Human body scoliosis evaluation system based on image analysis

The image analysis-based human scoliosis assessment system automatically processes spinal images, generates spinal curve fitting points, and calculates the Cobb angle, solving the problems of large errors and cumbersome operation of traditional diagnostic methods, and achieving efficient and accurate scoliosis assessment and monitoring.

CN118967584AInactive Publication Date: 2024-11-15FOURTH PEOPLES HOSPITAL OF SHAANXI PROVINCE
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410968867.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing methods for scoliosis diagnosis rely on doctors' subjective judgment, which has large errors and is cumbersome to operate, making them unsuitable for large-scale screening and follow-up monitoring.

Method used

An image-based human scoliosis assessment system is used to acquire spinal images through automated image processing technology, generate spinal curve fitting points, calculate the maximum Cobb angle, and assess the degree of scoliosis.

Benefits of technology

It improves the efficiency and accuracy of scoliosis assessment, automates and standardizes diagnosis, facilitates early detection and long-term monitoring, and enhances the overall level of clinical diagnosis and treatment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967584A_ABST
    Figure CN118967584A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of human body spine detection, in particular to a human body scoliosis evaluation system based on image analysis. Program instructions stored in a memory of the human body scoliosis evaluation system based on image analysis enable a processor of the human body scoliosis evaluation system to execute the following steps: acquiring a first spine image which is a normal phase spine image of a target object; obtaining a plurality of spine sub-images by using the first spine image; according to pixel distribution characteristics in the spine area, spine curve fitting points corresponding to the spine sub-images are generated respectively, and all the spine curve fitting points in the first spine image are utilized to fit a spine curve; and acquiring the maximum Cobb angle of the spine curve, and evaluating the scoliosis degree of the target object based on the maximum Cobb angle. The problems that a traditional manual evaluation method is high in subjectivity, insufficient in accuracy, tedious in operation and the like can be solved, and automation and standardization of scoliosis diagnosis can be achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of human spine detection, and in particular to a human scoliosis assessment system based on image analysis. Background Art

[0002] Scoliosis is a common spinal disease characterized by an abnormal curvature of the spine in the coronal plane, usually accompanied by spinal rotation. This disease can occur at any age, but is particularly common in children and adolescents. Scoliosis not only affects the patient's posture and appearance, but may also cause serious health problems such as back pain, breathing difficulties, and nerve damage. Therefore, early detection and continuous monitoring of the progression of scoliosis are critical to prevention and treatment.

[0003] Currently, the diagnosis of scoliosis mainly relies on doctors' visual assessment of X-rays and measurement of the Cobb angle. The Cobb angle is a standard angle for assessing the degree of scoliosis, which is usually calculated by manually drawing two straight lines perpendicular to the most curved part of the spine on an X-ray and measuring their angle. Although this method is intuitive, it is limited by the doctor's experience and subjective judgment, is prone to errors, and is cumbersome to operate, making it unsuitable for large-scale screening and follow-up monitoring. Summary of the invention

[0004] In response to the needs of practical applications, the present invention provides a human scoliosis assessment system based on image analysis, which aims to comprehensively assess scoliosis through automated and precise image processing technology, thereby addressing the shortcomings of traditional methods.

[0005] The human scoliosis assessment system based on image analysis provided by the present invention includes a memory, the memory stores a computer program, the computer program includes program instructions, and the program instructions, when executed by a processor, cause the processor to perform the following steps: obtain a first spinal image, the first spinal image is a frontal phase spinal image of a target object; use the first spinal image to obtain a plurality of spinal sub-images, any spinal sub-image includes a spinal region and / or a non-spinal region; generate spinal curve fitting points corresponding to the spinal sub-images according to pixel distribution characteristics in the spinal region, and use all the spinal curve fitting points in the first spinal image to fit the spinal curve; obtain the maximum Cobb angle of the spinal curve, and based on the maximum Cobb angle, assess the degree of scoliosis of the target object.

[0006] The human scoliosis assessment system based on image analysis provided by the present invention has the following benefits: the present invention utilizes stored computer program instructions to automatically obtain the anteroposterior spinal image of the target object and generate a number of spinal sub-images; then generates spinal curve fitting points based on the pixel distribution characteristics in the spinal region, performs curve fitting, and finally calculates the maximum Cobb; and comprehensively assesses the degree of scoliosis of the target object based on the maximum Cobb angle. The present invention can significantly improve the efficiency and accuracy of scoliosis assessment through automated and precise image processing technology, which not only solves the problems of strong subjectivity, insufficient accuracy, and cumbersome operation of traditional manual assessment methods, but also realizes the automation and standardization of scoliosis diagnosis, which is helpful for early detection and long-term monitoring of scoliosis, and contributes to the overall level of clinical diagnosis and treatment.

[0007] Furthermore, the method of obtaining a plurality of spinal sub-images using the first spinal image comprises the following steps: setting a capture window parameter according to a size of the first spinal image, the capture window parameter comprising a capture window length, a capture window width, a capture window moving step and a capture window moving direction; generating a spinal sub-image capture window based on the capture window length and the capture window width; and using the capture window to capture a plurality of spinal sub-images in the first spinal image based on the capture window moving step and the capture window moving direction.

[0008] Furthermore, the method of generating spinal curve fitting points of corresponding spinal sub-images according to pixel distribution characteristics in the spinal region includes the following steps: identifying the spinal region in each spinal sub-image, and cutting out the corresponding spinal region from the spinal sub-image according to the identification result of each spinal sub-image; generating spinal curve fitting points of corresponding spinal sub-images according to pixel distribution characteristics in the spinal region.

[0009] Furthermore, the method of generating spinal curve fitting points of corresponding spinal sub-images respectively according to the pixel distribution characteristics of the spinal region comprises the following steps: binarizing the spinal region and obtaining the coordinates of the pixel points distributed in the spinal region after binarization; obtaining the mean coordinates of the pixel point distribution using the pixel point coordinates distributed in the spinal region after binarization; subtracting the mean coordinates of the pixel point distribution from each pixel point coordinate respectively to obtain the corresponding decentralized pixel point coordinates; obtaining the pixel point distribution covariance matrix using the decentralized pixel point coordinates, and obtaining the eigenvector corresponding to the maximum eigenvalue of the pixel point distribution covariance matrix; generating the slope of the pixel distribution symmetry axis based on the eigenvector, and generating the pixel distribution symmetry axis based on the slope in the spinal region with the mean coordinates of the pixel point distribution as the base point; setting the intersection of the pixel distribution symmetry axis and the edge of the corresponding spinal region as the spinal curve fitting point.

[0010] Furthermore, the method of fitting the spinal curve using all the spinal curve fitting points in the first spinal image includes the following steps: summarizing all the spinal curve fitting points in the first spinal image; sorting the spinal curve fitting points from large to small based on the ordinate values ​​of the spinal curve fitting points to obtain a spinal curve fitting point sequence; and fitting the spinal curve using all the spinal curve fitting points in the spinal curve fitting point sequence.

[0011] Furthermore, the method of fitting a spinal curve using all spinal curve fitting points in the spinal curve fitting point sequence includes the following steps: setting fitting window parameters, the fitting window parameters including fitting window size and fitting window step size, and setting a curve fitting model and a weighting function within the fitting window; generating a fitting window based on the fitting window size, and intercepting a number of subsequences from the spinal curve fitting point sequence using the fitting window based on the fitting window step size; and adjusting the central spinal curve fitting points in a number of subsequences respectively using the curve fitting model and the weighting function, and sequentially connecting the adjusted central spinal curve fitting points to obtain a spinal curve.

[0012] Furthermore, the use of the curve fitting model and the weighting function to adjust the central spine curve fitting points in several subsequences respectively includes the following steps: selecting a central spine curve fitting point in a subsequence; using the weighting function to obtain the weight of each spine curve fitting point in the subsequence relative to the central spine curve fitting point, and using the weight to generate a weighting matrix; based on the curve fitting model, using all the spine curve fitting points in the subsequence to generate an input matrix and a response matrix; fitting curve fitting model parameters through the weighting matrix, the input matrix and the response matrix, and using the curve fitting model parameters in combination with the curve fitting model to adjust the central spine curve fitting points.

[0013] Furthermore, the curve fitting model is a third-order polynomial model, and the weighting function is a Gaussian weighting function.

[0014] Furthermore, the method of obtaining the maximum Cobb angle of the spinal curve includes the following steps: obtaining the inflection point of the spinal curve; sorting all the inflection points and all the endpoints of the spinal curve from large to small according to the ordinate values ​​to obtain a Cobb angle fitting sequence; on the spinal curve, respectively obtaining the tangent angle of any two points in the Cobb angle fitting sequence, and setting the largest tangent angle as the maximum Cobb angle.

[0015] Furthermore, the evaluating the scoliosis degree of the target object based on the maximum Cobb angle comprises the following steps: evaluating the scoliosis degree of the target object according to a scoliosis degree evaluation rule in combination with the maximum Cobb angle of the spinal curve, the scoliosis degree evaluation rule being: if the maximum Cobb angle of the spinal curve is within [0, 10°), the spine of the target object is normal; if the maximum Cobb angle of the spinal curve is within [10°, 20°), the spine of the target object is mild scoliosis; if the maximum Cobb angle of the spinal curve is within [20°, 40°), the spine of the target object is moderate scoliosis; if the maximum Cobb angle of the spinal curve is equal to or greater than 40°, the spine of the target object is severe scoliosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the structure of a human scoliosis assessment system based on image analysis provided by an embodiment of the present invention;

[0017] Figure 2 The present invention is a flow chart of a method for assessing human scoliosis executed by a processor in a human scoliosis assessment system based on image analysis. DETAILED DESCRIPTION

[0018] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0019] In one embodiment, see Figure 1 , Figure 1 A schematic diagram of the structure of a human scoliosis assessment system based on image analysis provided by an embodiment of the present invention.

[0020] like Figure 1 As shown, the human scoliosis assessment system based on image analysis provided by the embodiment of the present invention includes an input device, a processor, a memory and an output device, and the input device, the processor, the memory and the output device are interconnected.

[0021] Furthermore, the memory in the human scoliosis assessment system based on image analysis stores a computer program, and the computer program includes program instructions, and when the program instructions are executed by the processor, the processor executes the following steps: Figure 2 Steps for assessing scoliosis in a human are shown.

[0022] See also Figure 2 , Figure 2 The flowchart of the human scoliosis assessment method executed by the processor in the human scoliosis assessment system based on image analysis provided by the embodiment of the present invention. Figure 2 As shown, the human scoliosis assessment method comprises the following steps:

[0023] S01. Acquire a first spinal image, where the first spinal image is a anteroposterior spinal image of a target object.

[0024] Normally, medical images of the human spine can be obtained through medical imaging equipment such as ordinary X-ray machines (X-ray Machine), computer X-ray photography systems (CR, Computed Radiography), digital X-ray photography systems (DR, Digital Radiography), X-ray computed tomography devices (CT, Computed Tomography), and magnetic resonance imaging (MRI).

[0025] In this embodiment, the input device is a digital X-ray photography system, and the processor is signal-connected to the digital X-ray photography system. When the program instructions are executed by the processor, the processor obtains a full-length DR image of the human spine when the target object is in the positive phase, which is taken and stored by the digital X-ray photography system.

[0026] Furthermore, the anteroposterior spinal image refers to a medical spinal image taken from the front of the target object. In this embodiment, the full-length DR image of the human spine when the target object is in the anteroposterior phase refers to the entire medical spinal image of the human body from the cervical vertebrae, thoracic vertebrae, lumbar vertebrae to the sacrum obtained by taking the image from the front of the target object through a digital X-ray photography system.

[0027] It is understandable that in one or some other embodiments, the input device may also be other medical imaging devices connected to the processor signal, for capturing and storing partial or full-length images of the human spine when the target object is in the positive phase.

[0028] S02. Using the first spinal column image, obtain a plurality of spinal column sub-images, wherein any spinal column sub-image includes a spinal column region and / or a non-spinal column region.

[0029] Furthermore, the step S02 of using the first spine image to obtain a plurality of spine sub-images includes the following steps:

[0030] S021. Set capture window parameters according to the size of the first spine image.

[0031] In this embodiment, the interception window parameters include the interception window length, the interception window width, the interception window moving step length and the interception window moving direction.

[0032] It is understandable that the specific value of any parameter in the interception window parameters can be adjusted and optimized according to factors such as the size of the input first spine image and the spine curve fitting requirements.

[0033] Furthermore, the capture window length represents the number of pixels of the capture window in the vertical direction of the first spine image; the capture window width represents the number of pixels of the capture window in the horizontal direction of the first spine image; the capture window moving step represents the number of pixels that the capture window moves in the capture window moving direction each time when capturing a sub-image in the first spine image; the capture window moving direction refers to the direction of the capture window when capturing a sub-image in the first spine image.

[0034] S022. Generate a spine sub-image capture window based on the capture window length and the capture window width.

[0035] In this embodiment, a first spinal image with a size of L (length) × W (width) equals M (pixels) × N (pixels), and its capture window length l is much smaller than the length L of the first spinal image, and its capture window width w is consistent with the width W of the first spinal image, thereby ensuring that all spinal regions in the first spinal image can be gradually captured, thereby decomposing them into small spinal regions for subsequent data analysis.

[0036] S023. Based on the moving step length of the capture window and the moving direction of the capture window, use the capture window to capture a plurality of spine sub-images in the first spine image.

[0037] In this embodiment, based on the first spine image with a size of L (length) × W (width) of M (pixels) × N (pixels), the moving step of the cropping window is half the sub-image length l, and the moving direction of the cropping window is vertical to the first spine image.

[0038] Furthermore, based on the spine sub-image capture window parameters set in step S031, spine sub-images p1, p2, ..., p3, p4, ..., p5, ..., p6, ..., p7, ..., p8, ..., p9, ..., p10, ..., p11, ..., p12, ..., p13, ..., p14, ..., p15, ..., p16, ..., p17, ..., p18, ..., p19, ..., p n-1 and the spine sub-image p n etc. multiple consecutive spine sub-images.

[0039] S03. Generate spinal curve fitting points corresponding to the spinal sub-images according to pixel distribution characteristics in the spinal region, and fit the spinal curve using all the spinal curve fitting points in the first spinal image.

[0040] It is understandable that the spine pixels in the human spine medical image are usually distributed based on axisymmetry. Further, the step S03 described in which the spine curve fitting points corresponding to the spine sub-images are generated according to the pixel distribution characteristics in the spine region includes the following steps:

[0041] S0311. Identify the spinal column region in each spinal column sub-image, and cut out the corresponding spinal column region from the spinal column sub-image according to the identification result of each spinal column sub-image.

[0042] Any spinal sub-image captured from the first spinal image may be entirely non-spinal areas, or may include non-spinal areas and spinal areas. Therefore, in order to ensure the accuracy of subsequent spinal image symmetry analysis, the spinal sub-image needs to be further processed.

[0043] Furthermore, step S0311 can realize the recognition and selection of the spinal region in the spinal sub-image by deep learning technology. Specifically, a customized convolutional neural network (CNN) model is trained or a model improved or adjusted based on an existing image segmentation model (such as U-Net) is trained, and the verified model is used to realize the recognition and selection of the spinal region in each spinal sub-image.

[0044] S0312. Generate spine curve fitting points corresponding to the spine sub-images according to the pixel distribution characteristics of the spine region.

[0045] In this embodiment, the generation task of the spinal curve fitting points in different spinal regions in step S0312 can be performed step by step as a single generation task, or multiple generation tasks can be performed in parallel.

[0046] Furthermore, the task of generating a spinal curve fitting point in any spinal region in step S0312 includes the following steps:

[0047] S03121. Binarize the spine region, and obtain the coordinates of the pixel points distributed in the spine region after binarization.

[0048] Typically, human-computer spinal medical images are displayed in the form of grayscale images. In this embodiment, the first spinal image acquired in step S01 is a grayscale image, so the spinal region image captured from the first spinal image is also a grayscale image.

[0049] In this embodiment, by setting a grayscale threshold, the pixel value of each pixel in the spine area is binarized, and the specific binarization rules are as follows: if the pixel value exceeds the grayscale threshold, the pixel value is set to 1 (foreground); otherwise, it is set to 0 (background).

[0050] Furthermore, the grayscale threshold can be set and / or adjusted according to the pixel value in the spine region. Grayscale threshold The following model is satisfied:

[0051] Among them, p i represents the i-th spine sub-image intercepted from the first spine image, represents the spine region in the i-th spine sub-image, Indicates the spinal region The number of pixels in Indicates the spinal region One pixel coordinate inside, Indicates the spinal region The gray value of the pixel at the inner pixel coordinate (x, y), Indicates the spinal region The sum of the grayscale values ​​of all pixels in it.

[0052] Furthermore, the spine region in the i-th spine sub-image intercepted from the first spine image inside, if Less than the corresponding gray threshold Then set the pixel to 0 (background). Greater than or equal to the corresponding grayscale threshold Then set the pixel to 1 (foreground).

[0053] S03122. Using the binarized pixel coordinates distributed in the spine region, obtain the mean coordinates of the pixel distribution.

[0054] It can be understood that the pixel point coordinates described in step S03122 are the pixel point coordinates whose pixel value is 1 after binarization in step S03121.

[0055] In this embodiment, the spine area in the i-th spine sub-image The mean coordinates of the pixel distribution satisfy the following calculation formula:

[0056] in, Indicates the spinal region The horizontal coordinate of the mean pixel distribution, Spine Region The vertical coordinate of the mean pixel distribution, Indicates the spinal region The number of pixels with a pixel value of 1 after internal binarization. Indicates the spinal region The horizontal coordinate of a pixel point with a pixel value of 1 after internal binarization. Indicates the spinal region The horizontal coordinates and the horizontal coordinates of the pixels whose pixel values ​​are 1 after internal binarization. Indicates the spinal region The vertical coordinate of a pixel point with a pixel value of 1 after internal binarization. Indicates the spinal region The sum of the vertical coordinates of the pixels whose pixel value is 1 after internal binarization.

[0057] S03123. Subtract the pixel distribution mean coordinate from each pixel coordinate to obtain the corresponding decentralized pixel coordinate.

[0058] S03124. Utilize the decentralized pixel coordinates to obtain a pixel distribution covariance matrix, and obtain an eigenvector corresponding to the maximum eigenvalue of the pixel distribution covariance matrix.

[0059] In this embodiment, the spine area in the i-th spine sub-image The pixel distribution covariance matrix satisfies the following calculation model:

[0060] in, Indicates the spinal region S pi The coordinates of a pixel point with a pixel value of 1 after internal binarization The corresponding decentralized pixel coordinates, Indicates the spinal region The mean coordinates of the pixel distribution.

[0061] Furthermore, the pixel distribution covariance matrix obtained in step S03124 is The maximum eigenvalue corresponds to The eigenvector of Satisfies the following characterization model:

[0062] S03125. Based on the characteristic vector, generate the slope of the pixel distribution symmetry axis, and in the spine area, take the mean coordinates of the pixel point distribution as a base point, and generate the pixel distribution symmetry axis based on the slope.

[0063] In this embodiment, the spine area in the i-th spine sub-image The slope of the pixel distribution symmetry axis Among them, ν a With ν b Both are eigenvectors The element value in .

[0064] S03126. Set the intersection of the pixel distribution symmetry axis and the edge of the corresponding spine area as a spine curve fitting point.

[0065] In this embodiment, the spine region captured based on the spine sub-image is a rectangular region, so step S03126 is to obtain the intersection between the pixel distribution symmetry axis and the rectangular boundary of the spine region image.

[0066] Furthermore, the step S03 of fitting the spinal curve using all the spinal curve fitting points in the first spinal image comprises the following steps:

[0067] S0321. Summarize all the spinal curve fitting points in the first spinal image.

[0068] S0322. Based on the ordinate values ​​of the spinal curve fitting points, sort the spinal curve fitting points from large to small to obtain a spinal curve fitting point sequence.

[0069] S0323. Fit the spinal curve using all the spinal curve fitting points in the spinal curve fitting point sequence.

[0070] Furthermore, step S0323 uses a local polynomial fitting method to achieve spinal curve fitting. Specifically, the spinal curve fitting using all spinal curve fitting points in the spinal curve fitting point sequence includes the following steps:

[0071] S03231. Setting fitting window parameters, wherein the fitting window parameters include a fitting window size and a fitting window step size, and setting a curve fitting model and a weighting function within the fitting window.

[0072] In this embodiment, the fitting window size is 3 spinal curve fitting points; the fitting window step size is 1 spinal curve fitting point.

[0073] Furthermore, the curve fitting model is a third-order polynomial model: y = ax 3 +bx 2 +cx+d, where a, b, c and d are all parameters to be fitted of the third-order polynomial model; the weighting function is a Gaussian weighting function: Where w represents the horizontal coordinate x of a spine curve fitting point in the local fitting window relative to the horizontal coordinate x of the central spine curve fitting point. c ω represents the weight of the spinal curve fitting point contained in the local fitting window.

[0074] In one or some other embodiments, the local fitting window parameters can be optimized and adjusted. For example, the curve fitting model is set to a second-order polynomial model to improve the fitting speed; another example is setting the curve fitting model to a fourth-order polynomial model or a higher-order polynomial model to improve the accuracy of the fitting; another example is adjusting the order of the polynomial model according to the horizontal coordinate width corresponding to the spinal curve fitting points contained in the window: the wider the horizontal coordinate width, the higher the order of the polynomial model; the narrower the horizontal coordinate width, the lower the order of the polynomial model, so as to ensure that the spinal curves fitted by different subsequences are more in line with the actual situation.

[0075] S03232. Generate a fitting window based on the fitting window size, and based on the fitting window step size, use the fitting window to intercept several subsequences in the spinal curve fitting point sequence.

[0076] S03233. Utilize the curve fitting model and the weighting function to adjust the central spine curve fitting points in a plurality of subsequences respectively, connect the adjusted central spine curve fitting points in sequence, and obtain the spine curve.

[0077] Furthermore, in the process of adjusting the central spine curve fitting points in a plurality of subsequences respectively using the curve fitting model and the weighting function, adjusting the central spine curve fitting point in any subsequence comprises the following steps:

[0078] S032331. Select a central spine curve fitting point in a subsequence.

[0079] In this embodiment, the subsequence includes three spinal curve fitting points, and the second spinal curve fitting point is set as the central spinal curve fitting point.

[0080] S032332. Using the weighting function, obtain the weight of each spinal curve fitting point in the subsequence relative to the central spinal curve fitting point, and use the weight to generate a weighting matrix.

[0081] Furthermore, step S032332 uses the above-mentioned Gaussian weighting function to obtain the weight of each spinal curve fitting point relative to the central spinal curve fitting point, and the weighting matrix K generated based on the three weights satisfies the following characterization model: Among them, w1 represents the weight of the first spinal curve fitting point relative to the central spinal curve fitting point, w2=1, w2 represents the weight of the second spinal curve fitting point relative to the central spinal curve fitting point, and w3 represents the weight of the third spinal curve fitting point relative to the central spinal curve fitting point.

[0082] S032333. Based on the curve fitting model, use all the spinal curve fitting points in the subsequence to generate an input matrix and a response matrix.

[0083] In this embodiment, the curve fitting model is a third-order polynomial model: y=ax 3 +bx 2 +cx+d, and then based on the fitting characteristics of the third-order polynomial model, the horizontal and vertical coordinates of the three spinal curve fitting points are used to obtain the corresponding input matrix X and response matrix Y.

[0084] Furthermore, the input matrix X satisfies the following representation model: The input matrix Y satisfies the following representation model: Among them, (x1, y1) are the coordinates of the first spine curve fitting point of the subsequence, (x2, y2) are the coordinates of the second spine curve fitting point of the subsequence, which is also the coordinates of the central spine curve fitting point, and (x3, y3) are the coordinates of the third spine curve fitting point of the subsequence.

[0085] S032334. Fit the curve fitting model parameters through the weighting matrix, the input matrix and the response matrix, and use the curve fitting model parameters in combination with the curve fitting model to adjust the central spine curve fitting points.

[0086] In this embodiment, the curve fitting model parameters a, b, c and d can be obtained by the following calculation model: Among them, X is the input matrix, K is the weighting matrix, and Y is the output matrix.

[0087] Furthermore, based on the curve fitting model parameters fitted by the above-mentioned calculation model combined with the corresponding curve fitting model, the third-order polynomial equation corresponding to the central spine curve fitting point can be obtained. The horizontal coordinate of the central spine curve fitting point is substituted into this third-order polynomial equation to obtain the adjusted spine curve fitting point.

[0088] S04. Obtaining the maximum Cobb angle of the spinal curve, and based on the maximum Cobb angle, evaluating the degree of scoliosis of the target subject.

[0089] It is known that the Cobb angle is a standard method used in medical imaging to assess the degree of scoliosis. Further, the size of the Cobb angle is used to quantify the degree of curvature of the spine: the larger the Cobb angle, the higher the degree of scoliosis of the target subject.

[0090] In this embodiment, the maximum Cobb angle refers to the Cobb angle corresponding to the curved segment with the highest curvature in the spinal curve. Further, the step S04 of obtaining the maximum Cobb angle of the spinal curve includes the following steps:

[0091] S041. Obtain the inflection point of the spinal curve.

[0092] In this embodiment, the inflection point is the point where the spinal curve changes from convex to concave, or from concave to convex.

[0093] S042. Sort all the inflection points and all the endpoints of the spinal curve from large to small according to the ordinate values ​​to obtain a Cobb angle fitting sequence.

[0094] S043. On the spinal curve, respectively obtain the tangent angles of any two points in the Cobb angle fitting sequence, and set the largest tangent angle as the maximum Cobb angle.

[0095] In this embodiment, the tangent angle θ between any two points satisfies the following calculation model: Among them, k a represents the slope of a tangent line, k b Represents the slope of the other tangent line.

[0096] Furthermore, the step S04 of evaluating the scoliosis degree of the target object based on the maximum Cobb angle includes the following steps: evaluating the scoliosis degree of the target object according to a scoliosis degree evaluation rule in combination with the maximum Cobb angle of the spinal curve.

[0097] Among them, the scoliosis degree assessment rule is specifically as follows: if the maximum Cobb angle of the spinal curve is within [0,10°), the spine of the target object is normal; if the maximum Cobb angle of the spinal curve is within [10°,20°), the spine of the target object is mild scoliosis; if the maximum Cobb angle of the spinal curve is within [20°,40°), the spine of the target object is moderate scoliosis; if the maximum Cobb angle of the spinal curve is equal to or greater than 40°, the spine of the target object is severe scoliosis.

[0098] In this embodiment, based on the data required by the processor to execute corresponding program instructions, the input device in the human scoliosis assessment system based on image analysis is used to input data, including but not limited to spinal image data (such as the first spinal image mentioned above), user instructions (such as start, pause, end, etc. instructions), initial parameters (such as the above-mentioned capture window parameters, fitting window parameters, etc.), and target object information (such as name, age, gender, medical history, etc.).

[0099] In this embodiment, based on the data obtained by the processor executing the corresponding program instructions, the output device is used to output the evaluation results, including but not limited to the spinal curve, the maximum Cobb angle and its corresponding scoliosis degree evaluation results. Furthermore, the output device can be a display, a printer or other display and printing device, which can intuitively present the evaluation results for reference and diagnosis by doctors and related medical personnel.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents.

Claims

1. A human scoliosis assessment system based on image analysis, It is characterized in that The invention comprises a memory storing a computer program, wherein the computer program comprises program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the following steps: Acquire a first spinal image, wherein the first spinal image is a anteroposterior spinal image of a target object; Using the first spinal column image, obtaining a plurality of spinal column sub-images, any spinal column sub-image including a spinal column region and / or a non-spinal column region; Generating spinal curve fitting points corresponding to the spinal sub-images respectively according to the pixel distribution characteristics in the spinal region, and fitting the spinal curve using all the spinal curve fitting points in the first spinal image; The maximum Cobb angle of the spinal curve is obtained, and based on the maximum Cobb angle, the degree of scoliosis of the target subject is evaluated.

2. The human scoliosis assessment system based on image analysis according to claim 1, It is characterized in that The method of obtaining a plurality of spine sub-images by using the first spine image comprises the following steps: According to the size of the first spine image, setting interception window parameters, the interception window parameters including interception window length, interception window width, interception window moving step length and interception window moving direction; Generate a spine sub-image interception window based on the interception window length and the interception window width; Based on the moving step length of the capture window and the moving direction of the capture window, the capture window is used to capture a plurality of spine sub-images in the first spine image.

3. The human scoliosis assessment system based on image analysis according to claim 1, It is characterized in that The method of generating the spinal curve fitting points corresponding to the spinal sub-images according to the pixel distribution characteristics in the spinal region comprises the following steps: Identify the spinal region in each spinal sub-image, and cut out the corresponding spinal region from the spinal sub-image according to the identification result of each spinal sub-image; According to the pixel distribution characteristics of the spine area, the spine curve fitting points corresponding to the spine sub-images are generated respectively.

4. The human scoliosis assessment system based on image analysis according to claim 3, It is characterized in that The method of generating the spinal curve fitting points corresponding to the spinal sub-images according to the pixel distribution characteristics of the spinal region comprises the following steps: Binarizing the spine region and obtaining coordinates of pixel points distributed in the spine region after binarization; Using the coordinates of the pixels distributed in the spine area after binarization, the mean coordinates of the pixel distribution are obtained; Subtract the mean coordinate of the pixel distribution from each pixel coordinate to obtain the corresponding decentralized pixel coordinate; Using the decentralized pixel coordinates, a pixel distribution covariance matrix is ​​obtained, and an eigenvector corresponding to the maximum eigenvalue of the pixel distribution covariance matrix is ​​obtained; Based on the characteristic vector, the slope of the pixel distribution symmetry axis is generated, and in the spine region, the pixel distribution symmetry axis is generated based on the slope with the mean coordinates of the pixel point distribution as the base point; The intersection point between the pixel distribution symmetry axis and the edge of the corresponding spine region is set as the spine curve fitting point.

5. The human scoliosis assessment system based on image analysis according to claim 1, It is characterized in that The step of fitting the spinal curve using all the spinal curve fitting points in the first spinal image comprises the following steps: Summarizing all spine curve fitting points in the first spine image; Based on the ordinate values ​​of the spinal curve fitting points, the spinal curve fitting points are sorted from large to small to obtain a spinal curve fitting point sequence; The spinal curve is fitted using all the spinal curve fitting points in the spinal curve fitting point sequence.

6. The human scoliosis assessment system based on image analysis according to claim 5, It is characterized in that The step of fitting a spinal curve using all spinal curve fitting points in the spinal curve fitting point sequence comprises the following steps: Setting fitting window parameters, the fitting window parameters including fitting window size and fitting window step length, and setting a curve fitting model and a weighting function within the fitting window; Based on the fitting window size, a fitting window is generated, and based on the fitting window step size, a plurality of subsequences are intercepted from the spinal curve fitting point sequence using the fitting window; The curve fitting model and the weighting function are used to adjust the central spine curve fitting points in a plurality of subsequences respectively, and the adjusted central spine curve fitting points are connected in sequence to obtain the spine curve.

7. The human scoliosis assessment system based on image analysis according to claim 6, It is characterized in that The method of using the curve fitting model and the weighting function to adjust the central spine curve fitting points in a plurality of subsequences respectively comprises the following steps: Selecting a central spine curve fitting point in a subsequence; Using the weighting function, obtaining the weight of each spine curve fitting point in the subsequence relative to the central spine curve fitting point, and using the weight to generate a weighting matrix; Based on the curve fitting model, using all the spinal curve fitting points in the subsequence, generating an input matrix and a response matrix; The curve fitting model parameters are fitted through the weighting matrix, the input matrix and the response matrix, and the central spine curve fitting points are adjusted by using the curve fitting model parameters in combination with the curve fitting model.

8. The human scoliosis assessment system based on image analysis according to claim 7, It is characterized in that The curve fitting model is a third-order polynomial model, and the weighting function is a Gaussian weighting function.

9. The human scoliosis assessment system based on image analysis according to claim 1, It is characterized in that The step of obtaining the maximum Cobb angle of the spinal curve comprises the following steps: Obtaining the inflection point of the spinal curve; Arrange all the inflection points and all the endpoints of the spinal curve from large to small according to the ordinate values ​​to obtain a Cobb angle fitting sequence; On the spinal curve, the tangent angles of any two points in the Cobb angle fitting sequence are respectively obtained, and the largest tangent angle is set as the maximum Cobb angle.

10. The human scoliosis assessment system based on image analysis according to claim 1, It is characterized in that The method of evaluating the degree of scoliosis of the target subject based on the maximum Cobb angle comprises the following steps: According to the scoliosis degree assessment rule, combined with the maximum Cobb angle of the spinal curve, the scoliosis degree of the target object is assessed, and the scoliosis degree assessment rule is: If the maximum Cobb angle of the spinal curve is within [0,10°), the spine of the target subject is normal; If the maximum Cobb angle of the spinal curve is within [10°, 20°), the spine of the target subject is mild scoliosis; If the maximum Cobb angle of the spinal curve is within [20°, 40°), the spine of the target subject has moderate scoliosis; If the maximum Cobb angle of the spinal curve is equal to or greater than 40°, the spine of the target subject has severe scoliosis.

Citation Information

Cited By

  • Spine conservative treatment auxiliary decision-making method and system based on image analysis

    CN120912591A

  • Scoliosis identification method, device and equipment for back image

    CN121353814A