Pet cat sudden cardiac death risk assessment method

Abstract

The invention relates to the technical field of disease risk assessment, in particular to a pet cat sudden cardiac death risk assessment method, which comprises the following steps: collecting pet cat physical examination case information; constructing a sudden cardiac death risk assessment model of the pet cat according to the collected physical examination case information of the pet cat, and screening out characteristic indexes related to sudden cardiac death of the pet cat; and obtaining detection report data of a to-be-evaluated pet cat, and inputting the detection report data into the sudden cardiac death risk evaluation model of the pet cat to obtain a sudden cardiac death risk evaluation result of the pet cat. Scientific support can be provided for early diagnosis and identification of heart diseases of pet cats, and the risk level of potential sudden cardiac death can be predicted, so that timely and effective prevention measures can be taken, the life quality of pets can be improved, and scientization and standardization of pet medical services can be promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of disease risk assessment, and particularly to a method for assessing the risk of sudden cardiac death in pet cats. Background Art

[0002] Currently, it is considered that the common high-fatality diseases in pet cats mainly include: heart disease, chronic kidney disease, feline breast cancer, feline infectious peritonitis, and diseases caused by feline leukemia virus and feline immunodeficiency virus. It is worth noting that among all sudden death cases, the incidence of sudden cardiac death is the highest, accounting for about 85%. Sudden cardiac death refers to sudden, non-traumatic death caused by heart disease. Affected cats usually die rapidly within one hour after the onset of acute symptoms without warning. Such death events are unexpected and occur rapidly. Common causes may include coronary heart disease, myocardial infarction, and hypertrophic cardiomyopathy, etc.

[0003] Obviously, timely control of heart diseases in pet cats can effectively reduce the sudden death rate. However, early screening for feline heart diseases faces many difficulties. First, feline heart diseases lack obvious early symptoms. Many cats with heart diseases may not show any obvious clinical symptoms in the early stage of the disease. Moreover, cats are naturally good at hiding diseases. Therefore, even if their heart function has begun to decline, they may still appear normal. Second, considering that the screening for heart diseases usually requires relatively expensive medical tests, such as echocardiogram, electrocardiogram, and X-ray examination, these tests are not only costly but also require specialized equipment and trained professionals to operate. Pet owners may lack knowledge about the risk and early symptoms of heart diseases. Therefore, they may not seek veterinary help before the disease progresses to a later stage. Third, compared with human medicine, research in some aspects of veterinary medicine is relatively less. This means that the early screening and treatment guidelines for feline heart diseases may be imperfect or not updated in a timely manner. In view of a series of reasons such as the lack of obvious early symptoms, high diagnostic costs, insufficient awareness of pet owners' protection, and limitations of medical methods, we propose a method for assessing the risk of sudden cardiac death in pet cats to help early diagnosis and differentiation of feline heart diseases and predict the risk of potential sudden cardiac death through comprehensive analysis of routine physical examination data. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for assessing the risk of sudden cardiac death in pet cats to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A method for assessing the risk of sudden cardiac death in pet cats, comprising:

[0007] Collecting physical examination case information of pet cats, the physical examination case information includes: name, gender, weight, breed, age, routine blood cell test information, and routine blood biochemical test information;

[0008] Based on the collected physical examination case information of pet cats, a risk assessment model for sudden cardiac death of pet cats was constructed to screen out characteristic indicators related to sudden cardiac death of pet cats;

[0009] The test report data of the pet cat to be evaluated is obtained, wherein the test report data includes characteristic indicators related to sudden cardiac death of the pet cat, and the test report data is input into the pet cat sudden cardiac death risk assessment model to obtain the pet cat sudden cardiac death risk assessment result.

[0010] Preferably, the physical examination case information is divided into three levels of high risk, medium risk and low risk according to the physical examination results. Cases that have been diagnosed with heart disease are classified as high-risk cases; cases that have not been diagnosed but have shown relevant indications in physical examination are classified as medium-risk cases; and cases that are healthy during physical examination are classified as low-risk cases.

[0011] Preferably, the physical examination case information needs to be cleaned and preprocessed after collection, including:

[0012] Eliminate features and samples with too large missing proportions;

[0013] Missing data were handled by filling in the gaps to ensure the completeness of the sample;

[0014] The data is normalized and standardized to ensure that features of different dimensions can play a balanced role in the model.

[0015] Preferably, the processing of missing data by filling in the missing data comprises:

[0016] Collect the pet cat physical examination case information, randomly number it, and arrange it in sequence according to the number;

[0017] Find the K data points closest to the missing value and use the average of these points to fill the missing value, where the value range of K is 10 to 20.

[0018] Preferably, the construction of a risk assessment model for sudden cardiac death in pet cats includes:

[0019] The cleaned and preprocessed data set is randomly divided into a training set and a test set in a ratio of 4:1. The training set is used for model parameter fitting, and the test set is used for evaluating the model effect.

[0020] The XGBoost algorithm is used to fit the model to the training set data. By weighting each input metric, it learns the impact of different metric data on the risk of heart disease, integrates and constructs multiple decision trees, and finally outputs the classification result of the risk of sudden cardiac death;

[0021] The model effect is evaluated using the test set data to assess the generalization ability of the model.

[0022] Preferably, the use of the XGBoost algorithm to fit the model to the training set data includes:

[0023] The objective function of the model uses the Multiclass Log Loss:

[0024]

[0025] where n is the number of samples, and y im is an indicator variable. If sample i belongs to class m, then y im = 1, otherwise 0; is the probability that the model predicts that sample i belongs to class m; T is the number of trees; Ω(f t ) is the regularization term of the t-th tree, which is used to control the complexity of the model and prevent overfitting;

[0026] The regularization function is:

[0027]

[0028] where w t is the weight of the t-th tree, and γ and λ are regularization parameters;

[0029] In each iteration, XGBoost adds a new tree f t (x) to fit the negative gradient of all previous trees. The learning objective of the t-th tree is:

[0030]

[0031] where is the sum of the prediction scores of all previous trees before adding the t-th tree, is the contribution of the t-th tree to sample i belonging to class m;

[0032] When XGBoost constructs a decision tree, by calculating the gain of each feature and split point, it selects the feature and split point with the largest gain for tree splitting. The calculation of the gain considers the gradient and second derivative:

[0033]

[0034] where gj is the sum of gradients, H j is the sum of second-order derivatives, i.e., the sum of the diagonal elements of the Hessian matrix, and λ is the regularization parameter for the sum of squared gradients at leaf nodes;

[0035] XGBoost uses the learning rate η in each step to reduce the contribution of newly added trees to improve the generalization ability of the model:

[0036] f(x) = argmin f [L(f(x)) + ηΩ(f)]

[0037] where L(f(x)) is the loss function and η is the learning rate;

[0038] Finally, the predicted output of the XGBoost model is the sum of the prediction scores of all trees:

[0039]

[0040] Preferably, the model effect is evaluated using the test set data, and the evaluation metrics include accuracy, recall rate, F1-score, and confusion matrix.

[0041] Preferably, the risk assessment model selects ten characteristic indicators related to feline cardiogenic sudden death through training, and in order of importance, they are body weight, age, total platelet count, total monocyte count, creatinine ratio, total neutrophil count, total white blood cell count, total red blood cell count, urea, and total lymphocyte count.

[0042] Preferably, after being trained, the risk assessment model also needs to be verified using actual physical examination samples, including the following steps:

[0043] Input the physical examination data of the actual physical examination samples into the risk assessment model to obtain the cardiogenic sudden death risk assessment results of the actual physical examination samples;

[0044] Use echocardiography technology to further examine the actual physical examination samples to verify whether they are consistent with the cardiogenic sudden death risk assessment results of the actual physical examination samples.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] 1. This method for assessing the risk of feline cardiogenic sudden death can provide scientific support for the early diagnosis and differentiation of feline heart diseases, predict the risk level of potential cardiogenic sudden death, and thus take timely and effective preventive measures, which can not only improve the quality of life of pets but also promote the scientific and standardized development of pet medical services.

[0047] 2. This method for assessing the risk of sudden cardiac death in pet cats focuses on this specific group of pet cats and assesses and predicts the risk of heart diseases in pet cats, with high market demand and social value.

[0048] 3. This method for assessing the risk of sudden cardiac death in pet cats provides a practical and convenient health management method for pet owners, veterinarians, and stray animal shelter organizations, which helps to promote research results to a wider market, enabling more pets and their caregivers to benefit. At the same time, it also assists in promoting the social influence of stray animal shelter organizations and providing help for more stray small animals. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic flow chart of the method for assessing the risk of sudden cardiac death in pet cats of the present invention;

[0050] Figure 2 It is a predicted confusion matrix diagram for each category of the test set of the risk assessment model in the present invention;

[0051] Figure 3 It is 10 features related to heart diseases in pet cats trained by the risk assessment model in the present invention;

[0052] Figure 4 It is a conventional examination report diagram of Case 1 in the embodiment of the present invention;

[0053] Figure 5 It is a conventional examination report diagram of Case 2 in the embodiment of the present invention;

[0054] Figure 6 It is a cardiac color Doppler ultrasound examination report diagram of Case 1 in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0056] With the improvement of living standards, more and more families choose to raise pets, which has promoted the rapid growth of the pet market. The "2022 China Pet Consumption Report" shows that the scale of the urban pet consumption market in China reached 270.6 billion yuan in 2022, with an annual growth rate of 8.7%, and the number of pet owners increased to 70.43 million, a growth of 2.9%. In contrast, the scale of the pet market in the United States has reached 123.6 billion US dollars, highlighting the huge development space and potential of the pet industry in China. In the field of pet health, using artificial intelligence technology to achieve comprehensive, accurate, and real-time monitoring of pet health has become a major trend.

[0057] In addition, according to the "2023 White Paper on Pet Health Consumption in China", pet medical and health expenditures account for 93.6% of the pet market consumption structure, indicating the growing demand for pet health insurance and preventive medical services. Using the prediction results of the model of this application, insurance companies can more accurately assess risks and formulate reasonable insurance policies, which not only reduces the claim risk but also provides a more fair and reasonable insurance service.

[0058] In summary, the method for assessing the risk of feline cardiogenic sudden death of this application will effectively promote the modernization level of pet health management, improve the quality of medical services, and promote the rational development of the pet insurance market, and is expected to play an important role in the early diagnosis and preventive treatment of pet heart diseases.

[0059] Please refer to Figure 1 , a technical solution provided by the present invention:

[0060] A method for assessing the risk of feline cardiogenic sudden death, comprising the following steps S100 to S300:

[0061] S100. Collect the physical examination case information of pet cats, and the physical examination case information includes: name, gender, weight, breed, age, routine blood cell test information, and routine blood biochemistry test information.

[0062] The data mainly comes from a pet hospital in Zhenjiang City. The entire data set includes the physical examination information of 503 pet cats, which can be specifically divided into the following three categories:

[0063] ① Basic pet information: including the name, gender, weight, breed, age, etc. of pet cats. Although this kind of information is basic data, it also has a certain correlation with disease risks. For example, pet cats of different breeds or age groups may face different disease risks.

[0064] ② Routine blood cell test information: including red blood cell count, white blood cell count, neutrophil count, platelet count, etc. These routine blood test indicators are very crucial in evaluating the health status of pet cats, especially helpful for identifying potential diseases. For example, abnormal red blood cell count may indicate anemia, while abnormal white blood cell count may suggest infection or inflammatory response, and these factors may be related to the risk of heart diseases.

[0065] ③ Routine blood biochemistry test information: including more detailed and complex test information such as nucleated red blood cells, abnormal lymphocytes, and large immature cells. This kind of information can often provide a more in-depth analysis of the health status and help reveal more complex health problems. For example, the appearance of large immature cells may reflect abnormalities in the immune or hematopoietic system of pet cats, which may be related to the pathogenesis of heart diseases.

[0066] In some embodiments, the physical examination case information is divided into three levels: high risk, medium risk, and low risk according to the physical examination results. Cases that have been diagnosed with heart disease are classified as high-risk cases; cases that have not been diagnosed but have shown relevant indicators during the physical examination are classified as medium-risk cases; cases with a healthy physical examination are classified as low-risk cases, which is convenient for constructing a subsequent risk assessment model for sudden cardiac death in pet cats.

[0067] Since the data is sourced from the actual clinical cases of pet hospitals, there will inevitably be some data missing during the data collection process, and there may also be deviations in the case information entered manually. Therefore, the physical examination case information needs to be cleaned and preprocessed after collection, including but not limited to the following steps S110 - S130:

[0068] S110. Eliminate features and samples with an overly large missing ratio to avoid interference of these data on model training;

[0069] S120. Process the missing data through filling methods to ensure the integrity of the samples;

[0070] S130. Normalize and standardize the data to ensure that features with different dimensions can play a balanced role in the model.

[0071] In some embodiments, the process of processing the missing data through filling methods includes but not limited to the following steps S121 - S122:

[0072] S121. Collect the random number numbers of the physical examination case information of pet cats and arrange them in sequence according to the number numbers;

[0073] S122. Find the K data points closest to the missing value and fill the missing value with the average value of these points, where the value range of K is 10 - 20.

[0074] S200. According to the collected physical examination case information of pet cats, construct a risk assessment model for sudden cardiac death in pet cats and screen out the characteristic indicators related to sudden cardiac death in pet cats.

[0075] In some embodiments, the construction of the risk assessment model for sudden cardiac death in pet cats includes the following steps S210 - S230:

[0076] S210. Randomly divide the cleaned and preprocessed data set into a training set and a test set according to a ratio of 4:1, where the training set has 402 samples and the test set has 101 samples. The training set is used for parameter fitting of the model, and the test set is used for evaluation of the model effect;

[0077] S220. Use the XGBoost algorithm to perform model fitting on the training set data. By weighting each input metric, learn the impact of different metric data on the risk of heart disease, integrate and construct multiple decision trees, and finally output the classification result of the risk of sudden cardiac death.

[0078] The training process of XGBoost is based on the framework of Gradient Boosting Machine (GBM). Its core idea is to build multiple weak classifiers (decision trees), that is, train a decision tree for each class separately, and combine the outputs of these trees to form the final classification result. This method of ensemble learning can obtain a more powerful classifier than a single decision tree.

[0079] Specifically, the Gradient Boosting algorithm is an iterative algorithm. It starts from an initial model and then adds a new model in each step to reduce the residuals of all previous models. In this application, the model classifies pet cats into low-risk, medium-risk, and high-risk samples according to the input features. This problem belongs to a multi-classification problem, and the objective function of the model uses the Multiclass Log Loss:

[0080]

[0081] where n is the number of samples, y im is an indicator variable. If sample i belongs to class m, then y im = 1, otherwise 0; is the probability that the model predicts that sample i belongs to class m; T is the number of trees; Ω(f t ) is the regularization term of the t-th tree, which is used to control the complexity of the model and prevent overfitting;

[0082] The regularization function is:

[0083]

[0084] where w t is the weight of the t-th tree, and γ and λ are regularization parameters;

[0085] In each iteration step, XGBoost adds a new tree f t (x) to fit the negative gradient of all previous trees. The learning objective of the t-th tree is:

[0086]

[0087] where is the sum of the prediction scores of all previous trees before adding the t-th tree, is the contribution of the t-th tree to the sample i belonging to the class m;

[0088] When constructing a decision tree in XGBoost, by calculating the gain of each feature and split point, the feature and split point with the largest gain are selected for tree splitting. The calculation of the gain considers the gradient and the second derivative:

[0089]

[0090] where g i is the sum of gradients, and H j is the sum of second derivatives, that is, the sum of the diagonal elements of the Hessian matrix, and λ is the regularization parameter of the sum of squared gradients on the leaf nodes;

[0091] XGBoost uses the learning rate η in each step to reduce the contribution of the newly added tree to improve the generalization ability of the model:

[0092] f(x) = argmin f [L(f(x)) + ηΩ(f)]

[0093] where L(f(x)) is the loss function and η is the learning rate;

[0094] Finally, the predicted output of the XGBoost model is the sum of the predicted scores of all trees:

[0095]

[0096] Specifically, first use the xgb.DMatrix function to convert the input dataset into the DMatrix format. DMatrix is a dedicated data structure in XGBoost for efficiently processing input data. It is faster than ordinary arrays or data frames in processing speed and can better manage sparse data. Then, configure the model parameters for the specific problem. After debugging, the specific parameter configuration is as follows:

[0097] params = {

[0098] 'objective':'multi:softmax',

[0099] 'num_class': 3,

[0100] 'max_depth': 4,

[0101] 'eta': 0.1,

[0102] 'eval_metric':'mlogloss'

[0103] }

[0104] Among them, objective specifies the optimization objective, which is multi:softmax here, indicating multi-classification and using logistic regression to output probabilities; max_depth represents the maximum depth of constructing the decision tree, and this value controls the complexity of each tree. The greater the depth, the more complex the model, but it is also more likely to overfit. The value here is 4, indicating that the maximum depth of each tree is 4; eta represents the learning rate, which controls the contribution of each tree to the final prediction. A smaller learning rate can make the model more stable, but requires more iteration times. It is set to 0.1 here, indicating that the impact of each iteration on the final prediction is small; eval_metric represents the evaluation metric, which is mlogloss here, that is, multi-class logarithmic loss, and is commonly used in multi-class problems to evaluate the model performance.

[0105] Finally, based on the configured parameters, use the xgb.train function to train the model.

[0106] S230. Use the test set data to evaluate the model effect and evaluate the generalization ability of the model. The evaluation metrics for using the test set data to evaluate the model effect include precision, recall rate, F1-score, and confusion matrix.

[0107] After a certain number of iteration rounds, use the trained model to verify on the test set. Since the optimization objective is set to multi:softmax, the model can output the probabilities of the samples belonging to each category. In the test set, the prediction accuracy of the model is 97.03%, and the prediction confusion matrix of each category is as Figure 2 shown.

[0108] The classification report of the test set provides detailed information on multiple performance metrics such as precision, recall rate, and F1-score, and can be used as a detailed evaluation tool for the model performance.

[0109]

[0110] After model training, 10 features related to pet cat heart diseases are selected, as Figure 3 shown. The order of feature importance is weight, age, total platelet count, total monocyte count, creatinine ratio, total neutrophil count, total white blood cell count, total red blood cell count, urea, and total lymphocyte count in turn.

[0111] After the risk assessment model is trained, it also needs to be verified using actual physical examination samples, including the following steps:

[0112] Input the physical examination data of the actual physical examination samples into the risk assessment model to obtain the risk assessment results of sudden cardiac death for the actual physical examination samples;

[0113] Use echocardiography technology to further examine the actual physical examination samples, and verify whether the results of the risk assessment of sudden cardiac death in the actual physical examination samples are consistent.

[0114] After completing model training and parameter adjustment, two subsequent physical examination cases obtained Figure 4 , Case 1; Figure 5 , Case 2) are used for early diagnosis and differentiation of heart diseases. Although the five-category blood routine and biochemical test data of both cases do not exceed the normal value range shown in the figure, when we predict the risk levels of their potential sudden cardiac death based on 10 characteristics related to pet cat heart diseases (marked with red lines in the report), the results show that: Case 1 is at high risk, mainly manifested by a relatively high incidence of sudden cardiac death, and pet owners need to conduct regular physical examinations, pay attention to diet management, and make preparations for emergencies. Case 2 is at low risk, that is, it can be considered in a healthy state. Although there are several abnormal indicators, they have no direct relationship with heart diseases, and the probability of sudden cardiac death is relatively low. Pet owners only need to do regular physical examinations.

[0115] Subsequently, the echocardiogram report of Case 1 was further verified, as Figure 6 shown. Although the five-category blood routine and biochemical test data show normal, the echocardiogram diagnosed focal thickening of the papillary muscle of the cat's heart, and there is a possibility of sudden cardiac death, which is consistent with the prediction result of our model.

[0116] S300. Obtain the test report data of the pet cat to be evaluated. The test report data includes characteristic indicators related to sudden cardiac death of the pet cat, and input them into the risk assessment model of sudden cardiac death of the pet cat to obtain the risk assessment result of sudden cardiac death of the pet cat.

[0117] The risk assessment method for sudden cardiac death of pet cats in this embodiment can provide scientific support for the early diagnosis and differentiation of heart diseases in pet cats, predict the risk levels of potential sudden cardiac death, and thus take timely and effective preventive measures, which can not only improve the quality of life of pets, but also promote the scientific and standardized development of pet medical services.

[0118] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for assessing the risk of sudden cardiac death in pet cats, characterized in that: include: Collecting physical examination case information of pet cats, the physical examination case information includes: name, gender, weight, breed, age, routine blood cell test information, and routine blood biochemical test information; Based on the collected physical examination case information of pet cats, a risk assessment model for sudden cardiac death of pet cats was constructed to screen out characteristic indicators related to sudden cardiac death of pet cats; The test report data of the pet cat to be evaluated is obtained, wherein the test report data includes characteristic indicators related to sudden cardiac death of the pet cat, and the test report data is input into the pet cat sudden cardiac death risk assessment model to obtain the pet cat sudden cardiac death risk assessment result.

2. The method for assessing the risk of sudden cardiac death in pet cats according to claim 1, characterized in that: The physical examination case information is divided into three levels: high risk, medium risk and low risk according to the physical examination results. Cases that have been diagnosed with heart disease are classified as high-risk cases; cases that have not been diagnosed but have shown relevant indications in physical examinations are classified as medium-risk cases; and cases that are healthy during physical examinations are classified as low-risk cases.

3. The method for assessing the risk of sudden cardiac death in pet cats according to claim 1, characterized in that: The physical examination case information needs to be cleaned and preprocessed after collection, including: Eliminate features and samples with too large missing proportions; Missing data were handled by filling in the gaps to ensure the completeness of the sample; The data is normalized and standardized to ensure that features of different dimensions can play a balanced role in the model.

4. The method for assessing the risk of sudden cardiac death in pet cats according to claim 3, characterized in that: The missing data are processed by filling in the missing data, including: Collect the pet cat physical examination case information, randomly number it, and arrange it in sequence according to the number; Find the K data points closest to the missing value and use the average of these points to fill the missing value, where the value range of K is 10 to 20.

5. The method for assessing the risk of sudden cardiac death in pet cats according to claim 3, characterized in that: The construction of a risk assessment model for sudden cardiac death of pet cats comprises: The cleaned and preprocessed data set is randomly divided into a training set and a test set in a ratio of 4:

1. The training set is used for model parameter fitting, and the test set is used for evaluating the model effect. The XGBoost algorithm is used to fit the model to the training set data. By weighting each input indicator, the impact of different indicator data on the risk of heart disease is learned, and multiple decision trees are integrated to finally output the classification results of sudden cardiac death risk. Use the test set data to evaluate the model effect and assess the generalization ability of the model.

6. The method for assessing the risk of sudden cardiac death in pet cats according to claim 5, characterized in that: The use of the XGBoost algorithm to perform model fitting on the training set data includes: The objective function of the model uses multi-class log likelihood loss (Multiclass Log Loss): Where n is the number of samples, y im is an indicator variable. If sample i belongs to category m, then y im =1, otherwise 0; is the probability that the model predicts that sample i belongs to category m; T is the number of trees; Ω(f t ) is the regularization term of the tth tree, which is used to control the complexity of the model and prevent overfitting; The regularization function is: Among them, w t is the weight of the tth tree, γ and λ are regularization parameters; At each iteration, XGBoost adds a new tree f t (x) is used to fit the negative gradients of all previous trees. The learning goal of the tth tree is: in, is the sum of the prediction scores of all trees before adding the tth tree, is the contribution of the t-th tree to sample i belonging to category m; When XGBoost builds a decision tree, it calculates the gain of each feature and split point, selects the feature and split point with the largest gain to split the tree, and the calculation of the gain takes into account the gradient and the second-order derivative: Among them, g j is the gradient and H j is the sum of the second-order derivatives, that is, the sum of the diagonal elements of the Hessian matrix, and λ is the regularization parameter of the sum of squared gradients at the leaf nodes; XGBoost uses a learning rate η at each step to reduce the contribution of the newly added trees to improve the generalization ability of the model: f(x)=argmin f [L(f(x))+ηΩ(f)] Where L(f(x)) is the loss function and η is the learning rate; Ultimately, the prediction output of the XGBoost model is the sum of all the tree prediction scores:

7. The method for assessing the risk of sudden cardiac death in pet cats according to claim 6, characterized in that: The test set data is used to evaluate the model effect, and the evaluation indicators include precision, recall rate, F1-score, and confusion matrix.

8. The method for assessing the risk of sudden cardiac death in pet cats according to claim 7, characterized in that: The risk assessment model was trained to screen out ten characteristic indicators related to sudden cardiac death in pet cats, which were weight, age, total platelet count, total monocyte count, urinary anhydride ratio, total neutrophil count, total white blood cell count, total red blood cell count, urea and total lymphocyte count in order of importance.

9. The method for assessing the risk of sudden cardiac death in pet cats according to claim 8, characterized in that: After the risk assessment model is trained, it needs to be verified using actual physical examination samples, including the following steps: Inputting the physical examination data of the actual physical examination sample into the risk assessment model to obtain the sudden cardiac death risk assessment result of the actual physical examination sample; The actual physical examination samples were further examined using cardiac color ultrasound technology to verify whether the sudden cardiac death risk assessment results were consistent with those of the actual physical examination samples.

Images