Preeclampsia risk prediction method and system based on maternal factors and early pregnancy biomarkers

Through machine learning methods based on maternal factors and early pregnancy biomarkers, a preeclampsia risk prediction model is constructed, which solves the problems of poor prediction results and high cost in the existing technology, and achieves efficient and low-cost preeclampsia risk Forecasting improves prediction accuracy and model interpretability.

CN117747110BActive Publication Date: 2024-11-15NANJING DRUM TOWER HOSPITAL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311807416.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-26
Publication Date
2024-11-15
Estimated Expiration
2043-12-26

AI Technical Summary

Technical Problem

In the early prediction of preeclampsia in the prior art, traditional methods can only identify 30%-40% of high-risk pregnant women, the prediction effect is poor, and the biomarker detection cost is high and the stability is poor, so there is inconvenience in model promotion.

Method used

Using machine learning methods based on maternal factors and early pregnancy biomarkers, a preeclampsia risk prediction model is constructed through voting classifiers and feature selection algorithms (such as random forests and extreme random trees), combining 50% cross-validation and self-service method to calculate confidence range, improve forecasting efficiency and reduce costs.

Benefits of technology

It significantly improves the prediction accuracy of preeclampsia, has a simple model structure, can analyze the contribution of features, is low cost, is suitable for widespread applications, and the area under the predicted ROC curve reaches 0.885, and the preeclampsia reaches 0.972 in early onset preeclampsia.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117747110B_ABST
    Figure CN117747110B_ABST
Patent Text Reader

Abstract

The present invention provides a method and system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, wherein the method includes obtaining maternal factor data and early pregnancy biomarker data of patients; taking maternal factors and early pregnancy biomarkers of patients as input, and whether the patient's outcome is preeclampsia as output, and using a voting classifier of a machine learning method to train the input data; repeating the target number of times using a five-fold cross-validation to determine the process of hyperparameters of the machine learning model, and using a self-help method to calculate the confidence interval of the target percentage to obtain a stable result evaluation; evaluating and interpreting the machine learning model to predict the risk of preeclampsia. The model structure of the present invention is simple, and the model can be analyzed to obtain the contribution of each feature in the model, which helps to further understand the pathogenic principle of preeclampsia; at the same time, only maternal factors and a small number of test results are used to predict preeclampsia, and the implementation cost is low, which is conducive to wide use.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of disease risk prediction, and in particular relates to a method and system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers. Background Art

[0002] Preeclampsia is a common multi-system disease during pregnancy, with an incidence of about 3-8%. The main characteristics of preeclampsia are hypertension and proteinuria in pregnant women. Severe patients are accompanied by systemic multi-organ damage or functional failure. In severe cases, pregnant women may experience convulsions, coma, or even death, accounting for about 10-15% of the direct or indirect causes of maternal death. Although the cause of preeclampsia is not clear and there is no effective treatment except termination of pregnancy, if it can be detected as early as possible and timely intervention is carried out, the harm of related complications to pregnant women can be effectively reduced, thereby effectively preventing maternal deaths caused by preeclampsia.

[0003] Currently, the construction of a prediction model for preeclampsia has always been a research hotspot for the prevention of preeclampsia. The research on prediction models is mainly divided into two directions: on the one hand, it is about the development of new predictive factors or markers. Baseline factors of pregnant women, such as advanced age of pregnant women, primiparas who have never given birth, previous history of preeclampsia, pregnancy interval, use of assisted reproductive technology, family history of preeclampsia, obesity, hyperglycemia during pregnancy, previous chronic hypertension, kidney disease, and autoimmune diseases such as systemic lupus erythematosus and antiphospholipid syndrome, are all related to the occurrence of preeclampsia. However, the traditional method of screening high-risk groups for preeclampsia in early pregnancy is to evaluate through maternal characteristics and medical history, which can only identify 30%-40% of pregnant women with preeclampsia, and the prediction effect is poor. Therefore, it has become the main trend to construct a prediction model for PE by combining maternal baseline with biophysical or biochemical markers such as mean arterial pressure (MAP), uterine artery pulsatility index (UtA-PI), serum placental growth factor (PLGF) and pregnancy-associated plasma protein A (PAPP-A). When selecting predictive factors, we should not only consider the improvement of predictive efficiency or model properties, but also the cost-effectiveness and practical application value of the predictive factors when they are promoted on a large scale. For example, UtA-PI requires specific training for ultrasound physicians and is greatly affected by the level of ultrasound physicians. PLGF testing is not only expensive, but also sensitive to external temperature and soluble endothelial growth factor receptor 1 in serum, and the test results are less stable. On the other hand, the algorithm for constructing a prediction model for preeclampsia is also an important direction. The prediction models for early preeclampsia are mostly based on traditional Logistic regression modeling. The British Fetal Foundation proposed a competing risk model based on Bayesian theory, which takes maternal factors as prior risks and combines other biophysical or biochemical markers such as mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and pregnancy-associated plasma protein A at 11-13 weeks of pregnancy to predict the risk of preeclampsia, which greatly improves the predictive efficiency of preeclampsia. However, in the Bayesian competing risk model, the value of the biomarker needs to be expressed as a multiple of the median (MoM) and adjusted according to various maternal factors during pregnancy. When using this model, people from different sources need to make corrections, which makes it inconvenient to use and promote. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers.

[0005] In a first aspect, the present invention provides a method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, comprising:

[0006] Obtain the patient's maternal factor data and early pregnancy biomarker data; the maternal factor data include gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome and outcome; the early pregnancy biomarker data include mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and plasma protein A;

[0007] The voting classifier of the machine learning method is trained on the input data, with maternal factors and early pregnancy biomarkers as input and whether the patient's outcome is preeclampsia as output;

[0008] Repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for the target number of times, and use the bootstrap method to calculate the confidence interval of the target percentage to obtain a stable result evaluation;

[0009] Evaluate and interpret machine learning models for predicting risk of preeclampsia.

[0010] Furthermore, the method uses the patient's maternal factors and early pregnancy biomarkers as inputs and whether the patient's outcome is preeclampsia as output, and uses a voting classifier of a machine learning method to train the input data, including:

[0011] In the random forest classifier model, the information gain for each input variable is calculated according to the following formula:

[0012]

[0013] Where D is the data set of the current node on the decision tree; X i The input variables included maternal factors and early pregnancy biomarkers; IG (D, X i ) is the input variable X i The information gain of ; Entropy(D) is the entropy of D; p(y) is the probability that the output variable y in the data set D is preeclampsia; Y is the set of output variables; D v For X i The upper value is a subset of v; Entropy(D v ) is D v Entropy of Values(X i ) is the input variable X i A collection of values ​​of ;

[0014] Select the input feature X with the maximum information gain best Split the data set D to obtain multiple sub-data sets;

[0015] Each sub-dataset is segmented until a preset segmentation threshold is reached;

[0016] The predicted category Y of each leaf node is calculated according to the following formula leaf :

[0017] Y leaf =mode{y|(x,y)∈D leaf};

[0018] Where x is a sample in the data set, which is a vector containing all eigenvalues; D leaf is the leaf node set;

[0019] In the extreme random tree classifier model, a split point is randomly selected for each feature, and the best split point is used as the node splitting rule;

[0020] The information gain of the split point is calculated according to the following formula:

[0021]

[0022] Among them, IG(D p ,f) is the information gain of the selected split point; I is the impurity measure; D p is the data set of the parent node; N left N is the number of samples of the left child node after segmentation; right N is the number of samples of the right child node after segmentation; p is the number of samples of the parent node; D left is the data set of the left child node after segmentation; D right It is the data set of the right child node after segmentation.

[0023] Furthermore, the said evaluating and interpreting the effect of the machine learning model to predict the risk of preeclampsia includes:

[0024] Use the ROC curve and the area under the ROC curve and the coordinate axis to evaluate the performance of the machine learning model;

[0025] Sensitivity is used to measure the ability of machine learning models to identify positive examples;

[0026] Use specificity to measure the ability of machine learning models to identify negative examples;

[0027] The accuracy rate is used to measure the proportion of samples correctly classified by the machine learning model to the total samples;

[0028] Use the true positive rate when the false positive rate is 10% and the true positive rate when the false positive rate is 20% to evaluate the true positive rate of the machine learning model at the target false positive rate level;

[0029] The prediction probability accuracy of the machine learning model is measured using the calibration curve, calibration curve intercept and calibration curve slope;

[0030] The Brier score is used to measure the accuracy of the probabilistic predictions of machine learning models.

[0031] Furthermore, the method of evaluating and interpreting the effect of the machine learning model to predict the risk of preeclampsia also includes:

[0032] SHapleyAdditive exPlanations was used to interpret the model, and the SHAP value was calculated according to the following formula:

[0033]

[0034] Among them, Φ i is the SHAP value of the i-th feature; S is the feature set that does not contain feature i; N is the set of all features; |S| is the number of features in set S; |N| is the number of features in set N; f x (S) is the model prediction value when considering the feature set S, f x (S∪{i}) is the model prediction value when considering the feature set S and adding feature i.

[0035] In a second aspect, the present invention provides a system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, comprising:

[0036] An acquisition module is used to obtain maternal factor data and early pregnancy biomarker data of the patient; the maternal factor data include gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome and outcome; the early pregnancy biomarker data include mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and plasma protein A;

[0037] A training module is used to train the voting classifier using a machine learning method with maternal factors and early pregnancy biomarkers as input and whether the patient's outcome is preeclampsia as output;

[0038] The result evaluation module is used to repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for the target number of times, and to calculate the confidence interval of the target percentage using the bootstrap method to obtain a stable result evaluation;

[0039] A prediction module for performance evaluation and interpretation of machine learning models to predict the risk of preeclampsia.

[0040] In a third aspect, the present invention provides a computer device comprising a processor and a memory; wherein, when the processor executes a computer program stored in the memory, the steps of the method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers described in the first aspect are implemented.

[0041] In a fourth aspect, the present invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, the steps of the method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers described in the first aspect are implemented.

[0042] The present invention provides a method and system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, wherein the method includes obtaining maternal factor data and early pregnancy biomarker data of patients; taking maternal factors and early pregnancy biomarkers of patients as input, and whether the patient's outcome is preeclampsia as output, and using a voting classifier of a machine learning method to train the input data; repeating the target number of times using a five-fold cross-validation to determine the process of hyperparameters of the machine learning model, and using a self-help method to calculate the confidence interval of the target percentage to obtain a stable result evaluation; evaluating and interpreting the machine learning model to predict the risk of preeclampsia. The model structure of the present invention is simple, and the model can be parsed to obtain the contribution of each feature in the model, which helps to further understand the pathogenic principle of preeclampsia; at the same time, only maternal factors and a small number of test results are used to predict preeclampsia, and the implementation cost is low, which is conducive to wide use. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 A flowchart of a method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers provided in an embodiment of the present invention;

[0045] Figure 2 A flow chart of the queue entry provided by an embodiment of the present invention;

[0046] Figure 3 The ROC curve diagram of the overall preeclampsia validation set provided by the embodiment of the present invention;

[0047] Figure 4 The ROC curve diagram of the early-onset preeclampsia validation set provided by the embodiment of the present invention;

[0048] Figure 5A schematic diagram of the structure of a preeclampsia risk prediction system based on maternal factors and early pregnancy biomarkers provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In one embodiment, if Figure 1 As shown, an embodiment of the present invention provides a method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, comprising:

[0051] Step 101, obtaining patient's maternal factor data and early pregnancy biomarker data.

[0052] In this step, the information of 5144 pregnant women in the prospective cohort during the study period was obtained, and the python toolkit pandas was used to read the excel data file. The relevant data of the group were screened, and the samples with incomplete data were eliminated. Based on the data characteristics, the maternal factors and the characteristic values ​​of early pregnancy biomarkers were replaced with binary discrete numbers to represent the strings in the original data. Figure 2 As shown, 62 cases with incomplete baseline information or outcomes were excluded: 11 cases without baseline information, 5 cases without previous pregnancy records, 12 cases without mean arterial pressure, and 34 cases without uterine artery pulsatility index, with a total of 5082 cases after screening. On this basis, 163 cases of miscarriage or loss of follow-up before 28 weeks were excluded: 56 cases of spontaneous abortion before 28 weeks, 6 cases of termination of pregnancy due to personal factors, 18 cases of termination of pregnancy due to fetal malformation or chromosomal abnormality, and 83 cases of loss of follow-up or refusal to follow-up, leaving 4919 cases after screening. Finally, 275 cases with missing information on pregnancy-associated plasma protein A were excluded, leaving a total of 4644 cases. Among them, 49 cases were early-onset preeclampsia, 161 cases were late-onset preeclampsia, and 4434 cases were normal pregnancies.

[0053] Cohort information was collected, including maternal factors and early pregnancy biomarker test results. Maternal factor data included gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome, and outcome; the early pregnancy biomarker data included mean arterial pressure, uterine artery pulsatility index, serum placental growth factor, and plasma protein A.

[0054] Step 102, using the patient's maternal factors and early pregnancy biomarkers as input and whether the patient's outcome is preeclampsia as output, a voting classifier using a machine learning method is trained on the input data.

[0055] In this step, 500 decision trees are first constructed, and each node needs to be constructed by calculating the information gain for each tree. Exemplarily, in the random forest classifier model, the information gain of each input variable is calculated according to the following formula:

[0056]

[0057] Where D is the data set of the current node on the decision tree; X i The input variables included maternal factors and early pregnancy biomarkers; IG (D, X i ) is the input variable X i The information gain of ; Entropy(D) is the entropy of D; p(y) is the probability that the output variable y in the data set D is preeclampsia; Y is the set of output variables; D v For X i The upper value is a subset of v; Entropy(D v ) is D v Entropy of Values(X i ) is the input variable X i A collection of values.

[0058] Select the input feature X with the maximum information gain best Split the dataset D into multiple sub-datasets.

[0059]

[0060] Each sub-dataset is segmented until the preset segmentation times threshold is reached.

[0061] The predicted category Y of each leaf node is calculated according to the following formula leaf :

[0062] Y leaf =mode{y|(x,y)∈D leaf}.

[0063] Where x is a sample in the data set, which is a vector containing all eigenvalues; D leaf A set of leaf nodes.

[0064] In this process, each feature X i All factors will be considered, including gestational age, height, gestational weight, reproductive history, etc., and their contribution to the prediction of preeclampsia. At this point, a single decision tree is constructed.

[0065] For a single decision tree, its data set is to select samples from the original training data using Bootstrap sampling (i.e., random sampling with replacement) to construct 500 decision trees.

[0066] The final prediction of the random forest is based on the predictions of all decision trees using a majority voting method.

[0067] Similar to random forest, but when constructing decision tree nodes, instead of searching for the optimal split point like random forest, extreme random trees randomly select a split point for each feature and choose the best split point as the node split rule. For each tree, features and split points are randomly selected from all features.

[0068] In the extreme random tree classifier model, a split point is randomly selected for each feature, and the best split point is used as the node splitting rule;

[0069] The information gain of the split point is calculated according to the following formula:

[0070]

[0071] Among them, IG(D p ,f) is the information gain of the selected split point; I is the impurity measure; D p is the data set of the parent node; N left N is the number of samples of the left child node after segmentation; right N is the number of samples of the right child node after segmentation; p is the number of samples of the parent node; D left is the data set of the left child node after segmentation; D right It is the data set of the right child node after segmentation.

[0072] Finally, 500 decision trees were constructed, and the predicted classification of each sample was obtained by majority voting of the trees.

[0073] Taking these two models as base learners, two ensemble learning algorithms, Stacking and soft Voting, are used respectively to combine them into a new strong learner.

[0074] Step 103, repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for a target number of times (200 times), and use the bootstrap method to calculate the confidence interval of the target percentage (95%) to obtain a stable result evaluation.

[0075] Step 104, evaluating and interpreting the machine learning model to predict the risk of preeclampsia.

[0076] In this step, the ROC curve and the area under the ROC curve and the coordinate axis are used to evaluate the performance of the machine learning model.

[0077] Sensitivity is used to measure the ability of a machine learning model to identify positive examples.

[0078] Specificity is used to measure the ability of a machine learning model to identify negative examples.

[0079] The accuracy rate is used to measure the proportion of samples correctly classified by the machine learning model to the total samples.

[0080] The true positive rate when the false positive rate is 10% and the true positive rate when the false positive rate is 20% are used to evaluate the true positive rate of the machine learning model at the target false positive rate level.

[0081] The calibration curve, calibration curve intercept and calibration curve slope are used to measure the prediction probability accuracy of the machine learning model.

[0082] The Brier score is used to measure the accuracy of the probabilistic predictions of machine learning models.

[0083] SHapleyAdditive exPlanations was used to interpret the model, and the SHAP value was calculated according to the following formula:

[0084]

[0085] Among them, Φ i is the SHAP value of the i-th feature; S is the feature set that does not contain feature i; N is the set of all features; |S| is the number of features in set S; |N| is the number of features in set N; f x (S) is the model prediction value when considering the feature set S, f x (S∪{i}) is the model prediction value when considering the feature set S and adding feature i.

[0086] After comparing the above steps and results, the final model is determined to be a strong learner with random forest and extreme random tree as base learners and soft voting as ensemble learning method. The ROC curve results are as follows: Figure 3 and Figure 4 It can be seen that the area enclosed by the ROC curve and the coordinate axis for the prediction of preeclampsia risk of the present invention for overall preeclampsia is 0.885, while the area enclosed by the ROC curve and the coordinate axis for the prediction of early-onset preeclampsia is 0.972, which shows that the method for predicting preeclampsia provided by the present invention has a better effect.

[0087] The embodiment of the present invention provides a method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, wherein the model structure is simple, the model can be analyzed to obtain the contribution of each feature in the model, and help further understand the pathogenic mechanism of preeclampsia; at the same time, only maternal factors and a small amount of test results are used to predict preeclampsia, the implementation cost is low, and it is conducive to wide use.

[0088] Based on the same inventive concept, an embodiment of the present invention also provides a preeclampsia risk prediction system based on maternal factors and early pregnancy biomarkers. Since the principle of solving the problem by this system is similar to that of the preeclampsia risk prediction method based on maternal factors and early pregnancy biomarkers, the implementation of this system can refer to the implementation of the preeclampsia risk prediction method based on maternal factors and early pregnancy biomarkers, and the repeated parts will not be repeated.

[0089] In another embodiment, the present invention provides a system for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, such as Figure 5 As shown, including:

[0090] Acquisition module 10 is used to obtain the patient's maternal factor data and early pregnancy biomarker data; the maternal factor data include gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome and outcome; the early pregnancy biomarker data include mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and plasma protein A.

[0091] The training module 20 is used to train the input data using a voting classifier of a machine learning method, taking the patient's maternal factors and early pregnancy biomarkers as input and whether the patient's outcome is preeclampsia as output.

[0092] The result evaluation module 30 is used to repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for a target number of times, and to calculate the confidence interval of the target percentage using a bootstrap method to obtain a stable result evaluation.

[0093] The prediction module 40 is used to evaluate and interpret the effect of the machine learning model to predict the risk of preeclampsia.

[0094] For more specific working processes of the above modules, please refer to the corresponding contents disclosed in the above embodiments, which will not be repeated here.

[0095] In another embodiment, the present invention provides a computer device comprising a processor and a memory; wherein, when the processor executes a computer program stored in the memory, the steps of the above-mentioned method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers are implemented.

[0096] For more specific processes of the above method, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0097] In another embodiment, the present invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, the steps of the above-mentioned method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers are implemented.

[0098] For more specific processes of the above method, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system, device, and storage medium disclosed in the embodiment, since they correspond to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention or some parts of the embodiments.

[0101] The present invention has been described in detail above in conjunction with specific implementations and exemplary examples, but these descriptions cannot be understood as limiting the present invention. Those skilled in the art understand that, without departing from the spirit and scope of the present invention, a variety of equivalent substitutions, modifications or improvements may be made to the technical solution of the present invention and its implementation methods, all of which fall within the scope of the present invention. The scope of protection of the present invention shall be subject to the attached claims.

Claims

1. A method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers, characterized in that: include: Obtain the patient's maternal factor data and early pregnancy biomarker data; the maternal factor data include gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome and outcome; the early pregnancy biomarker data include mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and plasma protein A; The input data are maternal factors and early pregnancy biomarkers, and the output is whether the patient's outcome is preeclampsia. The voting classifier of the machine learning method is trained on the input data, including: In the random forest classifier model, the information gain for each input variable is calculated according to the following formula: Where D is the data set of the current node on the decision tree; X i The input variables included maternal factors and early pregnancy biomarkers; IG (D, X i ) is the input variable X i The information gain of ; Entropy(D) is the entropy of D; p(y) is the probability that the output variable y in the data set D is preeclampsia; Y is the set of output variables; D v For X i The upper value is a subset of v; Entropy(D v ) is D v Entropy of Values(X i ) is the input variable X i A collection of values ​​of ; Select the input feature X with the maximum information gain best Split the data set D to obtain multiple sub-data sets; Each sub-dataset is segmented until a preset segmentation threshold is reached; The predicted category Y of each leaf node is calculated according to the following formula leaf : AND leaf =mode{y|(x,y)∈D leaf }; Where x is a sample in the data set, which is a vector containing all eigenvalues; D leaf is the leaf node set; In the extreme random tree classifier model, a split point is randomly selected for each feature, and the best split point is used as the node splitting rule; The information gain of the split point is calculated according to the following formula: Among them, IG(D p ,f) is the information gain of the selected split point; I is the impurity measure; D p is the data set of the parent node; N left N is the number of samples of the left child node after segmentation; right N is the number of samples of the right child node after segmentation; p is the number of samples of the parent node; D left is the data set of the left child node after segmentation; D right It is the data set of the right child node after segmentation; Repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for the target number of times, and use the bootstrap method to calculate the confidence interval of the target percentage to obtain a stable result evaluation; Evaluate and interpret machine learning models to predict risk of preeclampsia, including: Use the ROC curve and the area under the ROC curve and the coordinate axis to evaluate the performance of the machine learning model; Sensitivity is used to measure the ability of machine learning models to identify positive examples; Use specificity to measure the ability of machine learning models to identify negative examples; The accuracy rate is used to measure the proportion of samples correctly classified by the machine learning model to the total samples; Use the true positive rate when the false positive rate is 10% and the true positive rate when the false positive rate is 20% to evaluate the true positive rate of the machine learning model at the target false positive rate level; The prediction probability accuracy of the machine learning model is measured using the calibration curve, calibration curve intercept and calibration curve slope; Use the Brier score to measure the accuracy of the probabilistic predictions of machine learning models; SHapley Additive exPlanations was used to interpret the model, and the SHAP value was calculated according to the following formula: Among them, Φ i is the SHAP value of the i-th feature; S is the feature set that does not contain feature i; N is the set of all features; |S| is the number of features in set S; |N| is the number of features in set N; f x (S) is the model prediction value when considering the feature set S, f x (S∪{i}) is the model prediction value when considering the feature set S and adding feature i.

2. A preeclampsia risk prediction system based on maternal factors and early pregnancy biomarkers, characterized in that: include: An acquisition module is used to obtain maternal factor data and early pregnancy biomarker data of the patient; the maternal factor data include gestational age, height, gestational weight, reproductive history, history of preeclampsia, history of diabetes, history of hypertension, assisted reproduction, family history of preeclampsia, systemic lupus erythematosus or antiphospholipid syndrome and outcome; the early pregnancy biomarker data include mean arterial pressure, uterine artery pulsatility index, serum placental growth factor and plasma protein A; A training module is used to train the voting classifier using a machine learning method with maternal factors and early pregnancy biomarkers as input and whether the patient's outcome is preeclampsia as output; The result evaluation module is used to repeat the process of determining the hyperparameters of the machine learning model using five-fold cross validation for the target number of times, and to calculate the confidence interval of the target percentage using the bootstrap method to obtain a stable result evaluation; A prediction module for evaluating and interpreting machine learning models to predict the risk of preeclampsia; The preeclampsia risk prediction system is used to implement the preeclampsia risk prediction method based on maternal factors and early pregnancy biomarkers as described in claim 1.

3. A computer device, characterized in that: It comprises a processor and a memory; wherein, when the processor executes the computer program stored in the memory, it implements the steps of the method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers as described in claim 1.

4. A computer-readable storage medium, characterized in that: Used to store a computer program; when the computer program is executed by a processor, the steps of the method for predicting the risk of preeclampsia based on maternal factors and early pregnancy biomarkers as described in claim 1 are implemented.