A risk prediction model for endometrial adenocarcinoma, a method for constructing the model, and a construction device thereof.

Abstract

This invention belongs to the field of disease risk prediction, specifically relating to a risk prediction model for endometrial adenocarcinoma, a method for constructing the model, and a construction apparatus. This invention utilizes high-throughput metabolomics data of free fatty acids in blood samples detected by liquid chromatography-mass spectrometry (LC-MS / MS) and combines this with artificial intelligence (AI) technology to construct a prediction model. This invention, relying on quantitative metabolomics data combined with AI technology, provides technical support for the construction of an endometrial adenocarcinoma risk prediction model, accurately presenting the risk prediction model and developing risk prediction software, with broad clinical application prospects.

Description

Technical Field

[0001] This invention belongs to the field of disease risk prediction, and specifically relates to a risk prediction model for endometrial adenocarcinoma, a method for constructing the model, and a construction device. Background Technology

[0002] Endometrial cancer is a common gynecological malignancy. In 2018, there were 89,000 deaths worldwide from endometrial cancer, and this number is increasing annually. Sheikh et al. projected an incidence rate of 42 per 100,000 by 2030, a 55% increase compared to 2010. The NCCN guidelines use CA125 as a commonly used tumor marker for endometrial cancer, but its specificity and sensitivity are unsatisfactory. Clinically, early diagnosis is commonly achieved through hysteroscopy or curettage for histopathological examination, both of which are invasive procedures with a sensitivity of only 86.4%, hindering the widespread adoption of endometrial cancer screening. Therefore, finding non-invasive biomarkers for endometrial cancer that facilitate screening and disease assessment is a pressing clinical challenge.

[0003] The most common pathological type of endometrial cancer is estrogen-dependent endometrial adenocarcinoma, accounting for approximately 80%-90%, and is closely related to metabolic syndrome. Katherine reported that metabolic syndrome increases the risk of endometrial cancer by 61%, and abnormal lipid metabolism is the most important risk factor for the development and progression of endometrial cancer. Free fatty acids (FFA) are intermediate products of lipid mobilization, mainly derived from the breakdown of lipids and triglycerides (TGs). The application of liquid chromatography-mass spectrometry for the spectroscopic quantitative detection of free fatty acids to construct screening models has not been reported in the field of endometrial cancer and has significant potential for widespread application.

[0004] Furthermore, traditional statistical methods for model building have high requirements for data distribution, while artificial intelligence technology has broader requirements for data. This study uses mass spectrometry detection of free fatty acid spectra as features to select and optimize five algorithms: XGBoost, support vector machine, decision tree, random forest, and Lasso regression. The optimal model and parameters suitable for this dataset are selected to construct a metabolomics prediction model for endometrial adenocarcinoma. The value of the model in the diagnosis and early detection of endometrial adenocarcinoma is evaluated, and software development is carried out. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides a risk prediction method / software that predicts the probability of developing endometrial adenocarcinoma based on the characteristics of the free fatty acid profile in the blood of the test subject, thereby determining the risk level of the subject and enabling early detection of the disease, thus helping clinicians to better formulate treatment plans.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] This invention provides a method for establishing a machine learning-based model for predicting the risk of endometrial adenocarcinoma, characterized by the following steps:

[0008] (1) Quantitative detection of free fatty acid spectra in blood samples of experimental and control groups was performed using liquid chromatography-mass spectrometry;

[0009] (2) Statistical analysis was performed using univariate or multivariate logistic regression analysis to identify risk factors associated with endometrial adenocarcinoma events and characteristics.

[0010] (3) The risk factors selected in (2) were used as features for the construction of artificial intelligence models. Decision trees, random forests, support vector machines, Lasso regressions and XGBoost models were constructed and their parameters were adjusted in the data environment of endometrial adenocarcinoma. The model suitable for endometrial adenocarcinoma metabolome data was selected.

[0011] (4) Select an external validation dataset to verify the accuracy of the XGBoost model.

[0012] Furthermore, the statistical analysis of univariate or multivariate analysis and endometrial adenocarcinoma-related events and characteristics described in step (2) includes:

[0013] (1) Univariate analysis of risk factors associated with the development of endometrial adenocarcinoma;

[0014] (2) Univariate analysis of risk factors associated with postoperative pathological risk of endometrial adenocarcinoma;

[0015] (3) Multivariate analysis of risk factors associated with the development and progression of endometrial adenocarcinoma.

[0016] Furthermore, the risk factors identified in step (2) include 23 free fatty acid metabolites identified by a test.

[0017] Preferably, the 23 free fatty acid metabolites are: decanoic acid, lauric acid, myristic acid, palmitic acid, stearic acid, arachidic acid, behenic acid, myristoleic acid, palmitoleic acid, oleic acid, eicosapentaenoic acid, erucic acid, nervonic acid, lignotaric acid, docosahexaenoic acid, eicosadienoic acid, eicosatrienoic acid, arachidonic acid, linoleic acid, linolenic acid, docosapentaenoic acid, docosatraenoic acid, and eicosapentaenoic acid.

[0018] Furthermore, the core idea of ​​the XGBoost model described in step (3) is:

[0019] (1) Continuously add trees and continuously perform feature splits to grow a tree. Each time a tree is added, it is actually learning a new function f(xi) to fit the residual of the previous prediction.

[0020] (2) When we have completed training and obtained k trees, we need to predict the score of a sample. In fact, based on the characteristics of the sample, it will fall into a corresponding leaf node in each tree, and each leaf node corresponds to a score.

[0021] (3) Finally, simply sum the scores corresponding to each tree to get the predicted value for that sample, and let the predicted value... Try to get as close as possible to the true value y i .

[0022] Preferably, the model construction is as follows:

[0023] S1: Iterative formula for XGBoost prediction results: in This is the result of the (k+1)th prediction. This is the prediction result for the kth time, f k+1 (x i ) is the function to be trained in the (k+1)th iteration, where x i It is a feature vector set consisting of 23 features;

[0024] S2: Objective function: The first part is the loss function, and the second part is the regularization.

[0025] The loss function is Regularization is expressed as

[0026] This invention also provides a machine learning-based method for predicting the risk of endometrial adenocarcinoma, characterized in that the method includes the following steps:

[0027] S1 Data Acquisition: Quantitative detection of free fatty acid spectra in blood samples from the experimental and control groups was performed using liquid chromatography-mass spectrometry.

[0028] S2 data preprocessing: Statistical analysis was performed using univariate or multivariate analysis to identify risk factors associated with endometrial adenocarcinoma events and characteristics;

[0029] S3 Training Prediction Model: Using the obtained risk factors as features for constructing artificial intelligence models, decision trees, random forests, support vector machines, Lasso regressions, and XGBoost models were constructed and their parameters were adjusted in the data environment of endometrial adenocarcinoma to select models suitable for endometrial adenocarcinoma metabolome data.

[0030] S4 Risk Prediction: Select external validation datasets to validate the accuracy of the XGBoost model, including selecting FIGO Phase I validation set to validate the value of XGBoost in the early diagnosis of endometrial adenocarcinoma and selecting external validation datasets to validate the accuracy of the XGBoost model in early diagnosis.

[0031] The present invention also provides an apparatus for a machine learning-based endometrial adenocarcinoma risk prediction model, characterized in that the apparatus comprises:

[0032] Memory;

[0033] Communication interface;

[0034] The processor is configured to perform the following steps:

[0035] (1) Quantitative detection of free fatty acid spectra in blood samples of experimental and control groups was performed using liquid chromatography-mass spectrometry;

[0036] (2) Statistical analysis was performed using univariate or multivariate analysis to identify risk factors associated with endometrial adenocarcinoma events and characteristics;

[0037] (3) Using the obtained risk factors as features for artificial intelligence model construction, decision tree, random forest, support vector machine, Lasso regression and XGBoost models were constructed and their parameters were adjusted in the data environment of endometrial adenocarcinoma to select suitable models for endometrial adenocarcinoma metabolome data.

[0038] (4) Select external validation datasets to validate the accuracy of the XGBoost model, including selecting FIGO Phase I validation set to validate the value of XGBoost in early diagnosis of endometrial adenocarcinoma and selecting external validation datasets to validate the accuracy of the XGBoost model in early diagnosis.

[0039] Furthermore, the processor is further configured to perform any or all of the method steps described in this invention.

[0040] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed any or all of the method steps described in the present invention.

[0041] The beneficial effects of the present invention compared with the prior art.

[0042] Endometrial adenocarcinoma is a metabolic disease lacking sensitive and specific tumor markers. Therefore, whether tumor markers for endometrial cancer can be extracted from metabolomics data is worth exploring. This invention utilizes LC-MS / MS detection of free fatty acids in blood samples to obtain high-throughput metabolomics data and combines it with artificial intelligence technology to construct a predictive model, filling a gap in this field.

[0043] The application of artificial intelligence (AI) technology in the field of endometrial cancer is mostly limited to single algorithms. This invention proposes the XGBoost model, a sophisticated evaluator within ensemble algorithms, which demonstrates superior computational power across various domains due to its step-by-step optimization capabilities; however, its application in endometrial cancer has not been reported. Combining quantitative metabolomics data with AI technology can provide technical support for constructing a risk prediction model for endometrial adenocarcinoma, enabling accurate risk prediction and the development of risk prediction software. Attached Figure Description

[0044] Figure 1 Expression levels of candidate differentially expressed metabolites and their importance in model construction. *P < 0.05, **P < 0.001; the numerical order after the metabolite name indicates the importance of the feature during modeling.

[0045] Figure 2 Accuracy of cross-validation of the five models.

[0046] Figure 3 ROC curves for the five models on the training and test sets. Where A represents the ROC curve for the five models on the training set; B represents the ROC curve for the five models on the test set.

[0047] Figure 4 ROC curves for the validation set in diagnosing endometrial adenocarcinoma. Validation 1: Validation set 1 validates the ROC curve of the XGBoost model in diagnosing endometrial adenocarcinoma; Validation 2: Validation set 2 validates the ROC curve of the XGBoost model in diagnosing early-stage endometrial adenocarcinoma; Validation 3: Validation set 3 validates the ROC curve of the XGBoost model in diagnosing early-stage endometrial adenocarcinoma. Detailed Implementation

[0048] The present invention will be described below through specific embodiments, but the present invention is not limited thereto. Unless otherwise specified, the experimental methods used in the following embodiments are conventional methods; unless otherwise specified, the reagents, biological materials, etc. used in the following embodiments are commercially available.

[0049] Example 1: Prediction Model Establishment.

[0050] 1. Materials and Methods.

[0051] 1.1 Sources of data and specimens:

[0052] We collected 143 cases of pathologically confirmed endometrial adenocarcinoma that underwent gynecological surgery at Liaoning Cancer Hospital between February 2020 and July 2021 as the experimental group, and selected 124 gynecological patients who underwent surgery for non-estrogen-dependent diseases at our hospital during the same period as the control group.

[0053] 1.2 Data and Specimen Collection:

[0054] Collect clinical data for all study cases, including age, clinical symptoms, menstrual history (age of menarche, menstrual period, menstrual cycle, age of menopause), height, weight, BMI; complete surgical data; postoperative specimen pathological type, tumor differentiation, lymph node metastasis, tumor size, muscle layer invasion, and disease stage.

[0055] 1.3 Data and Blood Sample Testing:

[0056] Fasting venous blood (2-4 mL) was collected from enrolled patients after 12 hours of fasting. After blood sample pretreatment, the samples were transferred to injection vials for detection by liquid chromatography-mass spectrometry. The obtained data were preprocessed to obtain quantitative detection results for data analysis and modeling.

[0057] 1.4 Research Methods.

[0058] (1) Liquid Chromatography-Mass Spectrometry Detector:

[0059] The samples were analyzed by an AB SCIEX Triple Quad 4500MD liquid chromatography-mass spectrometry system. The generated raw data were preprocessed using Analyst MD Software for data denoising, normalization, and standardization to obtain quantitative detection data.

[0060] (2) Statistical analysis and feature extraction:

[0061] The quantitative results of detected substances were analyzed using SPSS 21.0 statistical software. Quantitative data that did not conform to a normal distribution were described using the median and interquartile range (IQR). The Mann-Whitney U test was used to compare non-normally distributed quantitative data between two groups. Log transformation was performed on non-normally distributed data for non-linear transformation. Univariate and multivariate logistic regression analyses were conducted on clinicopathological characteristics. A p-value < 0.05 was considered statistically significant.

[0062] (3) Model building:

[0063] Data modeling and analysis were performed using the scikit-learn framework in Python. Data preprocessing methods such as standardization were used; feature selection was performed using filtering and embedding methods. Based on the characteristics of the dataset, several supervised models were selected, including decision trees, random forests, support vector machines, Lasso regression, and XGBoost. The optimal model was selected by comprehensively comparing accuracy and AUC values, and parameters were adjusted using learning curves and cross-validation.

[0064] (4) Receiver operating characteristic (ROC) curves were plotted to evaluate the diagnostic efficacy of the model. The optimal model was selected by comprehensively comparing accuracy and AUC values. External data was selected as validation set 1 for model validation. Early-stage cases were selected to validate the diagnostic value of XGBoost in early-stage patients, with identical model parameters.

[0065] 2. Results.

[0066] 2.1 Detection of serum free fatty acids by LC-MS / MS and selection of candidate differential metabolites:

[0067] A total of 23 free fatty acid metabolites and 8 ratios were detected. The detection items and quantitative results are shown in Table 1. The 31 detection items were divided into four categories according to whether the carbon chain contained unsaturated bonds and the number of double bonds: 1. Saturated fatty acids, 2. Monounsaturated fatty acids, 3. Polyunsaturated fatty acids, 4. Ratios and sums.

[0068] Table 1 Comparison of serum metabolite detection items and expression levels between experimental group and control group.

[0069]

[0070]

[0071]

[0072]

[0073] 2.1.1 Correlation between serum free fatty acids and the occurrence of endometrial adenocarcinoma

[0074] The training test focused on comparing the experimental and control groups in the following ways: saturated fatty acids: myristic acid, palmitic acid, and arachidic acid; monounsaturated fatty acids: myristoleic acid, palmitoleic acid, and oleic acid; polyunsaturated fatty acids: eicosadienoic acid, mead acid, linoleic acid, linolenic acid, docosapentaenoic acid (DPA), and docosatraenoic acid (DTA); ratios and sums of triene / tetraene ratio, total saturated fatty acids, total monounsaturated fatty acids, total polyunsaturated fatty acids, total w3, total w6, and total fatty acids. A total of 19 substances (fatty acids) showed significant differences (P < 0.05), and the median values ​​of the above metabolites in the experimental group were all higher than those in the control group. These substances were included as candidate differential metabolites, as detailed in Table 1.

[0075] 2.1.2 Correlation between serum free fatty acids and the development of endometrial adenocarcinoma

[0076] Univariate logistic regression analysis was performed on the clinicopathological characteristics of the experimental group. Indicators with statistically significant correlations (P < 0.05) between each characteristic and the metabolite were included as candidate differential metabolites. Further, using metabolites as independent variables and clinicopathological characteristics as dependent variables, multivariate analysis was conducted on the statistically significant metabolites from the univariate analysis. Indicators with P < 0.05 were included as features in the artificial intelligence model. Combined with the 19 metabolites obtained in section 2.1.1, a total of 23 metabolites were selected as candidate differential metabolites for the next step of artificial intelligence model construction. (See...) Figure 1 .

[0077] 2.2 Construction of artificial intelligence models using candidate differential metabolites and selection of candidate models.

[0078] Feature selection was performed on the 23 candidate differential metabolites included. After processing by filtering and embedding methods, all 23 features were found to carry valid information. Since the selected features were obtained from a single check, they were all retained, thus completing feature extraction.

[0079] The 223 patients in the training and test sets were randomly assigned to the training and test sets in a 9:1 ratio. Decision trees, random forests, and support vector machines exhibited severe overfitting issues, so "pruning" was performed on the trees. Lasso regression, due to its L1 regularization, did not show overfitting, but its cross-validation results were low. XGBoost, with its step-by-step approach, achieved a cross-validation accuracy exceeding 75% (see...). Figure 2 Furthermore, when performing predictions using modeling, the sensitivity, accuracy, AUC, and cross-validation of the five models were all higher than the other four models, both on the training and test sets. See details for further information. Figure 3 And Table 2. Based on the above research, the XGBoost model was selected as the diagnostic model for endometrial adenocarcinoma.

[0080] Table 2 Diagnostic efficacy of each model on the training and test sets.

[0081]

[0082] 2.3 Validation Set 1: Validating the XGBoost Model

[0083] To validate the accuracy of the XGBoost model in diagnosing endometrial adenocarcinoma, a validation set of 44 patients (13 in the experimental group and 31 in the control group) was selected. The trained model was used for validation, yielding an accuracy of 75%, specificity of 77.5%, sensitivity of 70%, and AUC of 73.3%. This indicates that the XGBoost model maintains high accuracy even after external validation, demonstrating high diagnostic value. Figure 4 ).

[0084] 2.4 Value of XGBoost Model in Early Diagnosis of Endometrial Adenocarcinoma

[0085] The training set and the FIGO I stage patients and control group from the test set were selected as validation set 2. The model's accuracy in diagnosing early-stage patients was 80.6%, specificity was 67.7%, sensitivity was 92%, and AUC was 80.1%. Further validation set 3 was selected from the FIGO I stage patients and control group from the validation set, yielding an accuracy of 72.5%, specificity of 77.4%, sensitivity of 55.6%, and AUC of 66.5% (see [link to validation set]). Figure 4 ).

[0086] 2.5 Software Development

[0087] This software primarily uses a B / S architecture, which offers greater adaptability on the client side; it can be used simply by installing the corresponding browser.

[0088] For the client side, this invention primarily uses HTML, CSS, and JavaScript, combined with Ajax. The client uses a browser to access the application and is mainly used to interact with end users (such as doctors) and display the results of the interaction.

[0089] The server program is primarily written using the MVC architecture, connecting to the administrator interface and database, and controlling business logic programmatically. Based on the RBAC model, different permissions are assigned to different user roles.

[0090] For the servers, since the data used in this invention is internal hospital data, public cloud cannot be used. Therefore, a private cluster was built as the server, deploying no fewer than three server devices with at least 128GB of memory, and providing storage space for the server devices via IPSAN. Considering that the amount of business data will be very large in the future, HDFS of open-source Hadoop was used as the file system for data storage, Hive was used as the data warehouse, and Sqoop was used to connect with the business relational database.

[0091] For modeling, this invention uses the NumPy and Pandas frameworks to process the data extracted from the data warehouse, uses the Python open-source framework scikit-learn for modeling, and deploys the model program on a designated data mining server. Crontab is used to run the model on a regular schedule to cope with the ever-increasing business datasets.

[0092] The above description of the embodiments is only for understanding the method and core ideas of the present invention. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from the principles of the invention, and these improvements and modifications will also fall within the protection scope of the claims of the present invention.

Claims

1. A method for establishing a machine learning-based model for predicting the risk of endometrial adenocarcinoma, characterized in that, The method includes the following steps: S1 used liquid chromatography-mass spectrometry to quantitatively detect the free fatty acid spectra of blood samples from the experimental and control groups; S2 used univariate or multivariate logistic regression analysis to statistically analyze the events and characteristics associated with endometrial adenocarcinoma, and screened out risk factors. The risk factors included 23 free fatty acid metabolites obtained from one test. The 23 free fatty acid metabolites were: decanoic acid, lauric acid, myristic acid, palmitic acid, stearic acid, arachidic acid, behenic acid, myristoleic acid, palmitoleic acid, oleic acid, eicosapentaenoic acid, erucic acid, nervonic acid, lignotaric acid, docosahexaenoic acid, eicosadienoic acid, eicosatrienoic acid, arachidonic acid, linoleic acid, linolenic acid, docosapentaenoic acid, docosatraenoic acid, and eicosapentaenoic acid. S3 uses the risk factors screened in S2 as features for the construction of artificial intelligence models. Decision trees, random forests, support vector machines, Lasso regressions, and XGBoost models are constructed and their parameters are tuned in the data environment of endometrial adenocarcinoma, and models suitable for endometrial adenocarcinoma metabolome data are selected. S4 selects an external validation dataset to validate the accuracy of the XGBoost model.

2. The method for establishing a machine learning-based endometrial adenocarcinoma risk prediction model according to claim 1, characterized in that, The univariate or multivariate logistic regression analysis described in step S2, which involves statistical analysis of events and characteristics associated with endometrial adenocarcinoma, includes: (1) Univariate logistic regression analysis of risk factors associated with the risk of developing endometrial adenocarcinoma; (2) Univariate logistic regression analysis of risk factors associated with postoperative pathological risk of endometrial adenocarcinoma; (3) Multivariate logistic regression analysis of risk factors associated with the development and progression of endometrial adenocarcinoma.

3. The method for establishing a machine learning-based endometrial adenocarcinoma risk prediction model according to claim 1, characterized in that, The core idea of ​​the XGBoost model described in step S3 is: (1) Continuously add trees and continuously perform feature splits to grow a tree. Each time a tree is added, it is actually learning a new function f(x). i ), to fit the residual of the previous prediction; where x i It is a feature vector set consisting of 23 features; (2) When we have completed training and obtained k trees, we need to predict the score of a sample. In fact, based on the characteristics of the sample, it will fall into a corresponding leaf node in each tree, and each leaf node corresponds to a score. (3) Finally, we only need to add up the scores corresponding to each tree to get the predicted value of the sample, and make the predicted value as close as possible to the true value.

4. A machine learning-based method for predicting the risk of endometrial adenocarcinoma, characterized in that, The method includes the following steps: S1 Data Acquisition: Quantitative detection of free fatty acid spectra in blood samples from the experimental and control groups was performed using liquid chromatography-mass spectrometry. S2 Data Preprocessing: Statistical analysis was performed using univariate or multivariate logistic regression analysis to identify events and characteristics associated with endometrial adenocarcinoma, and risk factors were screened. These risk factors included 23 free fatty acid metabolites obtained from a single test. These 23 free fatty acid metabolites were: decanoic acid, lauric acid, myristic acid, palmitic acid, stearic acid, arachidic acid, behenic acid, myristoleic acid, palmitoleic acid, oleic acid, eicosapentaenoic acid, erucic acid, nervonic acid, lignotaric acid, docosahexaenoic acid, eicosadienoic acid, eicosatrienoic acid, arachidonic acid, linoleic acid, linolenic acid, docosapentaenoic acid, docosatraenoic acid, and eicosapentaenoic acid. S3 Training Prediction Model: Using the obtained risk factors as features for constructing artificial intelligence models, decision trees, random forests, support vector machines, Lasso regressions, and XGBoost models were constructed and their parameters were tuned in the data environment of endometrial adenocarcinoma, and models suitable for endometrial adenocarcinoma metabolomics data were selected. S4 Risk Prediction: Select external validation datasets to validate the accuracy of the XGBoost model, including selecting FIGO Phase I validation set to validate the value of the XGBoost model in the early diagnosis of endometrial adenocarcinoma and selecting external validation datasets to validate the accuracy of the XGBoost model in early diagnosis.

5. A device for predicting the risk of endometrial adenocarcinoma based on machine learning, characterized in that, The device package include: Memory; Communication interface; The processor is configured to perform the following steps: (1) Quantitative detection of free fatty acid spectra in blood samples of experimental and control groups was performed using liquid chromatography-mass spectrometry. (2) Statistical analysis was performed using univariate or multivariate analysis to identify events and characteristics associated with endometrial adenocarcinoma, and risk factors were screened out. The risk factors included 23 free fatty acid metabolites obtained from one test. The 23 free fatty acid metabolites were: decanoic acid, lauric acid, myristic acid, palmitic acid, stearic acid, arachidic acid, behenic acid, myristoleic acid, palmitoleic acid, oleic acid, eicosapentaenoic acid, erucic acid, nervonic acid, lignotaric acid, docosahexaenoic acid, eicosadienoic acid, eicosatrienoic acid, arachidonic acid, linoleic acid, linolenic acid, docosapentaenoic acid, docosatraenoic acid, and eicosapentaenoic acid. (3) Using the obtained risk factors as features for artificial intelligence model construction, decision tree, random forest, support vector machine, Lasso regression and XGBoost models were constructed and their parameters were adjusted in the data environment of endometrial adenocarcinoma, and suitable models for endometrial adenocarcinoma metabolome data were selected. (4) Select external validation datasets to validate the accuracy of the XGBoost model, including selecting FIGO Phase I validation set to validate the value of the XGBoost model in the early diagnosis of endometrial adenocarcinoma and selecting external validation datasets to validate the accuracy of the XGBoost model in early diagnosis.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is adapted to be loaded by a processor and executed by a method for establishing a machine learning-based endometrial adenocarcinoma risk prediction model as described in any one of claims 1-3.

Images