A method, system, computer device and medium for predicting osteoarthritis

Abstract

The present invention provides an osteoarthritis disease prediction method, system, computer device and medium, the method comprising obtaining an osteoarthritis data set, and dividing the osteoarthritis data set into a training set, a validation set and a test set according to a preset ratio; the osteoarthritis data set comprises clinical sample data and x-ray joint space width data; according to the training set and the validation set, the optimal osteoarthritis disease prediction model is determined from different machine learning models; according to the test set and the optimal osteoarthritis disease prediction model, the corresponding osteoarthritis disease prediction results and influencing factors are obtained. While making up for the deficiencies of the prior art and providing convenience for clinicians to evaluate and diagnose the condition of osteoarthritis patients, the present invention effectively improves the accuracy of the prediction results, and further enhances the application value of osteoarthritis disease prediction, treatment guidance and treatment effect tracking.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to an osteoarthritis disease prediction method, system, computer equipment and storage medium. Background Art

[0002] Osteoarthritis is the most common form of arthritis. According to rough statistics, one-third of people over the age of 65 suffer from this disease. Studies have shown that the prevalence of osteoarthritis is gradually increasing with the increase of factors that cause osteoarthritis, such as obesity, lack of exercise or joint injuries. Patients with osteoarthritis often experience joint pain, which in turn leads to symptoms such as limited body movement function, lack of sleep, fatigue, depression and depression. As a result, the prediction and evaluation of osteoarthritis has become a research direction that many scholars are keen on.

[0003] The existing bone and joint disease prediction methods are mainly divided into statistical methods and model prediction. However, due to its own application limitations, the traditional statistical learning method cannot effectively process a large amount of data information, and cannot explore the potential internal correlation between the data, that is, it is difficult to give the contribution of different features to the assessment of the severity of osteoarthritis. In addition, although the existing model prediction method solves the shortcomings of traditional statistical methods to a certain extent, when it is used for prediction, it not only needs to be based on a large amount of high-quality real clinical data, but also requires that these data contain as many relevant factors as possible that cause the onset of osteoarthritis, and is also easily affected by the selected prediction model. It can be seen that the current clinical diagnostic procedures cannot fully meet the needs of clinicians to give accurate assessment and diagnosis of the severity of osteoarthritis in patients.

[0004] Therefore, there is an urgent need to provide a method that can accurately and effectively predict the severity of osteoarthritis, so as to improve the application value of osteoarthritis onset prediction, treatment guidance and treatment effect tracking. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting osteoarthritis. By collecting and processing patient clinical information and relevant biomarker test results (X-ray joint space width data) and other data, the optimal osteoarthritis prediction model is screened and determined from multiple different types of machine learning models, and the osteoarthritis condition is accurately and effectively predicted by the model, and the corresponding influencing factors are given. On the basis of making up for the deficiencies of the existing technology and fully meeting the needs of clinicians for predicting and diagnosing patients' osteoarthritis conditions, the accuracy of the prediction results is effectively improved, and the application value of osteoarthritis onset prediction, treatment guidance and treatment effect tracking is further enhanced.

[0006] In order to achieve the above objectives, it is necessary to provide an osteoarthritis disease prediction method, system, computer device and storage medium in response to the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides a method for predicting osteoarthritis, the method comprising the following steps:

[0008] Acquire an osteoarthritis dataset, and divide the osteoarthritis dataset into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis dataset includes clinical sample data and X-ray joint space width data;

[0009] Determining the optimal osteoarthritis disease prediction model from different machine learning models according to the training set and the validation set;

[0010] According to the test set and the optimal osteoarthritis disease prediction model, the corresponding osteoarthritis disease prediction results and influencing factors are obtained.

[0011] Furthermore, the step of obtaining an osteoarthritis dataset and dividing the osteoarthritis dataset into a training set, a validation set and a test set according to a preset ratio includes:

[0012] Screening the disease-irrelevant characteristic variables in the osteoarthritis data set, and deleting the disease-irrelevant characteristic variables to obtain a first osteoarthritis data set;

[0013] One-hot encoding is used to process the discrete feature variables in the first osteoarthritis data set into corresponding feature value vectors, and the feature value vectors are used to replace the corresponding discrete feature variables to obtain a second osteoarthritis data set;

[0014] Finding missing data samples in the second osteoarthritis data set, and filling the missing data samples to obtain a third osteoarthritis data set;

[0015] Standardizing the continuous characteristic variables in the third osteoarthritis data set to obtain a fourth osteoarthritis data set;

[0016] Merging the clinical sample data and the X-ray joint space width data corresponding to the same sample ID in the fourth osteoarthritis data set to obtain a fifth osteoarthritis data set;

[0017] According to the fifth osteoarthritis data set, a corresponding osteoarthritis disease score statistical distribution is obtained, and according to the osteoarthritis disease score statistical distribution, the fifth osteoarthritis data set is expanded to obtain a sixth osteoarthritis data set;

[0018] Obtaining the Pearson correlation coefficient between each characteristic variable in the sixth osteoarthritis data set and the corresponding Kellgren-Lawrence grading score, and deleting the characteristic variables whose Pearson correlation coefficient is greater than a preset correlation threshold, to obtain a seventh osteoarthritis data set;

[0019] The seventh osteoarthritis data set is divided into a training set, a validation set and a test set according to the preset ratio.

[0020] Furthermore, the step of searching for missing data samples in the second osteoarthritis dataset and filling the missing data samples to obtain a third osteoarthritis dataset includes:

[0021] Determine whether the data missing rate of the data missing sample is greater than a preset missing rate threshold. If so, delete the corresponding data missing sample. Otherwise, fill the missing feature variable corresponding to the data missing sample with the mean value of the corresponding feature variable of the data complete sample.

[0022] Further, the step of obtaining a corresponding osteoarthritis disease score statistical distribution according to the fifth osteoarthritis data set, and expanding the fifth osteoarthritis data set according to the osteoarthritis disease score statistical distribution to obtain a sixth osteoarthritis data set includes:

[0023] According to the Kellgren-Lawrence grading score corresponding to the x-ray joint space width data in the fifth osteoarthritis data set, statistically analyzing the osteoarthritis disease score distribution;

[0024] According to the statistical distribution of the osteoarthritis disease score, determining whether the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are evenly distributed;

[0025] If the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are unevenly distributed, SMOTE upsampling is performed on the sample data with the smallest proportion of Kellgren-Lawrence grading scores in the osteoarthritis disease score statistical distribution to obtain the sixth osteoarthritis data set.

[0026] Further, the machine learning model includes logistic regression, support vector machine, decision tree and XGBoost;

[0027] The step of determining the optimal osteoarthritis condition prediction model from different machine learning models based on the training set and the validation set comprises:

[0028] Establishing each machine learning model in advance, and using the training set to train each machine learning model respectively, and updating the model parameters of each machine learning model;

[0029] The validation set is used to train each machine learning model whose model parameters have been updated, and the model hyperparameters of each machine learning model are updated to obtain a corresponding osteoarthritis disease prediction model and MSE index value;

[0030] The MSE index values ​​corresponding to the various osteoarthritis condition prediction models are compared, and the osteoarthritis condition prediction model corresponding to the minimum MSE index value is determined as the optimal osteoarthritis condition prediction model.

[0031] Furthermore, the influencing factors include key risk characteristic variables and potential risk characteristic variables;

[0032] The step of obtaining the corresponding osteoarthritis condition prediction results and influencing factors according to the test set and the optimal osteoarthritis condition prediction model comprises:

[0033] Inputting the test set into the optimal osteoarthritis disease prediction model to obtain a corresponding osteoarthritis disease prediction result and a feature contribution score of each feature variable in the test set to the osteoarthritis disease prediction result;

[0034] Arrange the contribution scores of each feature in descending order to obtain the corresponding prediction result contribution ranking, and filter out a first preset number of feature contribution scores from top to bottom;

[0035] Acquire characteristic variables corresponding to the first preset number of characteristic contribution scores as risk characteristic variables, and select a second preset number of characteristic variables from the risk characteristic variables according to the characteristic contribution scores from high to low as risk characteristic variables to be analyzed;

[0036] According to the risk characteristic variables to be analyzed, the key risk characteristic variables and potential risk characteristic variables are determined.

[0037] Furthermore, the step of determining the key risk characteristic variables and the potential risk characteristic variables according to the risk characteristic variables to be analyzed includes:

[0038] Obtaining the Pearson correlation coefficient between each risk characteristic variable to be analyzed and the corresponding Kellgren-Lawrence grading score, and arranging the Pearson correlation coefficients in descending order to obtain the correlation coefficient ranking corresponding to each risk characteristic variable to be analyzed;

[0039] Determine whether the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are within the preset ranking range;

[0040] If the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are both within the preset ranking range, the risk characteristic variable to be analyzed is determined to be the key risk characteristic variable; otherwise, the risk characteristic variable to be analyzed is determined to be the potential risk characteristic variable.

[0041] In a second aspect, an embodiment of the present invention provides an osteoarthritis condition prediction system, the system comprising:

[0042] A data acquisition module, used to acquire an osteoarthritis data set, and divide the osteoarthritis data set into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis data set includes clinical sample data and X-ray joint space width data;

[0043] A model building module, used to determine the optimal osteoarthritis disease prediction model from different machine learning models based on the training set and the validation set;

[0044] The result prediction module is used to obtain the corresponding osteoarthritis condition prediction results and influencing factors according to the test set and the optimal osteoarthritis condition prediction model.

[0045] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0046] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0047] The above-mentioned application provides an osteoarthritis disease prediction method, system, computer device and storage medium. Through the method, an osteoarthritis data set including clinical sample data and X-ray joint space width data is obtained. After the necessary feature engineering preprocessing is performed on different feature variables in the data set in combination with statistical analysis to obtain an effective data set, it is divided into a training set, a validation set and a test set according to a preset ratio, and a plurality of preset different types of machine learning models are trained according to the training set and the validation set, and the optimal osteoarthritis disease prediction model is screened and determined from different machine learning models according to the corresponding model MES indicator size, and then the determined optimal osteoarthritis disease prediction model is used to predict the test set data to obtain the corresponding osteoarthritis disease prediction result, and the corresponding key risk factors and potential risk factors are further determined based on the osteoarthritis disease prediction result. Compared with the prior art, the present invention collects and processes patient clinical information and relevant biomarker test results, screens and establishes the optimal osteoarthritis disease prediction model, and determines the corresponding osteoarthritis disease prediction results and influencing factors through the optimal osteoarthritis disease prediction model. While making up for the deficiencies of the prior art and providing convenience for clinicians to evaluate and diagnose the condition of osteoarthritis patients, the present invention effectively improves the accuracy of the prediction results and further enhances the application value of osteoarthritis onset prediction, treatment guidance and treatment effect tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of application scenario of the osteoarthritis condition prediction method in an embodiment of the present invention;

[0049] Figure 2 is a flow chart of a method for predicting osteoarthritis in an embodiment of the present invention;

[0050] Figure 3 yes Figure 2 Step S11 is a flow chart of obtaining an osteoarthritis data set and dividing the data set;

[0051] Figure 4 is a schematic diagram of the statistical distribution of osteoarthritis disease scores of the OAI data set in an embodiment of the present invention;

[0052] Figure 5 a and b are respectively schematic diagrams of the statistical results of the correlation coefficients between the continuous feature variables, the discrete feature variables and the osteoarthritis disease score in the OAI data set in the embodiment of the present invention;

[0053] Figure 6 yes Figure 2 A schematic diagram of a process for determining an optimal osteoarthritis disease prediction model in step S12;

[0054] Figure 7: is a schematic diagram of the confusion matrix of the optimal osteoarthritis disease prediction model (XGBoost) on the test set in an embodiment of the present invention;

[0055] Figure 8 yes Figure 2 Step S13 is a flow chart of obtaining disease prediction results and influencing factors according to the optimal osteoarthritis disease prediction model;

[0056] Fig. 9 a, b, c and d are schematic diagrams of the ranking results of the contribution of feature variables obtained by the XGBoost model, decision tree (DT), support vector machine SVM and logistic regression in the embodiments of the present invention respectively;

[0057] Fig.10 is a schematic diagram of the structure of an osteoarthritis condition prediction system in an embodiment of the present invention;

[0058] Fig.11 It is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and beneficial effects of the present application clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the embodiments described below are part of the embodiments of the present invention and are only used to illustrate the present invention, but are not used to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The osteoarthritis disease prediction method provided by the present invention can be applied to Figure 1The terminal or server shown in the figure. The terminal may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices, and the server may be implemented as an independent server or a server cluster consisting of multiple servers. The server can adopt the osteoarthritis disease prediction method of the present invention according to actual application needs, based on the open source osteoarthritis data set obtained from the public osteoarthritis database (Osteoarthritis Initiative, OAI, https: / / data-archive.nimh.nih.gov / oai / ), and conduct a prediction comparison study on various different types of machine learning models to determine the optimal osteoarthritis disease prediction model, and use the optimal osteoarthritis disease prediction model to analyze and process the patient's disease data, and obtain accurate and effective disease prediction results for each patient, and the characteristic contribution of each characteristic variable to screen out a preset number of risk characteristic variables, and based on the obtained risk characteristic variables, screen out a certain number of key risk characteristic variables and potential risk characteristic variables, and give the influencing factors that may lead to the disease prediction results, and use the final osteoarthritis disease prediction results and influencing factors for subsequent research on the server, or send them to the terminal for further analysis and research by the terminal user, which can not only provide a reference for clinicians to evaluate and diagnose the disease of osteoarthritis patients, but also effectively improve the accuracy of the prediction results, thereby enhancing the application value of osteoarthritis onset prediction, treatment guidance and treatment effect tracking. The following embodiment will describe the osteoarthritis disease prediction method of the present invention in detail.

[0061] In one embodiment, Figure 2 As shown, a method for predicting osteoarthritis is provided, comprising the following steps:

[0062] S11, obtain an osteoarthritis data set, and divide the osteoarthritis data set into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis data set includes clinical sample data and x-ray joint space width data; wherein, the osteoarthritis data set can in principle use the existing osteoarthritis clinical data that can be collected, but since the clinical data of osteoarthritis patients generally involve patient privacy, and the disease diagnosis of osteoarthritis is a very professional issue, the amount of clinical data available is generally very small, and the Osteoarthritis Research Institute under the National Institutes of Health of the United States has obtained a large amount of clinical data through a multi-center osteoarthritis follow-up survey project (OAI) lasting 11 years, providing the public with freely available research data. In order to ensure the comprehensiveness of the research, as well as the effectiveness and accuracy of the optimal osteoarthritis disease prediction model, this embodiment selects 4796 osteoarthritis case samples from the osteoarthritis database (OAI) data set obtained as described above for research and analysis, which includes data of patients aged 45-79 years old, and each sample contains 1247 features. After obtaining the osteoarthritis dataset, the clinical sample data and x-ray joint space width data including different feature variables in the dataset are preprocessed with corresponding feature engineering, and then the sample feature variables are merged according to the patient ID, and the training set, validation set and test set are divided as required to establish an optimal osteoarthritis disease prediction model that meets the application requirements. It should be noted that the preset ratio of dividing the osteoarthritis dataset into training set, validation set and test set can be determined according to the actual application requirements, but it is necessary to ensure that the distribution of different disease levels in the divided training set, validation set and test set remains consistent. Specifically, if Figure 3 As shown, the step S11 of acquiring an osteoarthritis dataset and dividing the osteoarthritis dataset into a training set, a validation set and a test set according to a preset ratio includes:

[0063] S111, screening the disease-irrelevant characteristic variables in the osteoarthritis data set, and deleting the disease-irrelevant characteristic variables to obtain a first osteoarthritis data set; wherein, the disease-irrelevant characteristic variables refer to characteristic variables in the data set that are only used for statistical records and are irrelevant to the arthritis disease itself, such as the number, version number, survey completion date, X-ray clinical test number, first screening date and registration date, etc. Such characteristic variables will only increase the amount of data processing and will not have an essential impact on the research of osteoarthritis, but may cause the model to overfit, resulting in a lack of generalization ability. In order to effectively avoid overfitting and improve the generalization ability of the model, this embodiment preferably finds out the disease-irrelevant characteristic variables in the original data set and removes them.

[0064] S112. One-hot encoding is used to process the discrete feature variables in the first osteoarthritis data set into corresponding feature numerical vectors, and the feature numerical vectors are used to replace the corresponding discrete feature variables to obtain a second osteoarthritis data set; wherein, one-hot encoding is a single-bit valid encoding, and an N-bit state register is used to encode N states, each state has its own independent register bit, and only one bit is valid at any time, which requires that the classification value be mapped to an integer value, and each integer value is represented as a corresponding binary vector, the value corresponding to the index bit of the integer is marked as 1, and the other bits are all zero values. In this embodiment, due to the large number of discrete feature variables, the discrete feature variables that specifically need to be processed by one-hot encoding in the clinical sample data and the x-ray joint space width data are not listed here. The specific process of how to process each discrete feature variable using one-hot encoding can be referred to the existing implementation, and will not be repeated here.

[0065] S113. Search for data missing samples in the second osteoarthritis data set, and fill in the data missing samples to obtain a third osteoarthritis data set; wherein, the data missing samples cannot be used due to the missing of individual feature variable data, and in principle, all of them should be eliminated, but in order to ensure that there is enough data for modeling and prediction, this embodiment adopts a non-processing method for the data missing samples according to different missing rate situations.

[0066] Specifically, the step of searching for missing data samples in the second osteoarthritis data set and filling the missing data samples to obtain the third osteoarthritis data set includes: determining whether the missing data rate of the missing data samples is greater than a preset missing rate threshold, if so, deleting the corresponding missing data samples, otherwise, filling the missing feature variables corresponding to the missing data samples with the mean of the feature variables corresponding to the complete data samples. It should be noted that the preset missing rate threshold can be determined according to actual application requirements. If the missing data rate of a feature variable in a sample reaches 20%, even if the optimal filling method is used, the accuracy of the model prediction will be reduced. The corresponding missing rate threshold is set to 20%, and the samples exceeding the threshold are directly deleted. Otherwise, the feature variables corresponding to the missing data in the sample are effectively filled and continue to be used. The above-mentioned specific missing data filling method may not be limited to mean filling, and various methods such as default value filling, mode filling, KNN filling, and interpolation filling may also be used. In order to improve the data processing efficiency as much as possible while ensuring the availability of missing data filling, this embodiment preferably uses mean filling for missing data samples in clinical sample data and x-ray joint space width data, and the other filling methods mentioned above are in principle also within the protection scope of the present invention.

[0067] S114, normalizing the continuous characteristic variables in the third osteoarthritis data set to obtain a fourth osteoarthritis data set; wherein, in principle, the continuous characteristic variables can be normalized by any of the methods of min-max normalization, log function conversion, atan function conversion, z-score normalization (zero-menanormalization) and fuzzy quantization. And in this embodiment, the commonly used z-score normalization is preferably used to normalize each continuous characteristic variable in the clinical sample data and the x-ray joint space width data.

[0068] S115. Merge the clinical sample data and the x-ray joint space width data corresponding to the same sample ID in the fourth osteoarthritis data set to obtain a fifth osteoarthritis data set; wherein, as mentioned above, the clinical sample data and the x-ray joint space width data are both derived from the osteoarthritis database (OAI), and the data of the same patient in the clinical sample data and the x-ray joint space width data are recorded using the same sample ID number. Since the characteristic variables in the clinical sample data and the x-ray joint space width data will affect the prediction of osteoarthritis, in order to ensure the richness of the characteristic variables and the comprehensiveness of the research, and thereby improve the accuracy of the model prediction analysis, this embodiment merges the characteristic variables in the clinical sample data and the x-ray joint space width data corresponding to the same patient to obtain a sample with more characteristic variables, which is convenient for subsequent research use.

[0069] S116. According to the fifth osteoarthritis data set, the corresponding osteoarthritis disease score statistical distribution is obtained, and according to the osteoarthritis disease score statistical distribution, the fifth osteoarthritis data set is expanded to obtain the sixth osteoarthritis data set; wherein, the osteoarthritis disease score statistical distribution can be obtained by statistically analyzing the fifth osteoarthritis data set according to the Kellgren-Lawrence grading score. The Kellgren-Lawrence grading system for osteoarthritis is an assessment and grading method for the condition (severity) of knee osteoarthritis. It is divided into grades 0 (normal knee joint), I, II, III and IV (the most severe knee osteoarthritis) from mild to severe according to the performance of knee x-rays, and different osteoarthritis conditions in the OAI data set also use 0, 1, 2, 3 and 4 to represent the gradually increasing severity of osteoarthritis (different disease prediction grades), wherein 0 represents normal and 4 represents very severe. Figure 4As shown in the figure, the statistical distribution of osteoarthritis disease scores in the OAI dataset shows the proportion of different Kellgren-Lawrence grading scores in the dataset, which is convenient for intuitively understanding whether each type of sample is evenly distributed. The number of samples with very severe osteoarthritis in the figure is significantly less than the other four types, which shows that in the collected case samples, there is sample unevenness in the osteoarthritis disease prediction data (severity). At this time, in order to ensure the effectiveness of modeling prediction, it is necessary to expand the data samples that are significantly less than other types of samples to a certain extent.

[0070] Specifically, the step of obtaining a corresponding osteoarthritis disease score statistical distribution according to the fifth osteoarthritis data set, and expanding the fifth osteoarthritis data set according to the osteoarthritis disease score statistical distribution to obtain a sixth osteoarthritis data set includes:

[0071] According to the Kellgren-Lawrence grading score corresponding to the x-ray joint space width data in the fifth osteoarthritis data set, statistically analyzing the osteoarthritis disease score distribution;

[0072] According to the statistical distribution of the osteoarthritis disease score, determining whether the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are evenly distributed;

[0073] If the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are unevenly distributed, then SMOTE upsampling is performed on the sample data with the smallest proportion of Kellgren-Lawrence grading scores in the statistical distribution of the osteoarthritis disease score to obtain the sixth osteoarthritis data set. Among them, SMOTE upsampling includes four methods: SMOTE-Regular, SMOTE-Borderline1, SMOTE-Borderline2 and SMOTE-SVM. In specific practical applications, they can be selected according to application requirements and implemented with reference to the corresponding prior art. In this embodiment, the number of samples with very severe osteoarthritis in the data set as described above is significantly less than the other four types. Therefore, it is necessary to use SMOTE upsampling to expand the samples in the data set with a Kellgren-Lawrence grading score of less than 4 to balance the number of samples in the entire data set.

[0074] S117. Obtain the Pearson correlation coefficient between each characteristic variable in the sixth osteoarthritis data set and the corresponding Kellgren-Lawrence grading score, and delete the characteristic variables whose Pearson correlation coefficient is greater than a preset correlation threshold, to obtain the seventh osteoarthritis data set; wherein, the preset correlation threshold can be determined according to actual application requirements. In actual correlation analysis, it will be found that some characteristic variables have a very high correlation with the prediction results. If such characteristic variables are used for modeling, it will result in the inability to observe more other characteristic variables related to the prediction results, resulting in distorted prediction results. Therefore, in order to ensure the accuracy of model prediction while being able to effectively analyze the value of more characteristic variables to model prediction, this embodiment removes characteristic variables that exceed a certain preset correlation threshold from the data set and no longer uses them as modeling data. Figure 5 As shown in (a, b), the correlation between most of the characteristic variables in the sixth osteoarthritis data set and the corresponding Kellgren-Lawrence grading score is less than 0.3, while some characteristic variables such as whether there are osteophytes on the x-ray of the left and right knee joints, the pain symptoms of the patient's left and right knee joints, and whether the x-ray of the left and right knees shows narrowing of the medial joint space are all greater than 0.6, which is significantly higher than the Pearson correlation coefficients corresponding to most characteristic variables. In this case, such characteristic variables can be deleted.

[0075] S118, dividing the seventh osteoarthritis data set into a training set, a validation set and a test set according to the preset ratio. The preset ratio can be determined according to the actual model training requirements as described above, such as dividing the seventh osteoarthritis data set obtained after all preprocessing into a training set, a validation set and a test set according to the ratio of 65%, 25% and 10%. During the actual division, conventional statistical analysis methods can be used to check whether the statistical distribution of the osteoarthritis disease scores (osteoarthritis severity distribution) corresponding to the training set, validation set and test set after the division is different from the statistical distribution of the osteoarthritis disease scores of the data set before the division. If there is a discrepancy, corresponding adjustments need to be made, that is, on the basis of ensuring the randomness of the data set division process and meeting the training quantity requirements, it is possible to ensure that there is no significant difference in the distribution of each feature in the training set, validation set and test set, so as to provide reliable guarantee for obtaining the optimal osteoarthritis disease prediction model for subsequent training. As shown in Table 1, the training set, validation set, and test set divided are consistent with the distribution of the original sample data, but Table 2 shows that the samples with an osteoarthritis KL score of 4 in the training set only account for about 5% of the total samples, and the samples with an osteoarthritis KL score of 4 in the test set only account for 3.14%. The small number of samples with KL of 4 in the training set and the test set will lead to unsatisfactory model training effects and inaccurate prediction results. This division ratio is not recommended, and it is necessary to readjust and select a more appropriate preset ratio for reasonable data set division. It should be noted that the preset ratio given above is only an exemplary description and is not specifically limited here.

[0076] Table 1. Distribution of some feature variables in the training set, validation set, and test set

[0077]

[0078] Table 2. Distribution of osteoarthritis prediction results (severity) in the training set, validation set, and test set

[0079] K&L Training set (%) Validation set (%) Test set (%) Total(%) Level 0 19.19 19.36 19.90 19.29 Level 1 35.11 35.91 37.96 35.56 Level 2 18.89 18.68 20.42 18.96 Level 3 21.01 19.97 18.59 20.53 Level 4 5.80 6.07 3.14 5.66

[0080] S12. Determine the optimal osteoarthritis disease prediction model from different machine learning models based on the training set and validation set; wherein, in principle, the machine learning model can select all network learning models that can predict osteoarthritis disease based on the corresponding data set, such as logistic regression, support vector machine, decision tree, XGBoost and naive Bayes model, etc. However, in order to facilitate the subsequent analysis of the characteristic variables in the data set after obtaining the optimal osteoarthritis disease prediction model, and then determine the influencing factors related to the disease, the machine learning model in this embodiment selects four learning models that can perform characteristic variable contribution analysis, namely logistic regression, support vector machine, decision tree and XGBoost, for comparative study, and finally selects the model with the best performance from the four models as the optimal osteoarthritis disease prediction model. Specifically, Figure 6 As shown, the step S12 of determining the optimal osteoarthritis disease prediction model from different machine learning models based on the training set and the validation set includes:

[0081] S121, pre-establishing each machine learning model, and using the training set to train each machine learning model respectively, and updating the model parameters of each machine learning model;

[0082] S122, using the validation set to train each machine learning model whose model parameters have been updated, updating the model hyperparameters of each machine learning model, and obtaining a corresponding osteoarthritis condition prediction model and MSE index value;

[0083] S123, comparing the MSE index values ​​corresponding to the osteoarthritis condition prediction models, and determining the osteoarthritis condition prediction model corresponding to the minimum MSE index value as the optimal osteoarthritis condition prediction model.

[0084] Specifically, after setting the training parameters of four machine learning models, XGBoost, Decision Tree (DT), support vector machine (SVM) and logistic regression (Logisticregression), respectively, according to the following standards, the training set and validation set were used to predict the condition of osteoarthritis, and the corresponding performance is shown in Table 3:

[0085] (1) The XGBoost model comes from the XGBClassifier in the xgboost module, and its corresponding training parameters are learning rate = 0.1, nestimators = 100, max depth = 6, colsample btree = 0.8, and lossfunction = multi:softmax;

[0086] (2) The DT model comes from DecisionTreeClassifier in sklearn.tree, and its training parameters are all set to default values;

[0087] (3) The SVM model comes from SVC in the sklearn.svm module, and its corresponding model parameter is kernel=linear;

[0088] (4) The Logistic model comes from LogisticRegression in the sklearn.linear_model module, and its corresponding training parameters are C=0.1, penalty=elasticnet, and L1 ratio=0.5.

[0089] Table 3. Estimated F1 scores and MSE for each KL level for different machine learning models

[0090]

[0091] As shown in Table 3, the prediction effect of the tree structure model (XGBoost, DT) on the OAI dataset is relatively outstanding, while the corresponding prediction accuracy of the support vector machine SVM and logistic regression under the same data processing conditions is much lower than that of the XGBoost model, and the MSE of the XGBoost model is 0.812, which is significantly better than the other three machine learning models: DT (MSE is 1.317), SVM (MSE is 1.955) and Logistics (MSE is 4.745). It can be seen that the optimal osteoarthritis disease prediction model determined in this embodiment should be the XGBoost model.

[0092] In order to further verify the excellent performance of the XGBoost model on the OAI dataset, the F1-score score was selected for evaluation. As shown in Table 4, the F1-score of the test set was 0.69, and the F1-scores of the KL scores of 1, 2, and 3 for osteoarthritis prediction were significantly higher than those of the KL scores of 0 and 4. Through the precision and recall scores, it can be found that the prediction precision of KL 0 is the highest, and the recall rate of KL 1 is the highest. In addition, through Figure 7 The confusion matrix shown in the figure shows that XGBoost predicts 68% of the samples correctly in the test set, and the remaining 32% are incorrectly predicted. However, the prediction results of the incorrectly predicted samples are only close to the true KL score of the actual osteoarthritis condition prediction.

[0093] Table 4. Recall, precision and F1 score of each KL predicted by XGBoost validation set

[0094]

[0095] S13. Obtain corresponding osteoarthritis disease prediction results and influencing factors based on the test set and the optimal osteoarthritis disease prediction model; wherein the osteoarthritis disease prediction result, i.e., the osteoarthritis severity score, is the prediction result obtained by inputting the test set into the optimal osteoarthritis disease prediction model determined above, and the influencing factors are the characteristic variables that are highly correlated with the prediction results determined by using the optimal osteoarthritis disease prediction model to sort all characteristic variables involved in model training in the data set, and the influencing factors include key risk characteristic variables and potential risk characteristic variables. Among them, the key risk characteristic variables are characteristic variables that are determined to have a high impact on the osteoarthritis disease, and the potential risk characteristic variables are characteristic variables that have a potential impact on the osteoarthritis disease. Specifically, Figure 8 As shown, the step S13 of obtaining the corresponding osteoarthritis condition prediction results and influencing factors according to the test set and the optimal osteoarthritis condition prediction model includes:

[0096] S131, inputting the test set into the optimal osteoarthritis disease prediction model to obtain the corresponding osteoarthritis disease prediction result and the feature contribution score of each feature variable in the test set to the osteoarthritis disease prediction result; wherein, the optimal osteoarthritis disease prediction model has been verified in the aforementioned embodiment to be an XGBoost model on the OAI data set, then when the test set data is input into the optimal osteoarthritis disease prediction model to obtain the corresponding osteoarthritis disease prediction result, the XGBoost model can be used to score the contribution of all feature variables to the prediction result based on the classification and regression trees (CART) algorithm, using the Gini coefficient to obtain Fig. 9 Corresponding feature contribution score ranking. It should be noted that if the OAI dataset is not used for research in actual applications, the optimal osteoarthritis disease prediction model may be one of the other three. In this case, the feature contribution scoring algorithm of the corresponding model can be used to obtain the contribution of different feature variables in the dataset to the corresponding prediction results. For example, decision tree (DT) and XGBoost belong to the same tree structure model, and the same feature contribution scoring algorithm as XGBoost can be used to obtain the feature variable contribution ranking. The support vector machine SVM and logistic regression obtain their corresponding contribution score rankings based on the coefficients of each feature variable obtained through training. The specific implementation can refer to the existing technology and will not be repeated here.

[0097] S132, arranging each feature contribution score in descending order to obtain a corresponding prediction result contribution ranking, and screening out a first preset number of feature contribution scores from top to bottom;

[0098] Among them, after the prediction result contribution ranking is obtained according to the above method, the first preset number of feature contribution scores selected can be determined according to actual research needs. For example, the first preset number can be determined as 25 with reference to existing osteoarthritis research literature, that is, the top 25 feature variables with the highest contribution to the prediction result are selected from all feature variables currently involved in model training as shown in the following table for further analysis and research. It should be noted that the value of the first preset number given here is only for illustrative purposes, and the specific range is not limited.

[0099] Table 5. Top 25 feature variables selected by XGBoost for the prediction of osteoarthritis

[0100]

[0101]

[0102] Fig. 9 Among the top 25 feature variables with the highest contribution to the prediction results, the feature variables related to Radiograph information (x-ray joint space width data) contribute the most to the prediction results of the XGBoost model, and the top 5 risk feature variables (SVRKJSM, SVLKJSL, ELGRISK, RXRKOA, SVRKJSL) are all related to Radiographic information, which effectively proves that the relevant results of x-ray joint space width data have a direct guiding significance for assessing the condition (severity) of osteoarthritis; in addition, the feature variables corresponding to some key clinical indicators in the clinical sample data also contribute greatly to the prediction of the model, such as LKFHDEG, lfTHRL and RLVER, which are directly related to the degree of knee joint activity of the subjects, which also shows that these key clinical indicators can be used as an indirect way to predict the condition of osteoarthritis. At the same time, it also reveals that the two feature variables of age and weight are also two key factors directly related to the severity of osteoarthritis. The above prediction result contribution ranking also further verifies that the present invention combines clinical sample data and X-ray joint space width data as the basic data for predictive modeling, which not only ensures the integrity and diversity of the modeling data, but also helps to improve the comprehensiveness of the predictive research, improve the accuracy of the prediction results, and enhance the generalization ability of the model.

[0103] S133, obtaining the characteristic variables corresponding to the first preset number of characteristic contribution scores as risk characteristic variables, and selecting the second preset number of characteristic variables from the risk characteristic variables according to the characteristic contribution scores from high to low as risk characteristic variables to be analyzed; wherein the second preset number can also be determined according to actual research needs, and in principle, it is not greater than the first preset number. After selecting the second preset number of risk characteristic variables to be analyzed according to the above method, use them as the basis for selecting key risk characteristic variables and potential risk characteristic variables.

[0104] S134. Determine the key risk characteristic variables and potential risk characteristic variables according to the risk characteristic variables to be analyzed. The key risk characteristic variables and potential risk characteristic variables are further classifications of the risk characteristic variables to be analyzed based on the correlation between the risk characteristic variables to be analyzed and the corresponding Kellgren-Lawrence grading scores. Specifically, the step of determining the key risk characteristic variables and potential risk characteristic variables according to the risk characteristic variables to be analyzed includes:

[0105] Obtaining the Pearson correlation coefficient between each risk characteristic variable to be analyzed and the corresponding Kellgren-Lawrence grading score, and arranging the Pearson correlation coefficients in descending order to obtain the correlation coefficient ranking corresponding to each risk characteristic variable to be analyzed;

[0106] Determine whether the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are within the preset ranking range;

[0107] Among them, the preset ranking range can also be determined according to actual research needs. If only the correlation coefficient ranking and the prediction result contribution ranking of the risk characteristic variable to be analyzed are both in the top five range, then the risk characteristic variable to be analyzed is considered to be a key risk characteristic variable, and the preset ranking range can be set to 5. At the same time, if one of the correlation coefficient ranking and the prediction result contribution ranking of a risk characteristic variable to be analyzed is not in the top five range, then the risk characteristic variable to be analyzed is considered not to be a key risk characteristic variable but a potential risk characteristic variable, and so on.

[0108] If the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are both within the preset ranking range, the risk characteristic variable to be analyzed is determined to be the key risk characteristic variable; otherwise, the risk characteristic variable to be analyzed is determined to be the potential risk characteristic variable.

[0109] Specifically, as described above, the optimal osteoarthritis disease prediction model OAI data set is determined as an XGBoost model. After obtaining the first preset number of risk feature variables with the highest feature contribution score through XGBoost, the second preset number of feature variables to be analyzed are further screened from the first preset number of risk feature variables, and the correlation between them and the KL score of the osteoarthritis disease is statistically analyzed. For example, after obtaining 25 risk feature variables, the first 10 risk feature variables are screened as feature variables to be analyzed, and the Pearson correlation coefficient between each feature variable to be analyzed and the corresponding KL score of the osteoarthritis disease is further analyzed to obtain the correlation coefficient ranking shown in Table 6. From the results of Table 6, it can be seen that the two features (P01RXRKOA and P02ELGRISK) with the highest correlation (large correlation coefficient) also have a large contribution score to the model, and are feature variables in the clinical sample data and the x-ray joint space width data, respectively. Such features can be definitely the key risk feature variables affecting the osteoarthritis disease. However, P01SVRKJSM has no obvious correlation with osteoarthritis, but its contribution to the XGBoost model is the largest. This shows that the feature variable with greater correlation does not necessarily contribute more to the model. This feature variable cannot necessarily be used as a key risk feature variable that affects osteoarthritis, but it has a potential intrinsic connection with osteoarthritis. It also proves that the feature variable obtained by directly observing the gap between bones and joints through x-ray images can be used as a key evidence for doctors to determine whether the patient has osteoarthritis. Therefore, this type of risk feature variable to be analyzed can be used as a potential risk feature variable to provide data support for clinicians' diagnosis and research. It should be noted that the above key risk feature variables and potential risk feature variables only give corresponding examples, and do not represent the corresponding feature variable classification obtained in actual research.

[0110] Table 6. XGBoost’s top 10 important feature contribution scores and correlation coefficients for osteoarthritis disease prediction

[0111]

[0112] The embodiment of the present application proposes a method of combining the corresponding feature variables of the collected patient clinical information and X-ray detection results as a basic data set for osteoarthritis disease prediction after corresponding feature engineering processing is performed on the two, thereby providing rich and comprehensive data information for subsequent screening of the optimal osteoarthritis disease prediction model from multiple different machine learning models, ensuring that the selected optimal osteoarthritis disease prediction model has high accuracy and strong generalization ability, and based on this model, the osteoarthritis disease prediction results and influencing factors can be accurately and effectively predicted, which not only makes up for the deficiencies of the existing technology and provides convenience for clinicians to evaluate and diagnose the condition of osteoarthritis patients, but also effectively improves the accuracy of the prediction results, and also provides an effective classification method for different risk feature variables in the data set, further enhancing the application value of osteoarthritis onset prediction, treatment guidance and treatment effect tracking.

[0113] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0114] In one embodiment, Fig.10 As shown, a system for predicting osteoarthritis is provided, the system comprising:

[0115] Data acquisition module 1, used to acquire an osteoarthritis dataset, and divide the osteoarthritis dataset into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis dataset includes clinical sample data and X-ray joint space width data;

[0116] Model building module 2, used to determine the optimal osteoarthritis disease prediction model from different machine learning models based on the training set and the validation set;

[0117] The result prediction module 3 is used to obtain the corresponding osteoarthritis condition prediction results and influencing factors according to the test set and the optimal osteoarthritis condition prediction model.

[0118] It should be noted that the specific definition of the osteoarthritis disease prediction system can be found in the above-mentioned definition of the osteoarthritis disease prediction method, which will not be repeated here. Each module in the above-mentioned osteoarthritis disease prediction system can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0119] Fig.11FIG. 1 shows an internal structure diagram of a computer device in an embodiment, and the computer device may specifically be a terminal or a server. Fig.11 As shown, the computer device includes a processor, a memory, a network interface, a display and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting osteoarthritis is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0120] It can be understood by those skilled in the art that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied. The specific computing device may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.

[0121] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the above method are implemented when the processor executes the computer program.

[0122] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0123] In summary, an osteoarthritis disease prediction method, system, computer device and storage medium provided by an embodiment of the present invention realizes that the acquired clinical sample data and X-ray joint space width data are merged into an osteoarthritis data set, and after necessary feature engineering preprocessing is performed on different feature variables in combination with statistical analysis, the data set required for establishing a prediction model is divided according to a preset ratio, and a plurality of preset different types of machine learning models are trained according to a training set and a validation set, and the optimal osteoarthritis disease prediction model is determined according to the size of the MES index of each model, and then the determined optimal osteoarthritis disease prediction model is used to predict the test set data to obtain the corresponding osteoarthritis disease prediction result, and the corresponding osteoarthritis disease prediction result is further determined based on the osteoarthritis disease prediction result. The technical scheme of key risk variable characteristics (key influencing factors) and potential risk variable characteristics (potential influencing factors) is proposed. By combining the characteristic variables of clinical sample data and X-ray joint space width data as the basic data set for predicting osteoarthritis, a method of providing rich and comprehensive data information for the optimal osteoarthritis prediction model is provided. The method ensures that the selected optimal osteoarthritis prediction model has high precision and strong generalization ability, and improves the accuracy and effectiveness of predicting the osteoarthritis disease prediction results and influencing factors based on this model, providing an effective reference for clinicians to evaluate and diagnose the condition of osteoarthritis patients, and also provides an effective classification method for different risk characteristic variables in the data set, further enhancing the application value of osteoarthritis incidence prediction, treatment guidance and treatment effect tracking.

[0124] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0125] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A method for predicting osteoarthritis. It is characterized in that The method comprises the following steps: Acquire an osteoarthritis dataset, and divide the osteoarthritis dataset into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis dataset includes clinical sample data and X-ray joint space width data; Determining the optimal osteoarthritis disease prediction model from different machine learning models according to the training set and the validation set; the machine learning models include logistic regression, support vector machine, decision tree and XGBoost; According to the test set and the optimal osteoarthritis disease prediction model, corresponding osteoarthritis disease prediction results and influencing factors are obtained; The step of obtaining an osteoarthritis dataset and dividing the osteoarthritis dataset into a training set, a validation set and a test set according to a preset ratio comprises: Screening the disease-irrelevant characteristic variables in the osteoarthritis data set, and deleting the disease-irrelevant characteristic variables to obtain a first osteoarthritis data set; One-hot encoding is used to process the discrete feature variables in the first osteoarthritis data set into corresponding feature value vectors, and the feature value vectors are used to replace the corresponding discrete feature variables to obtain a second osteoarthritis data set; Finding missing data samples in the second osteoarthritis data set, and filling the missing data samples to obtain a third osteoarthritis data set; Standardizing the continuous characteristic variables in the third osteoarthritis data set to obtain a fourth osteoarthritis data set; Merging the clinical sample data and the X-ray joint space width data corresponding to the same sample ID in the fourth osteoarthritis data set to obtain a fifth osteoarthritis data set; According to the fifth osteoarthritis data set, a corresponding osteoarthritis disease score statistical distribution is obtained, and according to the osteoarthritis disease score statistical distribution, the fifth osteoarthritis data set is expanded to obtain a sixth osteoarthritis data set; Obtaining the Pearson correlation coefficient between each characteristic variable in the sixth osteoarthritis data set and the corresponding Kellgren-Lawrence grading score, and deleting the characteristic variables whose Pearson correlation coefficient is greater than a preset correlation threshold, to obtain a seventh osteoarthritis data set; Dividing the seventh osteoarthritis data set into a training set, a validation set and a test set according to the preset ratio; the statistical distribution of osteoarthritis disease scores of the training set, the validation set and the test set are kept consistent; The influencing factors include key risk characteristic variables and potential risk characteristic variables; the step of obtaining the corresponding osteoarthritis condition prediction results and influencing factors according to the test set and the optimal osteoarthritis condition prediction model includes: Inputting the test set into the optimal osteoarthritis disease prediction model to obtain a corresponding osteoarthritis disease prediction result and a feature contribution score of each feature variable in the test set to the osteoarthritis disease prediction result; Arrange the contribution scores of each feature in descending order to obtain the corresponding prediction result contribution ranking, and filter out a first preset number of feature contribution scores from top to bottom; Acquire characteristic variables corresponding to the first preset number of characteristic contribution scores as risk characteristic variables, and select a second preset number of characteristic variables from the risk characteristic variables according to the characteristic contribution scores from high to low as risk characteristic variables to be analyzed; Determining the key risk characteristic variables and potential risk characteristic variables according to the risk characteristic variables to be analyzed; The step of obtaining a corresponding osteoarthritis disease score statistical distribution according to the fifth osteoarthritis data set, and expanding the fifth osteoarthritis data set according to the osteoarthritis disease score statistical distribution to obtain a sixth osteoarthritis data set comprises: According to the Kellgren-Lawrence grading score corresponding to the x-ray joint space width data in the fifth osteoarthritis data set, statistically analyzing the osteoarthritis disease score distribution; According to the statistical distribution of the osteoarthritis disease score, determining whether the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are evenly distributed; If the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are unevenly distributed, performing SMOTE upsampling on the sample data with the smallest proportion of Kellgren-Lawrence grading scores in the statistical distribution of the osteoarthritis disease scores to obtain the sixth osteoarthritis data set; The step of determining the key risk characteristic variables and the potential risk characteristic variables according to the risk characteristic variables to be analyzed comprises: Obtaining the Pearson correlation coefficient between each risk characteristic variable to be analyzed and the corresponding Kellgren-Lawrence grading score, and arranging the Pearson correlation coefficients in descending order to obtain the correlation coefficient ranking corresponding to each risk characteristic variable to be analyzed; Determine whether the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are within the preset ranking range; If the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are both within the preset ranking range, the risk characteristic variable to be analyzed is determined to be the key risk characteristic variable; otherwise, the risk characteristic variable to be analyzed is determined to be the potential risk characteristic variable.

2. The method for predicting osteoarthritis according to claim 1, It is characterized in that The step of searching for missing data samples in the second osteoarthritis dataset and filling the missing data samples to obtain a third osteoarthritis dataset comprises: Determine whether the data missing rate of the data missing sample is greater than a preset missing rate threshold. If so, delete the corresponding data missing sample. Otherwise, fill the missing feature variable corresponding to the data missing sample with the mean value of the corresponding feature variable of the data complete sample.

3. The method for predicting osteoarthritis according to claim 1, It is characterized in that The step of determining the optimal osteoarthritis condition prediction model from different machine learning models based on the training set and the validation set comprises: Establishing each machine learning model in advance, and using the training set to train each machine learning model respectively, and updating the model parameters of each machine learning model; The validation set is used to train each machine learning model whose model parameters have been updated, and the model hyperparameters of each machine learning model are updated to obtain a corresponding osteoarthritis disease prediction model and MSE index value; The MSE index values ​​corresponding to the various osteoarthritis condition prediction models are compared, and the osteoarthritis condition prediction model corresponding to the minimum MSE index value is determined as the optimal osteoarthritis condition prediction model.

4. An osteoarthritis disease prediction system, It is characterized in that The system comprises: A data acquisition module, used to acquire an osteoarthritis data set, and divide the osteoarthritis data set into a training set, a validation set, and a test set according to a preset ratio; the osteoarthritis data set includes clinical sample data and X-ray joint space width data; A model building module, used to determine the optimal osteoarthritis disease prediction model from different machine learning models based on the training set and the validation set; the machine learning models include logistic regression, support vector machine, decision tree and XGBoost; A result prediction module, used to obtain corresponding osteoarthritis disease prediction results and influencing factors according to the test set and the optimal osteoarthritis disease prediction model; The step of obtaining an osteoarthritis dataset and dividing the osteoarthritis dataset into a training set, a validation set, and a test set according to a preset ratio includes: Screening the disease-irrelevant characteristic variables in the osteoarthritis data set, and deleting the disease-irrelevant characteristic variables to obtain a first osteoarthritis data set; One-hot encoding is used to process the discrete feature variables in the first osteoarthritis data set into corresponding feature value vectors, and the feature value vectors are used to replace the corresponding discrete feature variables to obtain a second osteoarthritis data set; Finding missing data samples in the second osteoarthritis data set, and filling the missing data samples to obtain a third osteoarthritis data set; Standardizing the continuous characteristic variables in the third osteoarthritis data set to obtain a fourth osteoarthritis data set; Merging the clinical sample data and the X-ray joint space width data corresponding to the same sample ID in the fourth osteoarthritis data set to obtain a fifth osteoarthritis data set; According to the fifth osteoarthritis data set, a corresponding osteoarthritis disease score statistical distribution is obtained, and according to the osteoarthritis disease score statistical distribution, the fifth osteoarthritis data set is expanded to obtain a sixth osteoarthritis data set; Obtaining the Pearson correlation coefficient between each characteristic variable in the sixth osteoarthritis data set and the corresponding Kellgren-Lawrence grading score, and deleting the characteristic variables whose Pearson correlation coefficient is greater than a preset correlation threshold, to obtain a seventh osteoarthritis data set; Dividing the seventh osteoarthritis data set into a training set, a validation set and a test set according to the preset ratio; the statistical distribution of osteoarthritis disease scores of the training set, the validation set and the test set are kept consistent; The influencing factors include key risk characteristic variables and potential risk characteristic variables; the corresponding osteoarthritis condition prediction results and influencing factors are obtained according to the test set and the optimal osteoarthritis condition prediction model, including: Inputting the test set into the optimal osteoarthritis disease prediction model to obtain a corresponding osteoarthritis disease prediction result and a feature contribution score of each feature variable in the test set to the osteoarthritis disease prediction result; Arrange the contribution scores of each feature in descending order to obtain the corresponding prediction result contribution ranking, and filter out a first preset number of feature contribution scores from top to bottom; Acquire characteristic variables corresponding to the first preset number of characteristic contribution scores as risk characteristic variables, and select a second preset number of characteristic variables from the risk characteristic variables according to the characteristic contribution scores from high to low as risk characteristic variables to be analyzed; Determining the key risk characteristic variables and potential risk characteristic variables according to the risk characteristic variables to be analyzed; The step of obtaining a corresponding osteoarthritis disease score statistical distribution according to the fifth osteoarthritis data set, and expanding the fifth osteoarthritis data set according to the osteoarthritis disease score statistical distribution to obtain a sixth osteoarthritis data set includes: According to the Kellgren-Lawrence grading score corresponding to the x-ray joint space width data in the fifth osteoarthritis data set, statistically analyzing the osteoarthritis disease score distribution; According to the statistical distribution of the osteoarthritis disease score, determining whether the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are evenly distributed; If the Kellgren-Lawrence grading scores of various levels in the fifth osteoarthritis data set are unevenly distributed, performing SMOTE upsampling on the sample data with the smallest proportion of Kellgren-Lawrence grading scores in the statistical distribution of the osteoarthritis disease scores to obtain the sixth osteoarthritis data set; Determining the key risk characteristic variables and potential risk characteristic variables according to the risk characteristic variables to be analyzed includes: Obtaining the Pearson correlation coefficient between each risk characteristic variable to be analyzed and the corresponding Kellgren-Lawrence grading score, and arranging the Pearson correlation coefficients in descending order to obtain the correlation coefficient ranking corresponding to each risk characteristic variable to be analyzed; Determine whether the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are within the preset ranking range; If the correlation coefficient ranking and prediction result contribution ranking of each risk characteristic variable to be analyzed are both within the preset ranking range, the risk characteristic variable to be analyzed is determined to be the key risk characteristic variable; otherwise, the risk characteristic variable to be analyzed is determined to be the potential risk characteristic variable.

5. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images