixel points on the edge of the spine.
[0106] In the several vertebral rectangles obtained in the previous step, the better the edge fitting effect, the less edge distortion caused by tissue influence; however, for the edges with poor rectangle fitting conditions, the width of the nearest aortic part is obtained, and the larger the width, the more distortion is caused by abdominal aortic aneurysm compression.
[0107] The shortest distance from each vertebral rectangle to the abdominal aorta region corresponds to the pixel point on the border of the abdominal aorta region, which is recorded as the main pixel point. On the border of the abdominal aorta region, starting from the main pixel point, the normal of the main pixel point is traversed one pixel at a time into the abdominal aorta region, and the other pixel points on the first traversed border of the abdominal aorta region are recorded as sub-pixel points. The distance from the main pixel point to the sub-pixel point is recorded as the width of the abdominal aorta closest to each vertebral rectangle.
[0108] According to the distance of each vertebral rectangle to the nearest abdominal aorta width, the number of vertebral rectangles and the fitting error of each vertebral rectangle, the interference distortion coefficient of the spine is obtained:
[0109]
[0110] In the formula, represents the interference distortion coefficient of the spine; Represents the number of vertebral rectangles; Indicates The fitting error of the vertebral rectangle; Indicates The width of the abdominal aorta closest to the vertebral rectangle; is the absolute value function; It is a hyperbolic function, and the result value is normalized in direct proportion.
[0111] It should be noted that for The fitting error of the vertebral rectangle , the larger the value, the worse the rectangular fitting effect. The width of the abdominal aorta closest to the vertebral rectangle The larger the value, the more likely the vertebral rectangle is to have spinal deformation caused by aneurysm. When the numerator and denominator are the same size, it means that the rectangular fitting effect is poor and the width of the nearest abdominal aorta is large. At this time, the spinal deformation caused by the abdominal aorta is low, so the interference distortion coefficient is low. When the numerator and denominator are the same, it means that the fitting effect with the rectangle is good and the width of the nearest abdominal aorta is small. At this time, the patient is in a normal state, so its interference distortion coefficient is low; when The difference between the numerator and the denominator indicates that the quality of the fitting to the rectangle has nothing to do with the width of the nearest abdominal aorta, so it is caused by interference factors, and therefore its interference distortion coefficient is high.
[0112] Since the spine is a connected domain of several vertebrae, for each edge of the spine, the greater the distance from the nearest edge, the more affected it is; and since the spine is relatively smooth, the rougher the edge of the spine, the more affected it is; and the larger the variance of the gradient of the spine edge, the greater the change in the edge gradient and the higher the degree of impact.
[0113] According to the distance between the spine edges, the fitting error of each spine edge and the gradient amplitude of all pixels on the spine edge, the degree of influence of the spine is obtained:
[0114]
[0115] In the formula, Indicates the degree to which the spine is affected; represents the number of spine edges; Indicates The minimum value of the closest distance between the edge of a spine and all other spine edges; Indicates The fitting error of the spine edge; Indicates The variance of the gradient amplitude of all pixels on the edge of the spine; It is a hyperbolic function, and the result value is normalized in direct proportion.
[0116] It should be noted that the minimum value of the closest distance between the spine edge and all other spine edges is The larger the value, the greater the fracture at the edge of the spine, so the higher the degree of impact; and the fitting error of the spine edge , the larger the value, the less smooth the spine edge is, so the more affected it is; and the variance of the gradient of all pixels on the spine edge is The larger the value, the greater the gradient change of the spine edge, and therefore the higher the degree of influence.
[0117] Step S005: Obtain the filtering weight of the grayscale image of the patient's spine according to the interference distortion coefficient and the degree of influence of the spine; filter the grayscale image of the patient's spine according to the filtering weight to obtain a filtered image; perform edge detection on the filtered image to obtain the edge image of the patient's spine, match the edge image of the patient's spine with the edge image of the healthy human spine, and obtain the patient's spine deformation detection result.
[0118] In order to eliminate the blurred interference of body tissue in the grayscale image of the patient's spine and restore the real spinal image, the Wiener filter is selected to process the image. It can estimate the real image information in the presence of noise and is generally used for image restoration. It controls the filtering intensity by adjusting the scale factor. Therefore, the scale factor of the Wiener filter is obtained according to the filtering weight, so that it can better restore the spinal edge information.
[0119] According to the interference distortion coefficient and the degree of influence of the spine, the filtering weight of the patient's spine grayscale image is obtained:
[0120]
[0121] In the formula, represents the filter weights of the patient's spine grayscale image; represents the interference distortion coefficient of the spine; Indicates the extent to which the spine is affected.
[0122] According to the filtering weight of the patient's spine grayscale image, the patient's spine grayscale image is subjected to Wiener filtering to obtain a filtered image, wherein Wiener filtering is a well-known technology;
[0123] Performing Canny edge detection on the filtered image to obtain an edge image of the patient's spine, wherein Canny edge detection is a well-known technology;
[0124] The image pairing technology is used to calculate the patient's spine edge image and the healthy human spine edge image, and the mean square error of all edge pixels after the spine edge image matching is obtained, and the mean square error is normalized;
[0125] It should be noted that the mean square error is a commonly used indicator for measuring the quality of image pairing, which can measure the difference between the pairing result and the actual situation. In the image pairing task, the mean square error is usually used to evaluate the pixel-level matching quality between two images, which is a well-known technology. In this embodiment, the matching results of multiple patient spine edge images and healthy human spine edge images are obtained, and the maximum and minimum norm method is used to normalize the mean square error, which is a well-known technology.
[0126] The preset mean square error threshold in this embodiment is 0.3, which is used as an example for description. Other values may be set in other implementations, which are not limited in this embodiment.
[0127] When the normalized value of the mean square error of the patient's spine edge image is greater than the mean square error threshold, it is determined that the patient's spine is deformed.
[0128] So far, the present invention is completed.
[0129] In summary, in an embodiment of the present invention, a grayscale image of the patient's spine and an edge image of the healthy human spine are obtained, and the abdominal aorta region and the spinal region are divided, a number of spinal edges are screened out from all edges, the interference distortion coefficient and the degree of influence of the spine are calculated, and then the filter weight of the grayscale image of the patient's spine is obtained, the grayscale image of the patient's spine is filtered to obtain a filtered image, and the edge detection is performed to obtain the edge image of the patient's spine, which is matched with the edge image of the healthy human spine to obtain the patient's spinal deformation detection result. The present invention obtains the abdominal aorta region through the abdominal aorta characteristics, screens the spinal edge according to the distribution characteristics between the spine and the abdominal aorta, and finally obtains a clear spinal image by filtering according to the degree to which the edge of the spine is affected, providing good data support for subsequent spinal deformation identification.
[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm, characterized in that: The method comprises the following steps: Acquire a grayscale image of the patient's spine including the patient's abdominal aorta and an image of the edge of a healthy human spine; The abdominal aorta region and the spinal column region are segmented from the patient's spinal column grayscale image; According to the distribution of the edges in the spinal region and the distance from each edge to the abdominal aorta region, a number of spinal edges are selected from all the edges; multiple rectangles are fitted to all the spinal edges to obtain a number of vertebral rectangles, the fitting error of each vertebral rectangle and the fitting error of each spinal edge; The interference distortion coefficient of the spine is obtained according to the distance between the pixel points on the boundary of the abdominal aorta area, the number of vertebral rectangles and the fitting error of each vertebral rectangle; the degree of influence of the spine is obtained according to the distance between the edges of the spine, the fitting error of each spine edge and the gradient amplitude of all pixel points on the edges of the spine; According to the interference distortion coefficient and the degree of influence of the spine, the filtering weight of the grayscale image of the patient's spine is obtained; according to the filtering weight, the grayscale image of the patient's spine is filtered to obtain a filtered image; the edge detection of the filtered image is performed to obtain the edge image of the patient's spine, and the edge image of the patient's spine is matched with the edge image of the healthy human spine to obtain the patient's spine deformation detection result.
2. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 1, characterized in that: The step of dividing the abdominal aorta region and the spinal column region from the patient's spinal column grayscale image includes the following specific steps: Obtaining a segmentation threshold of the patient's spine grayscale image, and recording the region formed by all pixels with grayscale values greater than the segmentation threshold in the patient's spine grayscale image as the abdominal aorta region; According to the preset first cluster number, a clustering operation is performed on the area except the abdominal aorta area in the patient's spine grayscale image to obtain a plurality of cluster areas; Use the Canny edge detection algorithm to perform operations on each cluster area to obtain several edge lines; Use edge tracking to obtain the bifurcation points on all edge lines, and use the bifurcation points as segmentation points to segment each edge line into several edges; Perform straight line fitting on all edges respectively to obtain the fitting error of each edge; The mean of the fitting errors of all edges in each cluster region is calculated, and the cluster region corresponding to the smallest mean of the fitting errors of all edges is recorded as the spine region.
3. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 1, characterized in that: The method of selecting a plurality of spinal edges from all the edges according to the distribution of the edges in the spinal region and the distance between each edge and the abdominal aorta region includes the following specific steps: Perform straight line fitting on all edges respectively to obtain the fitting straight line corresponding to each edge; Obtain the distances between all the connecting lines of any two pixels in the abdominal aorta region, and record the connecting line of the two pixels corresponding to the maximum value of the connecting line distances as the extended straight line segment of the abdominal aorta region; The minimum angle formed by the fitting straight line of each edge of the spine region and the straight line extending through the abdominal aorta region is recorded as the direction coefficient of each edge of the spine region; According to the direction coefficients of all the edges of the spine region and the shortest distances between all the edges of the spine region and the extended straight line segments of the abdominal aorta region, the collinearity coefficients of any two edges of the spine region are obtained; Among all the edges in the spine area, all the edges with direction coefficients less than the preset direction coefficient threshold are recorded as suspected edges on both sides of the spine; According to the collinearity coefficients of all the suspected edges on both sides of the spine, several spine edges are screened out from all the suspected edges on both sides of the spine.
4. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 3, characterized in that: The specific calculation formula for obtaining the collinearity coefficient of any two edges of the spine region based on the direction coefficients of all edges of the spine region and the shortest distance between all edges of the spine region and the extended straight line segments of the abdominal aorta region is: In the formula, The first Article and Collinearity coefficient of strip edges; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first The shortest distance between the edge of the strip and the extended straight line segment of the abdominal aorta region; The first Direction coefficient of the strip edge; The first Direction coefficient of the strip edge; is the absolute value function.
5. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 3, characterized in that: The method of selecting a plurality of spine edges from all the suspected spine edges on both sides according to the collinearity coefficients of all the suspected spine edges includes the following specific steps: In the spine region, the inverse of the collinearity coefficient of any two suspected spine edges is taken as the clustering distance, and according to the preset number of second clusters, all the suspected spine edges are clustered to obtain a number of clusters; The mean of the collinearity coefficients of all two suspected spine edges in each cluster is recorded as the spine edge possibility of each cluster; The first two clusters with the highest spine edge probability are recorded as the left spine edge cluster and the right spine edge cluster respectively; In the spine region, convex hull detection is performed on the edges on both sides of the suspected spine in the left spine edge cluster and the right spine edge cluster to obtain a convex hull region; The edges on both sides of the suspected spine that are not in the left spine edge cluster and the right spine edge cluster within the convex hull region are recorded as suspected vertebral edges; Multiple rectangles are fitted to all edges within the convex hull region to obtain the fitting error of each suspected vertebral edge; According to the direction coefficient and fitting error of each suspected vertebral edge, the vertebral edge possibility of each suspected vertebral edge is obtained; The suspected vertebral edge whose vertebral edge possibility is greater than a preset vertebral edge possibility threshold is recorded as a vertebral edge; The suspected two side edges of the spine and the vertebral edges in the left spine edge cluster and the right spine edge cluster are recorded as spine edges.
6. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 5, characterized in that: The specific calculation formula for obtaining the vertebral edge possibility of each suspected vertebral edge based on the direction coefficient and fitting error of each suspected vertebral edge is: In the formula, Indicates Possibility of vertebral margins of suspected vertebral margins; Indicates Direction coefficient of the suspected vertebral edge; is the preset angle threshold; is the absolute value function; Indicates The fitting error of the suspected vertebral edge; is an exponential function with a natural constant as its base.
7. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 1, characterized in that: The method of obtaining the interference distortion coefficient of the spine according to the distance between the pixel points on the boundary of the abdominal aorta region, the number of vertebral rectangles and the fitting error of each vertebral rectangle includes the following specific steps: The shortest distance from each vertebral rectangle to the abdominal aorta region corresponds to the pixel point on the border of the abdominal aorta region, which is recorded as the main pixel point. On the border of the abdominal aorta region, starting from the main pixel point, the normal line of the main pixel point is traversed to the inside of the abdominal aorta region pixel by pixel point, and the other pixel points on the first traversed border of the abdominal aorta region are recorded as sub-pixel points. The distance from the main pixel point to the sub-pixel point is recorded as the width of the abdominal aorta closest to each vertebral rectangle. The interference distortion coefficient of the spine is obtained according to the distance of each vertebral rectangle to the nearest abdominal aorta width, the number of vertebral rectangles and the fitting error of each vertebral rectangle.
8. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 7, characterized in that: The specific calculation formula for obtaining the interference distortion coefficient of the spine according to the width of the abdominal aorta closest to each vertebral rectangle, the number of vertebral rectangles, and the fitting error of each vertebral rectangle is: In the formula, represents the interference distortion coefficient of the spine; represents the number of vertebral rectangles; Indicates The fitting error of the vertebral rectangle; Indicates The width of the abdominal aorta closest to the vertebral rectangle; is the absolute value function; is a hyperbolic function.
9. A spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 1, characterized in that: The specific calculation formula for obtaining the degree of influence of the spine according to the distance between the spine edges, the fitting error of each spine edge and the gradient amplitude of all pixel points on the spine edge is: In the formula, Indicates the degree to which the spine is affected; represents the number of spine edges; Indicates The minimum value of the closest distance between the edge of a spine and all other spine edges; Indicates The fitting error of the spine edge; Indicates The variance of the gradient amplitude of all pixels on the edge of the spine; is a hyperbolic function.
10. The spinal deformity identification method for auxiliary diagnosis of abdominal aortic aneurysm according to claim 1, characterized in that: The specific steps of obtaining the filtering weight of the patient's spine grayscale image according to the interference distortion coefficient and the affected degree of the spine are as follows: The product of the interference distortion coefficient of the spine and the affected degree of the spine is recorded as the filtering weight of the patient's spine grayscale image.
