Method for constructing a maxillary sinus cyst prediction model based on machine learning algorithm, maxillary sinus cyst prediction method and device, and storage medium

Abstract

The present application relates to the technical field of computer data processing, and discloses a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, a maxillary sinus cyst prediction method and device, and a storage medium, wherein the construction method includes obtaining clinical data of patients, forming a data set, and performing data preprocessing on the data set; using chi-square test and LASSO regression method to screen characteristic variables in the data set, and constructing a training set, a validation set, and a test set based on the screened characteristic variables; constructing an initial prediction model based on an extreme gradient boosting algorithm; training the initial prediction model based on the training set, the validation set, and the test set, and iteratively adding trees, each time adding a tree to fit the residual of the previous round of prediction, and finally obtaining a maxillary sinus cyst prediction model whose root mean square error meets the preset requirements. This solves the problem in the prior art that the risk factors of maxillary sinus cysts need to be determined and the occurrence of maxillary sinus cysts needs to be predicted.

Description

Technical Field

[0001] The present application relates to the field of computer data processing technology, and specifically to a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, a maxillary sinus cyst prediction method and device, and a storage medium. Background Art

[0002] Maxillary sinus cysts are one of the types of maxillary sinus mucosal lesions, accounting for 89.5%-92.7% of all paranasal sinus cysts, with an incidence rate of 3.6%-35.6%. Maxillary sinus cysts are mostly asymptomatic in the early stage. After the cysts enlarge and fill the sinus cavity, symptoms such as cheek stuffiness, pressing pain, and toothache on the same side may occur. If the cysts enlarge and compress the nasal septum, it may cause nasal septum deviation; if the cysts block the maxillary sinus ostium, sinusitis-related symptoms such as nasal congestion and runny nose may occur. In clinical maxillary sinus floor lift surgery, in order to effectively treat maxillary sinus cysts, it is necessary to fully understand their pathogenic factors. Previous studies have reported anatomical factors and odontogenic factors related to maxillary sinus cysts, but no unified conclusion has been reached. In addition, the intricate nonlinear relationship between various influencing factors and maxillary sinus cysts poses a huge challenge to the application of traditional linear statistical methods. Considering the high incidence of maxillary sinus cysts, unclear influencing factors, impact on patients' quality of life and troubles for doctors, in order to improve the accuracy of prediction and early intervention of maxillary sinus cysts and improve patients' quality of life, it is urgent to identify the risk factors of maxillary sinus cysts and predict the occurrence of maxillary sinus cysts. Summary of the invention

[0003] The purpose of the present application is to provide a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, a maxillary sinus cyst prediction method and device, and a storage medium, so as to solve the problem in the prior art of needing to determine the risk factors of maxillary sinus cysts and predict the occurrence of maxillary sinus cysts.

[0004] To achieve the above-mentioned purpose, the embodiment of the present application provides a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, comprising the following steps: acquiring clinical data of a patient to form a data set, and performing data preprocessing on the data set;

[0005] Using chi-square test and LASSO regression method to screen the characteristic variables in the data set, and constructing a training set, a validation set and a test set based on the screened characteristic variables;

[0006] Build an initial prediction model based on the extreme gradient boosting algorithm;

[0007] The initial prediction model is trained based on the training set, validation set and test set, and trees are added iteratively, one tree is added each time to fit the residual of the previous round of prediction, and finally a maxillary sinus cyst prediction model with a root mean square error that meets the preset requirements is obtained.

[0008] Optionally, the obtaining of clinical data of the patient specifically includes:

[0009] The data on influencing factors of maxillary sinus cysts were obtained, including: age, gender, dentition, pulp condition, periodontal condition, different sides of the maxillary sinus, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0010] Optionally, the data preprocessing of the data set specifically includes:

[0011] Remove outliers and missing values ​​from the data set, and fill in the missing data by interpolation;

[0012] Normalize or standardize the numerical variables in the data set.

[0013] Optionally, the characteristic variables in the screening data set specifically include:

[0014] The categorical variables in the data set were expressed as numbers and percentages, and the chi-square test was used to screen out categorical variables with a two-sided P value less than 0.05;

[0015] LASSO regression was used to further screen feature variables, including: compressing the variable coefficients in the regression model by generating a penalty function to prevent overfitting, and changing the coefficients of the related independent variables to 0 through the correlation between the independent variables. LASSO regression set 10-fold cross validation and was executed through the R package.

[0016] Optionally, the screened characteristic variables include: gender, age, pulp condition, periodontal condition, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, whether maxillary sinusitis exists, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0017] Optionally, the parameters of the maxillary sinus cyst prediction model include: a learning rate of 0.1, a maximum tree depth of 6, a minimum bifurcation weight sum of 1, and an L2 regularization coefficient of 0.1.

[0018] To achieve the above purpose, the present application also provides a method for predicting maxillary sinus cysts, comprising the steps of:

[0019] Obtain relevant feature values ​​of the patient and input them into the maxillary sinus cyst prediction model constructed by the method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm as described in any of the preceding claims, wherein the feature values ​​include at least one of gender, age, pulp condition, periodontal condition, relationship between tooth root and maxillary sinus floor, maxillary sinus floor morphology, maxillary sinus lateral wall blood vessels, maxillary sinus mucosal condition, presence of maxillary sinusitis, maxillary sinus development, and maxillary sinus floor cortical bone condition;

[0020] The prediction model for maxillary sinus cyst outputs a prediction result and gives the probability of a patient developing a maxillary sinus cyst.

[0021] To achieve the above-mentioned purpose, the present application also provides a device for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, comprising: a memory; and,

[0022] A processor connected to the memory, the processor being configured to execute the steps of the method as described above.

[0023] To achieve the above objectives, the present application also provides a computer storage medium on which a computer program is stored, wherein the computer program implements the steps of the above method when executed by a machine.

[0024] The embodiments of the present application have the following advantages:

[0025] Through the above method, based on large-scale patient data, using chi-square test, LASSO regression and other methods, the risk factors of maxillary sinus cysts were screened out, and an efficient maxillary sinus cyst prediction model was constructed with the help of machine learning algorithm. The extreme gradient boosting algorithm has excellent performance, with the average area under the curve (AUC) values ​​of 0.939, 0.923 and 0.921 in the training set, validation set and test set, respectively. The maxillary sinus cyst prediction model constructed based on the XGBoost model can effectively predict the probability of maxillary sinus cysts in the future, thus solving the problem of determining the risk factors of maxillary sinus cysts and predicting the occurrence of maxillary sinus cysts in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings described below are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0027] Figure 1 A flowchart of a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0028] Figure 2 A schematic diagram of a LASSO coefficient path diagram of a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0029] Figure 3 A schematic diagram of a LASSO regression analysis cross-validation curve for a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0030] Figure 4 A schematic diagram of a training set ROC curve for comparing multiple machine learning models of a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0031] Figure 5 A schematic diagram of a validation set ROC curve for comparing multiple machine learning models of a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0032] Figure 6 A schematic diagram of an XGBoost algorithm model training set ROC curve for a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0033] Figure 7 A schematic diagram of an XGBoost algorithm model validation set ROC curve for a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application;

[0034] Figure 8 A SHAP summary diagram of a method for predicting a maxillary sinus cyst provided by at least one embodiment of the present application;

[0035] Fig. 9 A SHAP feature importance graph of a maxillary sinus cyst prediction method provided by at least one embodiment of the present application;

[0036] Fig.10 A SHAP graph of a sample of a method for predicting a maxillary sinus cyst provided by at least one embodiment of the present application;

[0037] Fig.11 A SHAP graph of another sample of a method for predicting a maxillary sinus cyst provided by at least one embodiment of the present application;

[0038] Fig.12 A module block diagram of a device for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application. DETAILED DESCRIPTION

[0039] The following is a description of the implementation of the present application by specific specific embodiments. People familiar with the technology can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0040] It should be noted that the steps in the claims and description of the present application may be executed substantially in parallel or in reverse order under appropriate circumstances, depending on the functions involved.

[0041] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0042] An embodiment of the present application provides a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, referring to Figure 1 , Figure 1 The present invention provides a flowchart of a method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm in at least one embodiment of the present application. It should be understood that the method may also include additional boxes not shown and / or the boxes shown may be omitted, and the scope of the present application is not limited in this respect.

[0043] In step 101 , clinical data of a patient is acquired to form a data set, and data preprocessing is performed on the data set.

[0044] In some embodiments, obtaining the patient's clinical data specifically includes:

[0045] Obtain data on factors affecting maxillary sinus cysts, including age, gender, dentition, pulp, periodontal conditions, different sides of the maxillary sinus, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0046] In some embodiments, the data preprocessing of the data set specifically includes:

[0047] Data cleaning: remove outliers and missing values ​​in the data set, and fill in the missing data through interpolation;

[0048] Standardization: Normalize or standardize the numerical variables in the data set to ensure that different feature quantities have the same scale.

[0049] At step 102, the characteristic variables in the data set are screened using the chi-square test and the LASSO regression method, and a training set, a validation set, and a test set are constructed based on the screened characteristic variables.

[0050] In some embodiments, the feature variables in the screening data set specifically include:

[0051] The characteristic variables included in this application are all categorical variables, including: age, gender, dentition, pulp, periodontal, different sides of the maxillary sinus, the relationship between the root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side wall, the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone of the maxillary sinus floor. The categorical variables in the data set are expressed as numbers and percentages, and the chi-square test is used to screen the categorical variables with bilateral P values ​​less than 0.05, which are considered to be statistically significant. The chi-square test analysis results refer to Table 1.

[0052] Table 1:

[0053]

[0054] refer to Figure 2 and Figure 3 , LASSO regression is used to further screen feature variables. Specifically, LASSO regression compresses the variable coefficients in the regression model by generating a penalty function to prevent overfitting; at the same time, LASSO regression changes the coefficients of related independent variables to 0 through the correlation between independent variables, thereby reducing the impact of multicollinearity on the regression results and solving serious collinearity problems. LASSO regression sets 10-fold cross validation and is executed through the R package "glmnet4.1.2".

[0055] exist Figure 2 In the figure, the upper horizontal axis represents the number of non-zero coefficients in the model, the vertical axis represents the value of the coefficient, the lower horizontal axis represents the standardized coefficient vector, and the 11 lines of different colors represent 11 variables, and each curve represents the change trajectory of each independent variable coefficient. Figure 3 The middle ordinate represents the cross-validation error (the smaller the ordinate axis, the better the LASSO fitting result), the upper abscissa represents the number of variables corresponding to different eigenvalues ​​λ, and the lower abscissa represents the logarithm of the λ penalty coefficient. The parameter (lambda.min) corresponding to the left dotted line in the graph has the smallest error, which determines how many variables can be used for analysis. The corresponding upper abscissa value is 11, that is, 11 variables are left.

[0056] In some embodiments, the characteristic variables finally screened out include: gender, age, pulp condition, periodontal condition, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0057] At step 103, an initial prediction model is constructed based on the extreme gradient boosting algorithm.

[0058] At step 104, the initial prediction model is trained based on the training set, the validation set and the test set, by iteratively adding trees, each time adding a tree to fit the residual of the previous round of prediction, and finally obtaining a maxillary sinus cyst prediction model whose root mean square error meets the preset requirements.

[0059] Specifically, the initial prediction model was constructed based on a variety of machine learning methods such as Logistic Regression (LR), eXtremeGradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Random Forest (RF), Adaptive Boosting (AdaBoost), Gaussian Naive Bayes (GNB), ComplementNaive Bayes (CNB), Multilayer Perceptron (MLP) and K-NearestNeighbors (KNN). The parameters of the Logistic Regression model are regularization factor: 1.0, number of iterations: 100, regularization type: l2, and convergence metric: 0.0001. The parameters of the LightGBM model are algorithm type: gbdt, learning rate: 0.1, maximum tree depth: -1, maximum number of trees: 100, and maximum number of leaves: 31. The parameters of the Random Forest model are metric: gini, maximum tree depth: None, minimum fork purity gain: 0.0, number of trees: 100. The parameters of the AdaBoost model are learning rate: 1.0, number of single models: 50. The parameters of the KNN model are number of neighbors: 5, weight type: uniform. The parameters of the SVM model are regularization factor: 1.0, kernel type: rbf, convergence metric: 0.001. The parameters of the GNB model are prior probability: None, var_smoothing: 1e-09. The parameters of the MLP model are nonlinear function: relu, hidden layer width: (20, 10), number of iterations: 200. XGBoost is determined to be the optimal model through the AUC value evaluation of the training set, validation set, and test set. Reference Figure 4 and Figure 5 , Figure 4 and Figure 5 The ROC curves of the training set and validation set are respectively used to compare multiple machine learning models. The ROC curve is a curve obtained by plotting the true positive rate and the false positive rate, which can be used to reflect the relationship between sensitivity and specificity. The AUC value represents the area under the ROC curve and the coordinate axis. The larger the AUC value, the better the effect of the machine learning model, that is, the higher the discrimination. Figure 6 and Figure 7 They are respectively a schematic diagram of the ROC curve of the XGBoost algorithm model training set and a schematic diagram of the ROC curve of the XGBoost algorithm model validation set.

[0060] Specifically, the Python function of the training dataset using the XGBoost algorithm model follows the XGBoost objective function and comes with detailed comments explaining the implementation of each step:

[0061] Specific functions:

[0062] import xgboost as xgb

[0063] import numpy as np

[0064] from sklearn.datasets import load_boston

[0065] from sklearn.model_selection import train_test_split

[0066] from sklearn.metrics import mean_squared_error

[0067] def train_xgboost_model(X, y, num_boost_rounds=100):

[0068] """

[0069] The XGBoost model is used to train the dataset.

[0070] parameter:

[0071] X (numpy.ndarray): Feature dataset.

[0072] y (numpy.ndarray): Labeled dataset.

[0073] num_boost_rounds (int): The number of iterations, i.e. the number of trees to train.

[0074] return:

[0075] model (xgb.Booster): The trained XGBoost model.

[0076] dtrain (xgb.DMatrix): DMatrix object for training.

[0077] """

[0078] # Split the data into training and validation sets

[0079] X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42)

[0080] # Convert the data into a DMatrix object, which is the internal data format of XGBoost

[0081] dtrain = xgb.DMatrix(X_train, label=y_train)

[0082] dval = xgb.DMatrix(X_val, label=y_val)

[0083] # Set model parameters

[0084] params = {'objective': 'reg:squarederror', # Use squared error as the objective function

[0085] 'max_depth': 6, # Maximum depth of the tree

[0086] 'eta': 0.3, # learning rate

[0087] 'subsample': 0.8, # The proportion of data used in each iteration

[0088] 'colsample_bytree': 0.8, # The proportion of features used by each tree

[0089] 'eval_metric': 'rmse' # Use root mean square error as the evaluation indicator}

[0090] # Train the model, evals is used to monitor the error of the validation set

[0091] evals = [(dtrain, 'train'), (dval, 'eval')]

[0092] model = xgb.train(params, dtrain, num_boost_round=num_boost_rounds, evals=evals, early_stopping_rounds=10)

[0093] # Return the model and training data

[0094] return model, dtrain

[0095] Code explanation:

[0096] Dataset segmentation and transformation:

[0097] Use train_test_split to split the data into training and validation sets.

[0098] Use xgb.DMatrix to convert data into XGBoost's built-in DMatrix format to improve computing efficiency.

[0099] Parameter settings:

[0100] Objective: Set the objective function. Here we use reg:squarederror (squared error).

[0101] max_depth: The maximum depth of the tree, to control overfitting.

[0102] eta: learning rate, a lower learning rate will make the model converge more slowly, but usually gives better results.

[0103] subsample and colsample_bytree: control the proportion of data and features used by each tree, helping to prevent overfitting.

[0104] eval_metric: Evaluation metric, here we use RMSE (root mean square error).

[0105] Model training:

[0106] Use xgb.train to train the model and monitor the error changes of the training set and validation set through the evals parameter.

[0107] early_stopping_rounds is used for the early stopping strategy. If the error on the validation set does not decrease within the set number of rounds, the training will stop early.

[0108] Prediction and evaluation:

[0109] After the model training is completed, predictions are made on the training set and RMSE is calculated.

[0110] RMSE (Root Mean Square Error) is a common metric used in machine learning to evaluate model prediction errors. It measures the gap between the predicted value and the true value, and gives a higher penalty to predictions with larger gaps. The calculation steps of RMSE are as follows:

[0111] Calculate the error: For each predicted value, calculate the difference between it and the corresponding true value: (e_i =\hat{y}_i - y_i), where (\hat{y}_i) is the (i)th predicted value and (y_i) is the (i)th true value.

[0112] Squared Error: Square each error to eliminate the cancellation of positive and negative errors: ( e_i^2 ).

[0113] Find the mean squared error (MSE): Add all the squared errors and divide by the total number of data points (n) to get the mean squared error:

[0114] [ MSE = \frac{1}{n} \sum_{i=1}^{n} e_i^2 ]

[0115] Root Mean Squared Error (RMSE): Finally, perform a square root operation on the MSE:

[0116] [ RMSE = \sqrt{MSE} ]

[0117] The formula is summarized as: [RMSE = \sqrt{ \frac{1}{n} \sum_{i=1}^{n} (\hat{y}_i - y_i)^2}]

[0118] The smaller the RMSE, the closer the model's predictions are to the actual values, and the better the model performance. It should be noted that the unit of RMSE is the same as that of the target variable, which makes it intuitive.

[0119] Specifically, the initial prediction model constructed based on the XGBoost algorithm model is trained based on the training set, validation set and test set. The initial prediction model is iteratively added with trees, each time adding a tree to fit the residual of the previous round of prediction, and finally a maxillary sinus cyst prediction model is obtained.

[0120] The specific process follows that the objective function Obj is: [\text{Obj}(\theta) = \sum_{i=1}^nl(y_i, \hat{y}i^{(t)}) + \sum{k=1}^T \Omega(f_k)]. Among them, (l) is the loss function, which is used to measure the deviation between the predicted value (\hat{y}_i^{(t)}) and the actual value (y_i); and (\Omega(f_k)) is the regularization term, which is used to control the complexity of the model, in the form of: [\Omega(f_k) = \gamma T + \frac{1}{2} \lambda \sum_{j=1}^T w_j^2]. Among them, (\gamma) and (\lambda) are regularization parameters, (T) is the number of trees, and (w_j) is the node weight of the tree.

[0121] In some embodiments, the maxillary sinus cyst prediction model is expressed as the following formula: [ \hat{y} = \sum_{k=1}^{K} f_k(\mathbf{x}) ]

[0122] Where ( \hat{y} ) is the predicted value, ( K ) is the number of trees, and ( f_k(\mathbf{x}) ) is the output of the ( k )th tree for the input feature ( \mathbf{x} ).

[0123] In some embodiments, the parameters of the maxillary sinus cyst prediction model include: a learning rate of 0.1, a maximum tree depth of 6, a minimum bifurcation weight sum of 1, and an L2 regularization coefficient of 0.1.

[0124] An embodiment of the present application further provides a method for predicting a maxillary sinus cyst, comprising the steps of: obtaining relevant feature values ​​of a patient, and inputting the feature values ​​into a maxillary sinus cyst prediction model constructed by any of the aforementioned embodiments of the method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, wherein the feature values ​​include at least one of gender, age, pulp condition, periodontal condition, relationship between tooth root and maxillary sinus floor, maxillary sinus floor morphology, maxillary sinus lateral wall blood vessels, maxillary sinus mucosal condition, presence of maxillary sinusitis, maxillary sinus development, and maxillary sinus floor cortical bone condition;

[0125] The prediction model for maxillary sinus cyst outputs a prediction result and gives the probability of a patient developing a maxillary sinus cyst.

[0126] In some embodiments, the method further includes: using the SHAP value to explain the maxillary sinus cyst prediction model, showing important features and their influence on the maxillary sinus cyst, and for each feature value (x_j) in the maxillary sinus cyst prediction model, calculating its SHAP value (\phi_j) to reflect the average contribution of different feature values ​​to the prediction results. Figure 8 , the SHAP summary diagram can show the nonlinear relationship between each influencing factor and maxillary sinus cyst, among which the periodontal condition and pulp condition are positively correlated with the occurrence of maxillary sinus cyst, that is, the more severe the periodontal and pulp conditions are, the higher the incidence of maxillary sinus cyst. The SHAP summary diagram combines feature importance with feature effect. Each point on the SHAP summary diagram is the Shapley value of a feature and an instance. The position on the ordinate axis is determined by the feature, and the position on the abscissa axis is determined by the Shapley value. Blue to red represents feature values ​​from low to high. The overlapping points jitter in the direction of the ordinate axis, so the distribution of the Shapley value of each feature can be understood. These features are ranked according to their importance. Reference Fig. 9 The feature importance ranking showed that periodontal condition, age, gender, and pulp condition were the characteristics of the maxillary sinus cyst prediction model, among which periodontal condition was the most important feature, which changed the predicted absolute probability of maxillary sinus cyst by about 13 percentage points on average. Fig.10 and Fig.11 , the model is further explained by two samples, where Fig.10 The corresponding prediction probability of the first sample is 0.988. Age, maxillary sinus mucosa and periodontal condition have a positive effect on the results. Fig.11 The corresponding predicted probability of the second sample is 0.995. The blood vessels on the side wall of the maxillary sinus have a promoting effect on the results, while the presence or absence of maxillary sinusitis has an inhibitory effect on the results.

[0127] Specifically, for a specific feature ( j ), the calculation formula for its Shapley value is:

[0128] [ \phi_j = \sum_{S \subseteq N \setminus {j}} \frac{|S|! (|N|-|S|-1)!}{|N|!} \left[ f_S(\mathbf{x}_S \cup {x_j}) - f_S(\mathbf{x}_S) \right] ]

[0129] in:

[0130] ( N ) is the set of features.

[0131] ( S \subseteq N \setminus { j} ) represents all feature subsets that do not contain feature ( j ).

[0132] ( f_S(\mathbf{x}_S) ) is the prediction made by the model using only the subset of features ( S ).

[0133] ( f_S(\mathbf{x}_S \cup { x_j}) ) is the model prediction for the feature subset ( S ) and feature ( j ) together.

[0134] In the maxillary sinus cyst prediction model, the SHAP value of each feature is calculated and regressed to feature importance. The SHAP library automatically handles these calculations. The following is the Python implementation code:

[0135] import xgboost as xgb

[0136] import shap

[0137] import matplotlib.pyplot as plt

[0138] from sklearn.datasets import load_boston

[0139] from sklearn.model_selection import train_test_split

[0140] # Load the dataset (dataset of maxillary sinus cyst)

[0141] boston = load_boston()

[0142] X, y = boston.data, boston.target

[0143] # Split the dataset

[0144] X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

[0145] # Train the XGBoost model

[0146] Model = xgb.XGBRegressor(objective='reg:squarederror', max_depth=6,eta=0.3, subsample=0.8, colsample_bytree=0.8)

[0147] model.fit(X_train, y_train)

[0148] # Explanation using SHAP values

[0149] explainer = shap.TreeExplainer(model)

[0150] shap_values ​​= explainer.shap_values(X_train)

[0151] # Plot feature importance graph

[0152] shap.summary_plot(shap_values, X_train, feature_names=boston.feature_names)

[0153] # Draw the SHAP value of a single sample

[0154] shap.initjs()

[0155] shap.force_plot(explainer.expected_value, shap_values[0,:], X_train[0,:], feature_names=boston.feature_names)

[0156] in:

[0157] Load and prepare data:

[0158] A dataset of maxillary sinus cysts was used.

[0159] Use train_test_split to split the data into training and test sets.

[0160] Train the XGBoost model:

[0161] Use xgb.XGBRegressor for model training, and select squared error reg:squarederror as the objective function.

[0162] Calculate the SHAP value:

[0163] Create a TreeExplainer object for explaining the XGBoost model.

[0164] Use shap_values ​​to get the SHAP value of the training set data.

[0165] Plot the SHAP values:

[0166] shap.summary_plot shows the overall feature importance.

[0167] shap.force_plot shows the specific impact of the features of a single sample on the prediction results.

[0168] Through the above method, based on large-scale patient data, using chi-square test, LASSO regression and other methods, the risk factors of maxillary sinus cysts were screened out, and an efficient maxillary sinus cyst prediction model was constructed with the help of machine learning algorithm. The extreme gradient boosting algorithm has excellent performance, with the average area under the curve (AUC) values ​​of 0.939, 0.923 and 0.921 in the training set, validation set and test set, respectively. The maxillary sinus cyst prediction model constructed based on the XGBoost model can effectively predict the probability of maxillary sinus cysts in the future, thus solving the problem of determining the risk factors of maxillary sinus cysts and predicting the occurrence of maxillary sinus cysts in the existing technology.

[0169] Fig.12 A block diagram of a device for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm provided in at least one embodiment of the present application. The device includes:

[0170] A memory 201; and a processor 202 connected to the memory 201, wherein the processor 202 is configured to: acquire clinical data of a patient, form a data set, and perform data preprocessing on the data set;

[0171] Using chi-square test and LASSO regression method to screen the characteristic variables in the data set, and constructing a training set, a validation set and a test set based on the screened characteristic variables;

[0172] Build an initial prediction model based on the extreme gradient boosting algorithm;

[0173] The initial prediction model is trained based on the training set, validation set and test set, and trees are added iteratively, one tree is added each time to fit the residual of the previous round of prediction, and finally a maxillary sinus cyst prediction model with a root mean square error that meets the preset requirements is obtained.

[0174] In some embodiments, the processor 202 is further configured to: obtain the clinical data of the patient, specifically including:

[0175] The data on influencing factors of maxillary sinus cysts were obtained, including: age, gender, dentition, pulp condition, periodontal condition, different sides of the maxillary sinus, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0176] In some embodiments, the processor 202 is further configured to: preprocess the data set, specifically including:

[0177] Remove outliers and missing values ​​from the data set, and fill in the missing data by interpolation;

[0178] Normalize or standardize the numerical variables in the data set.

[0179] In some embodiments, the processor 202 is further configured to: the feature variables in the screening data set specifically include:

[0180] The categorical variables in the data set were expressed as numbers and percentages, and the chi-square test was used to screen out categorical variables with a two-sided P value less than 0.05;

[0181] LASSO regression was used to further screen feature variables, including: compressing the variable coefficients in the regression model by generating a penalty function to prevent overfitting, and changing the coefficients of the related independent variables to 0 through the correlation between the independent variables. LASSO regression set 10-fold cross validation and was executed through the R package.

[0182] In some embodiments, the processor 202 is further configured to: the screened characteristic variables include: gender, age, pulp condition, periodontal condition, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side walls, the condition of the maxillary sinus mucosa, whether there is maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor.

[0183] In some embodiments, the processor 202 is further configured to: the parameters of the maxillary sinus cyst prediction model include: a learning rate of 0.1, a maximum tree depth of 6, a minimum bifurcation weight sum of 1, and an L2 regularization coefficient of 0.1.

[0184] The specific implementation method is referred to the aforementioned method embodiment and will not be described in detail here.

[0185] The present application may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present application.

[0186] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0187] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0188] The computer program instructions for performing the operation of the present application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present application.

[0189] Various aspects of the present application are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0190] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0191] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0192] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can also be executed in the opposite order sometimes, depending on the function involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a dedicated hardware-based system that performs the function or action of the specification, or can be realized by a combination of special-purpose hardware and computer instructions.

[0193] Note that, unless otherwise directly stated, all features disclosed in this specification (including any attached claims, abstracts and drawings) may be replaced by alternative features for achieving the same, equivalent or similar purposes. Therefore, unless otherwise explicitly stated, each feature disclosed is only an example of a group of equivalent or similar features. Where used, further, preferably, further and more preferably are simple beginnings for elaborating another embodiment based on the aforementioned embodiment, and the content of further, preferably, further or more preferably followed by the aforementioned embodiment is combined with the complete composition of another embodiment. Several further, preferably, further or more preferably settings following the same embodiment can be arbitrarily combined to form another embodiment.

[0194] Although the present application has been described in detail above with general descriptions and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements may be made to the present application. Therefore, these modifications or improvements made without departing from the spirit of the present application all fall within the scope of protection claimed in the present application.

Claims

1. A method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, characterized in that: The following steps are involved: Acquiring the clinical data of the patient to form a data set, and performing data preprocessing on the data set, wherein acquiring the clinical data of the patient specifically includes: Obtain the data on the influencing factors of maxillary sinus cysts, including age, gender, dentition, pulp, periodontal condition, different sides of the maxillary sinus, the relationship between the tooth root and the maxillary sinus floor, the morphology of the maxillary sinus floor, the blood vessels of the maxillary sinus side wall, the condition of the maxillary sinus mucosa, the presence of maxillary sinusitis, the development of the maxillary sinus, and the cortical bone condition of the maxillary sinus floor. The data preprocessing of the data set specifically includes: Remove outliers and missing values ​​from the data set, and fill in the missing data through interpolation. Normalizing or standardizing the numerical variables in the data set; The chi-square test and LASSO regression method are used to screen the characteristic variables in the data set, and based on the screened characteristic variables, a training set, a validation set and a test set are constructed. The characteristic variables in the screening data set specifically include: The categorical variables in the data set were expressed as numbers and percentages, and the chi-square test was used to screen out categorical variables with a bilateral P value less than 0.

05. LASSO regression was used to further screen the feature variables, including: compressing the variable coefficients in the regression model by generating a penalty function to prevent overfitting, and changing the coefficients of the related independent variables to 0 through the correlation between the independent variables. Among them, LASSO regression set 10-fold cross validation and was executed through the R package. The characteristic variables screened include: gender, age, pulp condition, periodontal condition, the relationship between tooth root and maxillary sinus floor, maxillary sinus floor morphology, maxillary sinus lateral wall blood vessels, maxillary sinus mucosa condition, the presence of maxillary sinusitis, maxillary sinus development, and maxillary sinus floor cortical bone condition; Build an initial prediction model based on the extreme gradient boosting algorithm; The initial prediction model is trained based on the training set, the validation set and the test set, and a tree is added iteratively, each time a tree is added to fit the residual of the previous round of prediction, so as to finally obtain a maxillary sinus cyst prediction model whose root mean square error meets the preset requirements; The maxillary sinus cyst prediction model is expressed as the following formula: [ \hat{y} = \sum_{k=1}^{K} f_k(\mathbf{x}) ], Where ( \hat{y} ) is the predicted value, ( K ) is the number of trees, and ( f_k(\mathbf{x}) ) is the output of the ( k )th tree for the input feature ( \mathbf{x} ); The parameters of the maxillary sinus cyst prediction model include: a learning rate of 0.1, a maximum tree depth of 6, a minimum bifurcation weight sum of 1, and an L2 regularization coefficient of 0.

1.

2. A method for predicting maxillary sinus cysts, characterized in that: Includes steps: Obtain relevant feature values ​​of the patient and input them into the maxillary sinus cyst prediction model constructed by the method for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm according to claim 1, wherein the feature values ​​include at least one of gender, age, pulp condition, periodontal condition, relationship between tooth root and maxillary sinus floor, maxillary sinus floor morphology, maxillary sinus lateral wall blood vessels, maxillary sinus mucosal condition, presence of maxillary sinusitis, maxillary sinus development, and maxillary sinus floor cortical bone condition; Outputting a prediction result through the maxillary sinus cyst prediction model to give the probability of the patient developing a maxillary sinus cyst; The SHAP value is used to explain the maxillary sinus cyst prediction model, showing the important features and their influence on the maxillary sinus cyst. For each feature value in the maxillary sinus cyst prediction model, its SHAP value is calculated to reflect the average contribution of different feature values ​​to the prediction results.

3. A device for constructing a maxillary sinus cyst prediction model based on a machine learning algorithm, characterized in that: include: Memory; as well as, A processor connected to the memory, the processor being configured to execute the steps of the method of claim 1.

4. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a machine, the steps of the method according to claim 1 are implemented.

Images