Scoliosis back image data classification method and system based on unified pinball loss

Abstract

This invention discloses a method, system, and medium for classifying scoliosis back image data based on a unified bouncing loss. The method includes: acquiring basic information about the patient's back and scoliosis back image data; preprocessing the scoliosis back image data to obtain preprocessed data, and extracting back spinal image features from the preprocessed data; constructing and training a scoliosis classification model based on a unified bouncing loss based on the preprocessed data, optimizing the model parameters, generating a pair of non-parallel separating hyperplanes, and obtaining hyperplane parameters; using the hyperplane parameters and new back spinal image features to make a classification decision, and obtaining a classification result; and outputting the classification result. This method optimizes the scoliosis classification model based on the unified bouncing loss function, enabling the model to classify and predict more accurately.

Description

Technical Field

[0001] This invention relates to the field of medical image processing and pattern recognition technology, specifically to a method, system, and medium for classifying scoliosis back image data based on uniform bouncing loss. Background Technology

[0002] Scoliosis is a complex three-dimensional spinal deformity and a common spinal disease. If it is not intervened and corrected in time, it may lead to serious health problems, including respiratory dysfunction, spinal pain, and limited motor function.

[0003] The cause of scoliosis is unknown. Without early prevention and treatment, it carries the risk of serious complications. As patients grow and develop, they are prone to chronic back pain, leading to more noticeable changes in the back, including asymmetry in the hips and shoulders, prominent ribs, and lateral tilting of the lower back and trunk. More seriously, the ribs may compress the lungs, causing breathing difficulties. While early etiological studies have shown that patients with a history of conditions such as cerebral palsy and muscular dystrophy are more likely to develop scoliosis than others, the cause of scoliosis in most adolescents remains unknown.

[0004] Currently, common methods for detecting scoliosis include:

[0005] Visual assessment: Doctors observe the shape of the back to determine if the shoulders are asymmetrical or if the spine deviates from its normal axis. This method relies on the doctor's experience, is prone to misjudgment, and is highly subjective.

[0006] Cobb angle measurement via X-ray: This is a commonly used diagnostic method in clinical practice. The patient must maintain a fixed posture, and the doctor measures the Cobb angle using X-ray images to determine the severity of scoliosis. While this method is accurate, frequent and prolonged use of X-rays is detrimental to health and is not suitable for large-scale screening.

[0007] Traditional scoliosis screening methods primarily rely on physical examinations and X-ray imaging. While these methods are routinely used in clinical practice, they have several limitations. First, traditional physical examinations, including simple physical checks and measurements, are non-invasive and safe, but their positive predictive value is low, often leading to unnecessary referrals and further imaging examinations. While X-rays are the gold standard, frequent radiation exposure poses potential health risks and is particularly unsuitable for long-term monitoring. Furthermore, existing technologies, such as the X-ray image-based scoliosis detection method proposed in Chinese patent "CN201210441954.9," which involves identifying the curvature region of the spine in the image using deep learning and calculating the Cobb angle, still rely on X-ray imaging, making it difficult to address the issue of radiation exposure.

[0008] The above screening methods are not only costly and inefficient, but also difficult to be widely applied to the initial screening of large populations.

[0009] With the rapid development of computer vision, artificial intelligence, and machine learning technologies, automated image classification methods are increasingly being applied to the diagnosis and screening of scoliosis. These methods offer advantages such as non-invasiveness, low cost, and no need for radiation exposure, reducing human error and improving screening efficiency. However, existing automated methods still face numerous challenges when dealing with large-scale data, particularly when encountering noisy or outlier data, which can affect the model's accuracy and robustness. Therefore, there is an urgent need for an algorithm that can improve the accuracy and efficiency of scoliosis image classification to address the challenges of noise and large-scale data. Summary of the Invention

[0010] To address the shortcomings of existing technologies, this invention provides a method and system for classifying scoliosis back image data based on uniform ball loss, thus solving the technical problems existing in the prior art.

[0011] In a first aspect, the present invention provides a method for classifying scoliosis back image data based on uniform ball loss, comprising the following steps:

[0012] Acquire basic information about the patient's back and image data of the back with scoliosis;

[0013] The back image data of the scoliosis is preprocessed to obtain preprocessed data, and back spinal image features are extracted from the preprocessed data.

[0014] Based on the preprocessed data, a scoliosis classification model based on unified bouncing loss is constructed and trained. The model parameters are optimized, and the model generates a pair of non-parallel separating hyperplanes, from which the hyperplane parameters are obtained.

[0015] The classification decision is made using the hyperplane parameters and the new back spine image features to obtain the classification result.

[0016] Output the classification results.

[0017] Secondly, embodiments of the present invention provide a classification system for scoliosis back image data based on unified ball loss, including: an image acquisition module, a data processing module, a model training module, a classification decision module and an output module, wherein the image acquisition module is used to acquire basic information of the patient's back and scoliosis back image data;

[0018] The data processing module is used to preprocess the scoliosis back image data to obtain preprocessed data, and extract back spine image features from the preprocessed data.

[0019] The model training module is used to construct and train a scoliosis classification model based on unified bouncing loss based on preprocessed data, optimize model parameters, and generate a pair of non-parallel separating hyperplanes to obtain hyperplane parameters.

[0020] The classification decision module uses hyperplane parameters and novel back spine image features to make classification decisions and obtain classification results.

[0021] The output module is used to output the classification results.

[0022] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the method described in the above embodiments.

[0023] The beneficial effects of this invention are:

[0024] This invention provides a method, system, and medium for classifying scoliosis back image data based on a unified bouncing loss. It utilizes machine learning methods to map data from patient back images using a Gaussian kernel function, ultimately constructing a scoliosis classification model based on a unified bouncing loss function for classifying the degree of scoliosis. Furthermore, the process of minimizing the bouncing loss is added to reduce misclassification, thereby optimizing the model and enabling more accurate classification and prediction. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0026] Figure 1 The flowchart of a traditional machine learning algorithm for classification is shown;

[0027] Figure 2 A flowchart of a method for classifying scoliosis back image data based on uniform bouncing loss, provided in the first embodiment of the present invention, is shown.

[0028] Figure 3 The composition diagram of UPT BLDC is shown;

[0029] Figure 4 A flowchart illustrating the workflow of the unified bouncing loss scoliosis classification model in the first embodiment of the present invention is shown.

[0030] Figure 5This diagram illustrates a structural block diagram of a scoliosis back image data classification system based on unified ball loss, according to another embodiment of the present invention.

[0031] Figure 6 A timing diagram of a back image data classification system for scoliosis based on uniform bouncing loss, according to another embodiment of the present invention, is shown. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0034] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0035] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0036] like Figure 1 As shown, a flowchart of a traditional machine learning algorithm for image classification is presented. When performing image classification, the Support Vector Machine does not take the original image as input but first performs a feature extraction step.

[0037] The parameters involved in the methods of this invention are all determined by cross-validation, rather than by fixed values.

[0038] like Figure 2 As shown, the first embodiment of the present invention provides a method for classifying scoliosis back image data based on uniform bouncing loss, comprising:

[0039] S1. Obtain basic information about the patient's back and image data of the back of the patient with scoliosis.

[0040] In this embodiment, the collected scoliosis back image data is publicly available scoliosis back image data on Kaggle. The test set consists of 16 scoliosis images and 20 non-scoliosis images, while the training set consists of 120 scoliosis images and 103 non-scoliosis images. To improve the accuracy of the test set, this embodiment of the invention uses k-fold cross-validation and conducts multiple experiments to test the prediction accuracy.

[0041] S2. Acquire image data of the back of the scoliosis and perform preprocessing to obtain preprocessed data. Extract back spinal image features from the preprocessed data.

[0042] In this embodiment, the preprocessing includes: image segmentation, grayscale conversion, filtering and denoising, size standardization, and image normalization. Finally, the image region is cropped to a 224×224 format to improve the accuracy of subsequent machine learning processing. The dataset is divided into a 75% training set and a 25% test set.

[0043] Specifically, the back image is segmented to extract the region of interest (ROI), namely the spine region.

[0044] Local Binary Method (LBP) is used to improve texture representation. The LBP operator assigns a value to a pixel in the image based on whether it is greater than or equal to the center pixel. If the value of S is greater than or equal to the center pixel at that location, the pixel is set to 1; otherwise, it is set to 0. Subsequently, LBP codes for the neighborhood are created using P=8 and R=1.

[0045] x i Let S represent the position of the i-th adjacent pixel, and Z represent the pixel intensity. Where S = Z(X... i The pixel allocation principle of the LBP method is as follows: -Z(X)

[0046] according to Normalization is performed to eliminate the influence of factors such as brightness.

[0047] S3. Based on the preprocessed data, establish a scoliosis classification model based on the unified bouncing loss and train the model to solve the optimization problem of maximizing the classification margin between samples with different degrees of scoliosis and minimizing the unified bouncing loss function to reduce the impact of misclassification.

[0048] like Figure 3 As shown, a diagram of the UPT BLDC composition is presented.

[0049] Parameter initialization settings: The parameter σ of the Gaussian kernel function is set using the standard deviation of the sample distances in the dataset; the penalty parameter c is initialized to 1; the tradeoff parameter λ is initially set to 0.5, similar to the slack variable ξ in the soft-margin model, which is initially chosen to be 1 to control the penalty ratio; the parameter τ, which determines the penalty ratio, is initialized to 0.5. It should be noted that the values ​​of the above parameters need to be adjusted through cross-validation.

[0050] Constructing a scoliosis classification model based on a unified ball-drop loss, specifically including:

[0051] The unified bouncing loss function is defined as:

[0052]

[0053] Where y is the true label of the sample, f(x) is the predicted output of the model, τ is a parameter controlling the smoothness of the bouncing loss, and τ takes values ​​in the interval [-1,0)∪(0,1], u=1-yf ω,b (x) represents the distance of a sample point relative to its corresponding class hyperplane.

[0054] Reference for process steps in specific applications Figure 4 , Figure 4 This is a flowchart illustrating the workflow of a scoliosis classification model based on the unified bouncing loss. The unified bouncing loss aims to address the problem of difficult-to-adjust hyperparameters in machine learning, which negatively impact classification performance.

[0055] Further introducing the concept of twins, this study seeks a pair of non-parallel hyperplane decision functions f based on a unified ball-drop loss scoliosis classification model. + (x) and f - (x),

[0056]

[0057] Among them, f + (x) and f - (x) represent the decision functions of two non-parallel hyperplanes, respectively. and These are the weight parameters of the classification hyperplane; It is a Gaussian kernel function used to map the features of input scoliosis image data to a high-dimensional space.

[0058] The distance from a single data sample to the hyperplane is calculated as follows:

[0059]

[0060] in, Let represent the distance from the i-th positive class sample to the hyperplane. Let l represent the distance from the i-th negative class sample to the hyperplane, and l1 and l2 represent the number of positive and negative class samples, respectively. and Let i represent the i-th positive and negative samples, where the positive sample is... This category belongs to the more severe type of scoliosis, and is a negative sample. This falls into the category of mild or normal scoliosis; and This represents the true class label of the i-th sample.

[0061] In the calculation of the interval mean and interval variance based on the unified bouncing loss scoliosis classification model, the interval mean and interval variance are divided into positive and negative classes. The calculation references the distance to the other class of hyperplanes. The relationship between the interval mean and interval variance is calculated using the following formula:

[0062]

[0063] Considering the relative distances between hyperplanes, the distance from a single sample to a hyperplane can be expressed in matrix form:

[0064]

[0065] and This represents the mapping function in the high-dimensional feature space. This approach can better distinguish different categories of scoliosis and improve the performance of the classification model. Q1 and Q2 are symmetric matrices, and their calculation formulas are as follows:

[0066]

[0067] Based on the unified bouncing loss scoliosis classification model, a pair of unknown vectors ω is sought by simultaneously maximizing the mean of the positive and negative intervals and minimizing the variance of the positive and negative intervals. + ,ω - That is, consider the following two optimization problems:

[0068]

[0069] Where τ∈[-1,1], λ1, λ2, λ3, λ4 are the trade-off parameters between interval variance and interval mean, λ1,λ2,λ3,λ4>0, and ||ω|| - || 2 c1 and c2 are the regularization terms, c3 and c4 are the hyperparameters corresponding to the regularization terms, and c3 and c4 are parameters that weigh the total error. and It is the interval mean. and It is the interval variance.

[0070] Based on this, let the slack variable ξi Satisfy ζ i =∑ i L τ (1-y i (ω T φ(x i ))), where L τ Let τ be the bouncing loss function, and τ be a parameter that determines the penalty ratio.

[0071]

[0072] Based on the constraints, the optimal classification parameters are solved. This invention leverages the symmetry of the unified bouncing ball loss scoliosis classification model, which addresses two pairs of small-scale QPP problems. The optimal hyperplane parameters corresponding to the scoliosis image samples are solved based on the constraints. Furthermore, the optimal solution is re-expressed as w. + =φ(X)η1 and w - =φ(X)η2, consider and The resulting constraints are as follows:

[0073]

[0074] Further derivation and substitution yields...

[0075]

[0076] By introducing Lagrange multiplier vectors α and β, and considering the classification requirements of scoliosis image data, the optimization objective of the classifier is transformed into a Lagrange form:

[0077]

[0078] in Let be the Lagrange multiplier vector, let It can be obtained

[0079] Convert to Wolfe dual form

[0080]

[0081] st-τc3e2≤α1-β1≤c3e2,

[0082]

[0083] st-τc4e1≤α2-β2≤c4e1,,

[0084] in Let Lagrange multiplier vector be the vector.

[0085]

[0086] S4. Based on the calculated hyperplane parameters and the feature vector of the new back image data, a classification decision is made to obtain the classification result. Specifically, classification is performed by calculating the relative positions of the hyperplane parameters and the feature vector of the new back image data.

[0087] The Wolfe dual problem and the parameter vectors η1 and η2 associated with the optimal hyperplane are derived. A new input vector is then obtained. The class to which a sample belongs depends on its distance from two non-parallel hyperplanes. The class to which the hyperplane is closer belongs to the input sample.

[0088] The final decision function of the scoliosis classification model based on unified bouncing loss is expressed as:

[0089] Among them, the kernel matrix The similarity between the input sample x and the training sample X is represented by a Gaussian kernel function. These kernel values, combined with the dual parameter η, are used for the final classification decision to determine the degree of scoliosis in the patient.

[0090] S5. Output the classification results for doctors' reference.

[0091] This invention provides a method for classifying scoliosis back image data based on unified ball loss. It utilizes machine learning methods to map data from patient back images using a Gaussian kernel function, ultimately constructing a scoliosis classification model based on unified ball loss for classifying the degree of scoliosis. Furthermore, the method incorporates a process of minimizing ball loss to reduce the impact of misclassification, thereby optimizing the model and enabling it to classify and predict more accurately.

[0092] Example 2

[0093] like Figure 5 As shown in another embodiment of the present invention, a scoliosis back image data classification system based on unified ball loss is provided, comprising: an image acquisition module, a data processing module, a model training module, a classification decision module, and an output module. The image acquisition module is used to acquire basic information of the patient's back and scoliosis back image data; the data processing module is used to preprocess the scoliosis back image data to obtain preprocessed data, and extract back spinal image features from the preprocessed data; the model training module is used to construct and train a scoliosis classification model based on unified ball loss according to the preprocessed data, optimize the model parameters, and generate a pair of non-parallel separating hyperplanes to obtain hyperplane parameters; the classification decision module uses the hyperplane parameters and the new back spinal image features to make a classification decision and obtain a classification result; the output module is used to output the classification result.

[0094] Figure 6 This is a time-series diagram of a scoliosis back image data classification system based on a unified bouncing loss provided by this invention. The image acquisition module allows users to automatically or manually upload back images, which are then input into the data processing module. The data processing module performs preprocessing operations on the images, including image segmentation, grayscale conversion, filtering and denoising, size standardization, and image normalization. Back spine image features, including texture, shape, and statistical features, are extracted from the preprocessed data, and the dataset is divided into a 75% training set and a 25% test set. Key regions are identified through ROI detection, and the image data is cropped to 224×224 pixels. Local Binary Analysis (LBP) is used to improve texture features and avoid brightness interference. The model training module constructs a scoliosis classification model based on the unified bouncing loss. The unified bouncing loss function updates the parameters of the twin large-margin distributed machine model, and the distance from sample points to the hyperplane is calculated to obtain the hyperplane parameters, achieving optimal classification performance. The classification decision module classifies the scoliosis image into "scoliosis" and "non-scoliosis" based on the final scoliosis image classification decision function.

[0095] This invention provides a scoliosis back image data classification system based on unified ball loss. It utilizes machine learning methods to map data from patient back images using a Gaussian kernel function, ultimately constructing a scoliosis classification model based on unified ball loss for classifying the degree of scoliosis. Furthermore, the process of minimizing ball loss is added to reduce the impact of misclassification, thereby optimizing the model and enabling it to classify and predict more accurately.

[0096] Example 3

[0097] In another embodiment of the present invention, a computer-readable storage medium is also provided, which stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method described in the first embodiment above.

[0098] The computer-readable storage medium can be an internal storage unit of the terminal described in the foregoing embodiments, such as the terminal's hard drive or memory. The computer-readable storage medium can also be an external storage device of the terminal, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the terminal. The computer-readable storage medium is used to store the computer program and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0100] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the terminals and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, or may be electrical, mechanical or other forms of connection.

[0102] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for classifying scoliosis back image data based on uniform bouncing loss, characterized in that, Includes the following steps: Acquire basic information about the patient's back and image data of the back with scoliosis; The back image data of the scoliosis is preprocessed to obtain preprocessed data, and back spinal image features are extracted from the preprocessed data. Based on the preprocessed data, a scoliosis classification model based on unified bouncing loss is constructed and trained. The model parameters are optimized, and the model generates a pair of non-parallel separating hyperplanes, from which the hyperplane parameters are obtained. The classification decision is made using the hyperplane parameters and the new back spine image features to obtain the classification result. Output the classification results; The specific steps of constructing and training a scoliosis classification model based on unified bouncing loss using preprocessed data include: The unified bouncing loss function is defined as: Where y is the true label of the sample, f(x) is the predicted output of the model, τ is a parameter controlling the smoothness of the bouncing loss, and τ takes values ​​in the interval [-1,0)∪(0,1], u=1-yf ω,b (x) represents the distance of a sample point relative to its corresponding class hyperplane; Based on the unified bouncing loss scoliosis classification model, a pair of non-parallel hyperplane decision functions f are identified. + (x) and f - (x); Among them, f + (x) and f - (x) represent the decision functions of two non-parallel hyperplanes, respectively. and These are the weight parameters of the classification hyperplane; It is a Gaussian kernel function used to map the features of an input scoliosis image to a high-dimensional space; The distance from a single data sample to the hyperplane is calculated as follows: in, Let represent the distance from the i-th positive class sample to the hyperplane. Let l represent the distance from the i-th negative class sample to the hyperplane, and l1 and l2 represent the number of positive and negative class samples, respectively. and Let i represent the i-th positive and negative samples, where the positive sample is... This category belongs to the more severe type of scoliosis, and is a negative sample. This falls into the category of mild or normal scoliosis; and This represents the true class label of the i-th sample.

2. The method according to claim 1, characterized in that, The method for preprocessing the back image data specifically includes: image segmentation, grayscale conversion, filtering and noise reduction, size standardization, and image normalization.

3. The method according to claim 1, characterized in that, The back spine image features include texture features, shape features, and statistical features.

4. The method according to claim 1, characterized in that, The optimized model parameters include: solving for maximizing the classification margin between samples with different degrees of scoliosis and minimizing the uniform bouncing loss function, specifically: A pair of unknown vectors ω is calculated by simultaneously maximizing the mean of the positive and negative intervals and minimizing the variance of the positive and negative intervals. + ,ω - The calculation formula is: Where τ∈[-1,1], λ1,λ2,λ3,λ4 are the trade-off parameters between interval variance and interval mean, λ1,λ2,λ3,λ4>0, and ||ω|| - || 2 c1 and c2 are the regularization terms, c3 and c4 are the hyperparameters corresponding to the regularization terms, and c3 and c4 are parameters that weigh the total error. and It is the interval mean. and It is the interval variance; The margin mean and margin variance of the Siamese large-margin classifier based on the unified bouncing loss function are calculated separately for the positive and negative classes. The calculation references the distance to the other class's hyperplane. The relationship between the margin mean and margin variance is calculated using the following formula:

5. The method according to claim 4, characterized in that, The specific method for classification decision-making includes: calculating the relative positional relationship between the new back spine image features and hyperplane parameters, and classifying based on the relative positional relationship.

6. A classification system for scoliosis back image data based on uniform bouncing loss, characterized in that, include: The system includes an image acquisition module, a data processing module, a model training module, a classification decision module, and an output module. The image acquisition module is used to acquire basic information about the patient's back and image data of the back of the patient with scoliosis. The data processing module is used to preprocess the scoliosis back image data to obtain preprocessed data, and extract back spine image features from the preprocessed data. The model training module is used to construct and train a scoliosis classification model based on unified bouncing loss based on preprocessed data, optimize model parameters, and generate a pair of non-parallel separating hyperplanes to obtain hyperplane parameters. The classification decision module uses hyperplane parameters and novel back spine image features to make classification decisions and obtain classification results. The output module is used to output the classification results; The model training module, which constructs and trains a scoliosis classification model based on unified bouncing loss using preprocessed data, specifically includes: The unified bouncing loss function is defined as: Where y is the true label of the sample, f(x) is the predicted output of the model, τ is a parameter controlling the smoothness of the bouncing loss, and τ takes values ​​in the interval [-1,0)∪(0,1], u=1-yf ω,b (x) represents the distance of a sample point relative to its corresponding class hyperplane; Based on the unified bouncing loss scoliosis classification model, a pair of non-parallel hyperplane decision functions f are identified. + (x) and f - (x); Among them, f + (x) and f - (x) represent the decision functions of two non-parallel hyperplanes, respectively. and These are the weight parameters of the classification hyperplane; It is a Gaussian kernel function used to map the features of an input scoliosis image to a high-dimensional space; The distance from a single data sample to the hyperplane is calculated as follows: in, Let represent the distance from the i-th positive class sample to the hyperplane. Let l represent the distance from the i-th negative class sample to the hyperplane, and l1 and l2 represent the number of positive and negative class samples, respectively. and Let i represent the i-th positive and negative samples, where the positive sample is... This category belongs to the more severe type of scoliosis, and is a negative sample. This falls into the category of mild or normal scoliosis; and This represents the true class label of the i-th sample.

7. The system according to claim 6, characterized in that, The preprocessing includes image segmentation, grayscale conversion, filtering and noise reduction, size standardization, and image normalization.

8. The system according to claim 6, characterized in that, The back spine image features include texture features, shape features, and statistical features.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-5.

Images