Cobb Angle Measurement Method, Device, Equipment, and Medium Based on Spine Curve Detection

Abstract

This invention discloses a method, apparatus, device, and medium for measuring Cobb angle based on spinal curve detection. The method includes: acquiring a target bare back image of a target object in a natural standing state; inputting the target bare back image into a spinal curve detection model to obtain a target curve coordinate probability matrix and a target curve start-end point probability matrix; the curve coordinate probability matrix describes the probability that each pixel in the bare back image belongs to the spinal curve; determining the target spinal curve based on the expectation of the target curve coordinate probability matrix, and determining the target curve start-end point row coordinates based on the expectation of the target curve start-end point probability matrix; determining multiple target row coordinates on the target spinal curve based on the target curve start-end point row coordinates, and determining the Cobb angle of the target object based on the target curvature corresponding to adjacent target row coordinates. This solution can achieve high-precision calculation of the Cobb angle using optical images.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, device and medium for measuring Cobb angle based on spinal curve detection. Background Technology

[0002] Scoliosis is a common three-dimensional spinal deformity characterized by a lateral curvature of more than 10 degrees in the coronal plane, often accompanied by vertebral rotation and changes in sagittal physiological curvature. This condition commonly occurs during adolescence and is known as adolescent idiopathic scoliosis (AIS). Without timely intervention, it can develop into severe physical deformities, affecting cardiopulmonary function and leading to chronic pain and psychological disorders. In clinical practice, Cobb angle measurement is the gold standard for assessing the severity of scoliosis; therefore, accurate and reliable Cobb angle measurement is crucial for developing AIS treatment plans and evaluating treatment effectiveness.

[0003] Currently, the diagnosis and follow-up of scoliosis mainly rely on full-spine standing X-rays, requiring doctors to manually mark the vertebral body edges and measure the Cobb angle. However, this method has significant limitations: firstly, the issue of ionizing radiation exposure needs to be considered, as frequent X-rays pose a cumulative radiation risk to adolescent patients who are in the growth and development stage; secondly, the measurement process is cumbersome, and the results are easily affected by factors such as the doctor's subjective judgment and the patient's posture, resulting in low repeatability.

[0004] To overcome the aforementioned problems, radiation-free, non-contact optical image Cobb angle measurement technology has become a major research hotspot. Existing optical image Cobb angle measurement methods rely on specialized equipment such as depth cameras and probe balls to assist in detecting spinal structures, making them difficult to operate independently of specific data acquisition hardware and thus limiting their applicability. Furthermore, compared to X-ray images, spinal structures are difficult to observe directly in optical images, making it impossible to obtain specific spinal segmentation annotations, increasing the difficulty of algorithm development, and compromising the accuracy of Cobb angle measurement. Summary of the Invention

[0005] This invention provides a method, device, equipment, and medium for measuring Cobb angle based on spinal curve detection. It utilizes a neural network model to detect the spinal curve in a bare back image and determines the Cobb angle based on the curve detection results. This method does not rely on special data acquisition equipment and avoids the risk of ionizing radiation exposure. It can improve the accuracy, safety, convenience, and adaptability of Cobb angle measurement based on optical image detection.

[0006] According to one aspect of the present invention, a method for measuring Cobb angle based on spinal curve detection is provided, the method comprising:

[0007] Obtain an optical image of the bare back of the target object in a natural standing position as the target bare back image;

[0008] The target bare back image is input into a pre-trained spinal curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, wherein the start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve.

[0009] The target spine curve corresponding to the target bare back image is determined based on the expectation of the target curve coordinate probability matrix, and the row coordinates of the target curve start and end points corresponding to the target bare back image are determined based on the expectation of the target curve start and end point probability matrix.

[0010] Based on the row coordinates of the start and end points of the target curve, multiple target row coordinates are determined on the target spine curve, and the Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates.

[0011] According to another aspect of the present invention, a Cobb angle measuring device based on spinal curve detection is provided, the device comprising:

[0012] The image acquisition module is used to acquire an optical image of the bare back of the target object in a natural standing position as the target bare back image;

[0013] The model prediction module is used to input the target bare back image into a pre-trained spinal curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, wherein the start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve;

[0014] The curve estimation module is used to determine the target spine curve corresponding to the target bare back image based on the expectation of the target curve coordinate probability matrix, and to determine the row coordinates of the target curve start and end points corresponding to the target bare back image based on the expectation of the target curve start and end point probability matrix.

[0015] The sampling calculation module is used to determine multiple target row coordinates on the target spine curve based on the row coordinates of the start and end points of the target curve, and to determine the Cobb angle of the target object according to the target curvature corresponding to adjacent target row coordinates.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the Cobb angle measurement method based on spinal curve detection according to any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the Cobb angle measurement method based on spine curve detection as described in any embodiment of the present invention.

[0019] The technical solution of this invention first acquires an optical image of the bare back of a target object in a natural standing state as the target bare back image; then, the target bare back image is input into a pre-trained spinal curve detection model to obtain a target curve coordinate probability matrix and a target curve start-end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start-end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, with the start point being the intersection of the shoulder line and the spinal curve, and the end point being the intersection of the hip line and the spinal curve; then, the target spinal curve corresponding to the target bare back image is determined according to the expectation of the target curve coordinate probability matrix, and the row coordinates of the target curve start-end point corresponding to the target bare back image are determined according to the expectation of the target curve start-end point probability matrix; furthermore, multiple target row coordinates are determined on the target spinal curve based on the row coordinates of the target curve start-end point, and the Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates. This technical solution utilizes a neural network model to detect the spinal curve in bare back images and determines the Cobb angle based on the curve detection results. It does not rely on special data acquisition equipment and avoids the risk of ionizing radiation exposure. It can improve the accuracy, safety, convenience and adaptability of Cobb angle measurement based on optical image detection.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a Cobb angle measurement method based on spinal curve detection according to an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of a Cobb angle measurement method based on spinal curve detection according to an embodiment of the present invention;

[0024] Figure 3 This is a flowchart of another Cobb angle measurement method based on spinal curve detection provided by an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of a Cobb angle measuring device based on spinal curve detection according to an embodiment of the present invention;

[0026] Figure 5 This is a schematic diagram of the structure of an electronic device that implements a Cobb angle measurement method based on spinal curve detection according to an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first," "second," "target," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] Example 1

[0030] Figure 1 This is a flowchart of a Cobb angle measurement method based on spinal curve detection provided in Embodiment 1 of the present invention. This embodiment is applicable to situations requiring high-precision and efficient measurement of the Cobb angle. The method can be executed by a Cobb angle measurement device based on spinal curve detection, which can be implemented in hardware and / or software. This Cobb angle measurement device based on spinal curve detection can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:

[0031] S110, acquire an optical image of the bare back of the target object in a natural standing position as the target bare back image.

[0032] The target object can refer to the object for which Cobb angle measurement is required. According to the definition given by the International Society for Scoliosis Research, the Cobb angle is the angle formed between the superior endplate of the uppermost vertebra and the inferior endplate of the lowermost vertebra in the spinal curvature. The optical image of the exposed back can refer to an image formed by imaging the exposed back based on optical imaging principles. For example, the optical image can be a three-channel RGB image. The target exposed back image can refer to an optical image of the exposed back of the target object in a natural standing position. In this embodiment, a target exposed back image is first obtained by optical imaging of the exposed back of the target object in a natural standing position using a visible light camera, which serves as the base image for measuring the target object's Cobb angle.

[0033] S120: Input the target bare back image into the pre-trained spine curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix.

[0034] The spinal curve detection model refers to a neural network model capable of detecting spinal curves. This model takes a bare back image as input and outputs a curve coordinate probability matrix and a curve start-end point probability matrix. The model output serves as an important basis for spinal curve detection. Specifically, the curve coordinate probability matrix describes the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start-end point probability matrix describes the row coordinate probabilities of the start and end points of the spinal curve. The start point is the intersection of the line connecting the shoulders and the spinal curve, and the end point is the intersection of the line connecting the hips and the spinal curve. The positions of the shoulders and hips are defined with reference to the corresponding shoulder and hip keypoints in the COCO Keypoints public human pose estimation dataset.

[0035] It should be noted that this embodiment describes the pixel coordinates in the image based on the image coordinate system. The image coordinate system has its origin at the top left corner of the image, with vertical coordinates defined as row coordinates and horizontal coordinates defined as column coordinates.

[0036] In this embodiment, a spinal curve detection model needs to be pre-trained using a neural network model. Optionally, the training process of the spinal curve detection model includes: acquiring an optical image of the bare back of a reference object in a natural standing state as a reference bare back image; acquiring first label information of the reference bare back image, and performing one-hot encoding on the first label information to obtain second label information; wherein, the first label information includes curve coordinate labels and curve start and end point row coordinate labels, and the second label information includes a curve coordinate label matrix and a curve start and end point label matrix; supervising the training of the neural network model based on the reference bare back image and its corresponding first and second label information, and determining the trained neural network model as the spinal curve detection model.

[0037] Here, the reference object can refer to the object used in model training. To ensure the accuracy and generalization of model training, multiple reference objects are usually selected. The reference bare back image can refer to the optical image of the bare back of the reference object in a natural standing position.

[0038] Specifically, the first step is to acquire a three-channel RGB image of the exposed back of a reference subject in a natural standing position using a visible light camera. As a reference image of a bare back, among which, Indicates the image height. This represents the image width. Then, the spine curve of the reference bare back image is manually annotated to obtain the first label information, specifically including curve coordinate labels. and the coordinate labels of the start and end points of the curve .in, This represents the coordinates of discrete points on the spine curve in a reference bare back image. This represents the row coordinates of the starting point of the spine curve in the reference naked back image. This represents the row coordinates of the endpoint of the spine curve in the reference naked back image. Furthermore, the curve coordinate labels are... and the coordinate labels of the start and end points of the curve One-hot encoding is performed separately to obtain the corresponding curve coordinate label matrix. and curve start and end point label matrix This serves as the second label information for the reference bare back image. A training dataset is then constructed based on the reference bare back image and its corresponding first and second label information. The neural network model is then trained under supervision using this training dataset. Once training is complete, the spine curve detection model is obtained.

[0039] In this embodiment, optionally, supervised training of the neural network model is performed based on a reference bare back image and its corresponding first and second label information, including: inputting the reference bare back image into the neural network model to obtain a reference curve coordinate probability matrix and a reference curve start and end point probability matrix; determining the reference spine curve corresponding to the reference bare back image based on the expectation of the reference curve coordinate probability matrix, and determining the reference curve start and end point row coordinates corresponding to the reference bare back image based on the expectation of the reference curve start and end point probability matrix; determining the overall loss function based on the reference spine curve, the reference curve start and end point row coordinates, the reference curve coordinate probability matrix, the reference curve start and end point probability matrix, the first label information, and the second label information, and supervised training of the neural network model based on the overall loss function.

[0040] Specifically, the reference bare back image is first input into the neural network model to obtain the reference curve coordinate probability matrix. and the probability matrix of the start and end points of the reference curve Optionally, the spine curve detection model uses the Softmax function as the activation function, that is, the neural network model uses the Softmax function as the activation function to ensure the model output... The sum of the probabilities of each row is 1 (i.e.) ), and the model output The sum of the probabilities in each column is 1 (i.e.) ),in This represents the row and column coordinate indices in the probability matrix. Then, the reference spine curve corresponding to the reference bare back image can be determined based on the expectation of the reference curve coordinate probability matrix, specifically as follows: Simultaneously, the row coordinates of the reference curve's start and end points corresponding to the reference bare back image can be determined based on the expected value of the reference curve's start and end point probability matrix, specifically expressed as follows: , .in, This indicates the row coordinates of the starting point of the spinal curve predicted by the model corresponding to the reference bare back image. This indicates the coordinates of the endpoint row of the spinal curve predicted by the model corresponding to the reference bare back image.

[0041] Then, it can be based on the reference spinal curve Reference curve start and end point line coordinates ( ), reference curve coordinate probability matrix Reference curve start and end point probability matrix First tag information ( and ) and second tag information ( and The overall loss function is determined, and the neural network model is trained under supervision based on the overall loss function. Optionally, the overall loss function is determined based on the reference spine curve, the row coordinates of the reference curve's start and end points, the reference curve's coordinate probability matrix, the reference curve's start and end point probability matrix, the first label information, and the second label information. This includes: determining the expected position loss function based on the reference spine curve, the row coordinates of the reference curve's start and end points, and the first label information; determining the probability distribution loss function based on the reference curve's coordinate probability matrix, the reference curve's start and end point probability matrix, and the second label information; determining the first derivative loss function based on the curve coordinate labels in the reference spine curve and the first label information; and obtaining the overall loss function by weighted summation of the expected position loss function, the probability distribution loss function, and the first derivative loss function.

[0042] For example, the expected loss function corresponding to the reference bare back image can be determined by calculating the L1 loss, specifically expressed as: Furthermore, the probability distribution loss function corresponding to the reference bare-back image can be determined by calculating the cross-entropy loss, specifically expressed as: .in, and Represents the first element in the corresponding matrix. The vector of rows, and Represents the first element in the corresponding matrix. A vector of columns. The smoothness of the curve can also be predicted by constructing a supervised model using a first-order derivative loss function, which can be achieved through the difference of the curve's column coordinates, as shown below: Next, the position expectation loss function, probability distribution loss function, and first derivative loss function are weighted and summed, which can be expressed as follows: This serves as the overall loss function for model training.

[0043] The neural network model is then trained under supervised supervision based on the overall loss function, and it is determined whether the overall loss function meets the preset iteration conditions. For example, the preset iteration conditions can be set as the overall loss function value being less than a first preset threshold, or the number of model iterations being greater than a second preset threshold. When the overall loss function meets the preset conditions, it indicates that the model training is complete. At this point, the model training process ends, and the neural network model at the end of training is used as the spine curve detection model.

[0044] In this embodiment, after obtaining the target bare back image, the target bare back image can be input into a pre-trained spine curve detection model, and the curve coordinate probability matrix output by the model (see...) can be used to... Figure 2 The coordinate probability matrix in the model is used as the target curve coordinate probability matrix, and the curve start and end point probability matrix output by the model (see...) is used as the target curve coordinate probability matrix. Figure 2 The probability matrix of the start and end points in the target curve is used as the probability matrix of the start and end points of the target curve.

[0045] S130, determine the target spine curve corresponding to the target bare back image based on the expectation of the target curve coordinate probability matrix, and determine the row coordinates of the target curve start and end points corresponding to the target bare back image based on the expectation of the target curve start and end point probability matrix.

[0046] In this embodiment, as Figure 2 As shown, after obtaining the target curve coordinate probability matrix and the target curve start and end point probability matrix, the target spine curve corresponding to the target bare back image can be determined according to the expectation of the target curve coordinate probability matrix. The specific implementation process can refer to the determination process of the reference spine curve. At the same time, the row coordinates of the target curve start and end points corresponding to the target bare back image can be determined according to the expectation of the target curve start and end point probability matrix. The specific implementation process can refer to the determination process of the row coordinates of the reference curve start and end points, which will not be elaborated here.

[0047] S140, based on the row coordinates of the start and end points of the target curve, determine multiple target row coordinates on the target spine curve, and determine the Cobb angle of the target object according to the target curvature corresponding to the adjacent target row coordinates.

[0048] In this embodiment, as Figure 2 As shown, after determining the target spine curve and the row coordinates of the target curve's start and end points, the Cobb angle of the target object can be calculated. Specifically, firstly, multiple points on the target spine curve are sampled within the row coordinate range defined by the row coordinates of the target curve's start and end points, and the sampled row coordinates are used as the target row coordinates. For example, random sampling can be used to sample on the target spine curve. individual row coordinates and guarantee Then, calculate the target curvature corresponding to every two adjacent target row coordinates, which can be specifically expressed as: .in, This represents the target spine curve.

[0049] Furthermore, a maximum value is determined from all target curvatures as the maximum curvature. Simultaneously, a minimum value is determined from all target curvatures as the minimum curvature. Then, based on the maximum and minimum curvature, the Cobb angle of the target object is calculated, which can be specifically expressed as: .

[0050] The technical solution of this invention first acquires an optical image of the bare back of a target object in a natural standing state as the target bare back image; then, the target bare back image is input into a pre-trained spinal curve detection model to obtain a target curve coordinate probability matrix and a target curve start-end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start-end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, with the start point being the intersection of the shoulder line and the spinal curve, and the end point being the intersection of the hip line and the spinal curve; then, the target spinal curve corresponding to the target bare back image is determined according to the expectation of the target curve coordinate probability matrix, and the row coordinates of the target curve start-end point corresponding to the target bare back image are determined according to the expectation of the target curve start-end point probability matrix; furthermore, multiple target row coordinates are determined on the target spinal curve based on the row coordinates of the target curve start-end point, and the Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates. This technical solution utilizes a neural network model to detect the spinal curve in bare back images and determines the Cobb angle based on the curve detection results. It does not rely on special data acquisition equipment and avoids the risk of ionizing radiation exposure. It can improve the accuracy, safety, convenience and adaptability of Cobb angle measurement based on optical image detection.

[0051] Example 2

[0052] Figure 3 This is a flowchart of a Cobb angle measurement method based on spinal curve detection provided in Embodiment 2 of the present invention. This embodiment is an optimization based on the above embodiment. Specifically, the optimization is as follows: multiple target row coordinates are determined on the target spinal curve based on the row coordinates of the start and end points of the target curve; the Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates, including: determining the target sampling range based on the row coordinates of the start and end points of the target curve; uniformly sampling the target spinal curve within the target sampling range to obtain multiple target row coordinates; sequentially determining the target curvature corresponding to every two adjacent target row coordinates; and determining the Cobb angle of the target object based on the maximum and minimum values ​​of the target curvature.

[0053] like Figure 3 As shown, the method in this embodiment specifically includes the following steps:

[0054] S210, acquire an optical image of the bare back of the target object in a natural standing position as the target bare back image.

[0055] S220, input the target bare back image into the pre-trained spine curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix.

[0056] Among them, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve. The start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve.

[0057] S230, determine the target spine curve corresponding to the target bare back image based on the expectation of the target curve coordinate probability matrix, and determine the row coordinates of the target curve start and end points corresponding to the target bare back image based on the expectation of the target curve start and end point probability matrix.

[0058] S240: Determine the target sampling range based on the row coordinates of the start and end points of the target curve, and uniformly sample the target spine curve within the target sampling range to obtain multiple target row coordinates.

[0059] For example, suppose the row coordinate of the starting point of the target curve is... The target curve endpoint row coordinates are At this point, the target sampling range can be expressed as In this embodiment, a uniform step size method is used to uniformly sample the target spine curve within the target sampling range to obtain multiple target row coordinates, thereby eliminating the influence of sampling step size differences on the accuracy of Cobb angle calculation and helping to further improve the accuracy of Cobb angle calculation.

[0060] S250, sequentially determine the target curvature corresponding to every two adjacent target row coordinates, and determine the Cobb angle of the target object based on the maximum and minimum values ​​of the target curvature.

[0061] It should be noted that, compared with existing Cobb angle measurement methods, this invention has the following advantages:

[0062] 1. Convenient Deployment and Use: The input image required by this invention is a bare back image, which can be acquired using a general-purpose visible light camera, without relying on special data acquisition equipment. Therefore, this invention can be deployed on easily accessible data acquisition devices such as mobile phone cameras, providing convenience for widespread adoption and application.

[0063] 2. Radiation-free screening: Current Cobb angle measurements rely on the imaging and analysis of X-ray films, which carries the risk of ionizing radiation exposure. The frequency of imaging, especially for adolescents, needs careful consideration. This invention only requires visible light images of the bare back for analysis, avoiding the risk of excessive ionizing radiation exposure. Therefore, it can be used repeatedly over a long period.

[0064] 3. Optimization of Annotation Difficulty and Measurement Accuracy: Compared to automated Cobb angle measurement using X-ray images, measuring Cobb angle in naked back images presents a significant technical challenge: annotators cannot directly observe the spinal structure, making it impossible to provide accurate vertebral position annotations. This hinders the application of mature X-ray image-based solutions, such as calculating Cobb angles by vertebral localization and segmentation. Existing optical imaging solutions often use keypoint detection for estimation, but a small number of keypoints makes accurate Cobb angle calculation difficult. This invention first transforms the problem from vertebral segmentation to curve detection and calculates Cobb angles based on curve curvature. Within this framework, annotators only need to draw the spinal curve, reducing annotation difficulty compared to vertebral segmentation solutions while improving accuracy compared to existing keypoint detection methods.

[0065] 4. Dense Probability-Based Curve Detection: To address the problem of spine curve detection in bare back images, this invention defines the model's prediction objective as predicting the probability that each pixel in the image belongs to the spine location, i.e., obtaining a dense probability map to predict curve coordinates. Specifically, a neural network model is used to estimate the probability of curve coordinates and the probability of the curve's start and end points, and then the expected position of the curve coordinates is calculated based on the probabilities. Experimental results on publicly available datasets show that this technical solution can obtain continuous and relatively accurate spine curve prediction results, thus providing strong support for downstream Cobb angle calculation.

[0066] 5. Intelligent Analysis: In manual Cobb angle calculations, variations in the selected vertebrae can lead to differences in the calculated Cobb angle, introducing errors due to subjective judgment. This invention utilizes an artificial intelligence model to achieve intelligent and automated analysis, providing stable and consistent results for the same image, thus enhancing objectivity. Furthermore, the model results offer good interpretability; by displaying the predicted spinal curve, users can confirm the Cobb angle calculation process and determine whether to accept the model's prediction, thereby improving the user experience.

[0067] The technical solution of this invention utilizes a neural network model to detect the spine curve in a bare back image and determines the Cobb angle based on the curve detection results. This method does not rely on special data acquisition equipment and avoids the risk of ionizing radiation exposure. It can improve the accuracy, safety, convenience, and adaptability of Cobb angle measurement based on optical image detection. At the same time, it uses a uniform step size method to uniformly sample the target spine curve within the target sampling range, which can effectively eliminate the influence of sampling step size differences on the accuracy of Cobb angle calculation and help to further improve the accuracy of Cobb angle calculation.

[0068] Example 3

[0069] Figure 4This is a schematic diagram of a Cobb angle measurement device based on spinal curve detection provided in Embodiment 3 of the present invention. This device can execute the Cobb angle measurement method based on spinal curve detection provided in any embodiment of the present invention, and possesses the corresponding functional modules and beneficial effects for executing the method. For example... Figure 4 As shown, the device includes:

[0070] Image acquisition module 310 is used to acquire an optical image of the bare back of the target object in a natural standing state as the target bare back image;

[0071] The model prediction module 320 is used to input the target bare back image into a pre-trained spinal curve detection model to obtain a target curve coordinate probability matrix and a target curve start and end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, wherein the start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve;

[0072] The curve estimation module 330 is used to determine the target spine curve corresponding to the target bare back image based on the expectation of the target curve coordinate probability matrix, and to determine the row coordinates of the target curve start and end points corresponding to the target bare back image based on the expectation of the target curve start and end point probability matrix.

[0073] The sampling calculation module 340 is used to determine multiple target row coordinates on the target spine curve based on the row coordinates of the start and end points of the target curve, and to determine the Cobb angle of the target object according to the target curvature corresponding to adjacent target row coordinates.

[0074] Optionally, the apparatus further includes: a model training module, used for:

[0075] Obtain an optical image of the bare back of the reference object in a natural standing position as a reference bare back image;

[0076] Obtain the first label information of the reference bare back image, and perform one-hot encoding on the first label information to obtain the second label information; wherein, the first label information includes curve coordinate labels and curve start and end point row coordinate labels, and the second label information includes curve coordinate label matrix and curve start and end point label matrix;

[0077] The neural network model is trained under supervision based on the reference bare back image and its corresponding first and second label information, and the trained neural network model is determined as the spine curve detection model.

[0078] Optionally, the model training module is further configured to:

[0079] The reference bare back image is input into a neural network model to obtain the reference curve coordinate probability matrix and the reference curve start and end point probability matrix;

[0080] The reference spine curve corresponding to the reference bare back image is determined based on the expectation of the reference curve coordinate probability matrix, and the row coordinates of the reference curve start and end points corresponding to the reference bare back image are determined based on the expectation of the reference curve start and end point probability matrix.

[0081] The overall loss function is determined based on the reference spine curve, the row coordinates of the reference curve start and end points, the reference curve coordinate probability matrix, the reference curve start and end point probability matrix, the first label information, and the second label information. The neural network model is then trained under supervision based on the overall loss function.

[0082] Optionally, the model training module is further configured to:

[0083] The position expected loss function is determined based on the reference spine curve, the row coordinates of the start and end points of the reference curve, and the first label information.

[0084] The probability distribution loss function is determined based on the reference curve coordinate probability matrix, the reference curve start and end point probability matrix, and the second label information.

[0085] The first derivative loss function is determined based on the reference spine curve and the curve coordinate labels in the first label information;

[0086] The overall loss function is obtained by weighted summation of the position expectation loss function, the probability distribution loss function, and the first derivative loss function.

[0087] Optionally, the position expectation loss function is determined based on L1 loss, and the probability distribution loss function is determined based on cross-entropy loss.

[0088] Optionally, the spinal curve detection model uses the Softmax function as the activation function.

[0089] Optionally, the sampling calculation module 340 is used for:

[0090] The target sampling range is determined based on the row coordinates of the start and end points of the target curve.

[0091] Multiple target row coordinates are obtained by uniformly sampling the target spine curve within the target sampling range;

[0092] The target curvature corresponding to every two adjacent target row coordinates is determined sequentially;

[0093] The Cobb angle of the target object is determined based on the maximum and minimum values ​​of the target curvature.

[0094] The Cobb angle measuring device based on spinal curve detection provided in this embodiment of the invention can execute the Cobb angle measuring method based on spinal curve detection provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0095] Example 4

[0096] Figure 5 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0097] like Figure 5 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0098] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0099] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the Cobb angle measurement method based on spine curve detection.

[0100] In some embodiments, the Cobb angle measurement method based on spinal curvature detection can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the Cobb angle measurement method based on spinal curvature detection described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the Cobb angle measurement method based on spinal curvature detection by any other suitable means (e.g., by means of firmware).

[0101] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0102] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0103] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0104] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0105] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0106] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0107] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0108] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for measuring Cobb angle based on spinal curve detection, characterized in that, The method includes: Obtain an optical image of the bare back of the target object in a natural standing position as the target bare back image; The target bare back image is input into a pre-trained spinal curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, wherein the start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve. The target spine curve corresponding to the target bare back image is determined based on the expectation of the target curve coordinate probability matrix, and the row coordinates of the target curve start and end points corresponding to the target bare back image are determined based on the expectation of the target curve start and end point probability matrix. Based on the row coordinates of the start and end points of the target curve, multiple target row coordinates are determined on the target spine curve, and the Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates.

2. The method according to claim 1, characterized in that, The training process of the spinal curve detection model includes: Obtain an optical image of the bare back of the reference object in a natural standing position as a reference bare back image; Obtain the first label information of the reference bare back image, and perform one-hot encoding on the first label information to obtain the second label information; wherein, the first label information includes curve coordinate labels and curve start and end point row coordinate labels, and the second label information includes curve coordinate label matrix and curve start and end point label matrix; The neural network model is trained under supervision based on the reference bare back image and its corresponding first and second label information, and the trained neural network model is determined as the spine curve detection model.

3. The method according to claim 2, characterized in that, Supervised training of the neural network model is performed based on the reference bare back image and its corresponding first and second label information, including: The reference bare back image is input into a neural network model to obtain the reference curve coordinate probability matrix and the reference curve start and end point probability matrix; The reference spine curve corresponding to the reference bare back image is determined based on the expectation of the reference curve coordinate probability matrix, and the row coordinates of the reference curve start and end points corresponding to the reference bare back image are determined based on the expectation of the reference curve start and end point probability matrix. The overall loss function is determined based on the reference spine curve, the row coordinates of the reference curve start and end points, the reference curve coordinate probability matrix, the reference curve start and end point probability matrix, the first label information, and the second label information. The neural network model is then trained under supervision based on the overall loss function.

4. The method according to claim 3, characterized in that, The overall loss function is determined based on the reference spine curve, the row coordinates of the reference curve's start and end points, the reference curve's coordinate probability matrix, the reference curve's start and end point probability matrix, the first label information, and the second label information, including: The position expected loss function is determined based on the reference spine curve, the row coordinates of the start and end points of the reference curve, and the first label information. The probability distribution loss function is determined based on the reference curve coordinate probability matrix, the reference curve start and end point probability matrix, and the second label information. The first derivative loss function is determined based on the reference spine curve and the curve coordinate labels in the first label information; The overall loss function is obtained by weighted summation of the position expectation loss function, the probability distribution loss function, and the first derivative loss function.

5. The method according to claim 4, characterized in that, The position expectation loss function is determined based on L1 loss, and the probability distribution loss function is determined based on cross-entropy loss.

6. The method according to any one of claims 1-4, characterized in that, The spinal curve detection model uses the Softmax function as the activation function.

7. The method according to claim 1, characterized in that, Based on the row coordinates of the start and end points of the target curve, multiple target row coordinates are determined on the target spine curve. The Cobb angle of the target object is determined according to the target curvature corresponding to adjacent target row coordinates, including: The target sampling range is determined based on the row coordinates of the start and end points of the target curve. Multiple target row coordinates are obtained by uniformly sampling the target spine curve within the target sampling range; The target curvature corresponding to every two adjacent target row coordinates is determined sequentially; The Cobb angle of the target object is determined based on the maximum and minimum values ​​of the target curvature.

8. A Cobb angle measuring device based on spinal curve detection, characterized in that, The device includes: The image acquisition module is used to acquire an optical image of the bare back of the target object in a natural standing position as the target bare back image; The model prediction module is used to input the target bare back image into a pre-trained spinal curve detection model to obtain the target curve coordinate probability matrix and the target curve start and end point probability matrix; wherein, the curve coordinate probability matrix is ​​used to describe the probability that each pixel in the bare back image belongs to the spinal curve, and the curve start and end point probability matrix is ​​used to describe the row coordinate probability of the start and end points of the spinal curve, wherein the start point is the intersection of the shoulder line and the spinal curve, and the end point is the intersection of the hip line and the spinal curve; The curve estimation module is used to determine the target spine curve corresponding to the target bare back image based on the expectation of the target curve coordinate probability matrix, and to determine the row coordinates of the target curve start and end points corresponding to the target bare back image based on the expectation of the target curve start and end point probability matrix. The sampling calculation module is used to determine multiple target row coordinates on the target spine curve based on the row coordinates of the start and end points of the target curve, and to determine the Cobb angle of the target object according to the target curvature corresponding to adjacent target row coordinates.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the Cobb angle measurement method based on spinal curve detection as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the Cobb angle measurement method based on spinal curve detection as described in any one of claims 1-7.

Images