Obstetrical complication prediction system and method based on AI

Abstract

The invention discloses an obstetrical complication prediction system and method based on AI, and relates to the technical field of complication prediction. The obstetrical complication prediction system based on the AI comprises a data acquisition module, a data preprocessing module, a feature engineering module, a complication prediction module, a prevention measure module and a result display module. According to the method, the psychological state data of the pregnant woman is taken into consideration, that is, the psychological state data of the pregnant woman is also taken as partial data of model input and training, so that the obstetrical complication prediction model can more comprehensively understand the health state of a patient, thereby helping the model to make more accurate prediction, and improving the accuracy of the obstetrical complication prediction. Meanwhile, the model is optimized through an incremental learning method, so that the model can be gradually corrected by introducing psychological state data, overfitting of historical data is avoided, computing resources can be saved, computing cost and time expenditure can be reduced, a large amount of historical data does not need to be stored, only necessary parameters and statistical information need to be reserved, and the method is suitable for large-scale popularization and application. Therefore, the storage requirement is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of complication prediction, and in particular to an AI-based obstetric complication prediction system and method. Background Art

[0002] The obstetric complication prediction system based on artificial intelligence (AI) is a technology that uses machine learning and data analysis technology to analyze and model a large amount of clinical data to predict possible complications during pregnancy and delivery in advance. This system can help doctors and medical institutions take preventive measures to reduce the risk of complications and improve maternal and child safety.

[0003] The existing complication prediction system will first collect the patient's biological information, reference vital sign data and corresponding complication information at various stages, and then bring the data into the neural network model for training, and obtain the complication prediction model of the corresponding disease, and then input the latest physiological data of the patient to be predicted into the obstetric complication prediction model for analysis and prediction to obtain the prediction result. Although this method will analyze and predict based on the patient's various physiological data and indicators, this system does not take into account that pregnant women are prone to depression during the delivery period and affect various physiological data of the body, and thus cannot fully reflect the patient's physical condition. In addition, the existing prediction model obtains more and more complex physiological data, and a large number of features will be extracted during training, resulting in high computational costs and model overfitting.

[0004] Based on the above situation, the present invention proposes an AI-based obstetric complications prediction system and method with small computational complexity and good prediction effect. Summary of the invention

[0005] In order to overcome the shortcomings of existing complication prediction systems and methods, which fail to take into account the fact that pregnant women are prone to depression during delivery and this affects various physiological data of the body when training the model, thereby failing to fully reflect the patient's physical condition, and the existing prediction models acquire more and more complex physiological data, which will extract a large number of features during training and lead to high computational costs, as well as model overfitting, the present invention proposes an AI-based obstetric complication prediction system and method with low computational complexity and good prediction effect.

[0006] The AI-based obstetric complication prediction method includes the following steps: A pregnant woman sample data set is obtained through the hospital's historical database, wherein the pregnant woman sample data set includes the physiological data of pregnant women at various stages of pregnancy and corresponding complication information, the pregnant woman sample data set is preprocessed to obtain a valid pregnant woman sample data set, features are generated for the valid pregnant woman sample data set to obtain an initial fusion feature set, and an obstetric complication prediction model based on the RNN network model is trained with the help of the initial fusion feature set to obtain an initial obstetric complication prediction model; Obtain a psychological state data set and a physiological data set of the pregnant woman to be predicted, obtain a new feature set after preprocessing and feature generation of the obtained data, add an input layer and a fusion layer to the network structure of the initial obstetric complications prediction model, obtain a final obstetric complications prediction model, input the new feature set into the final obstetric complications prediction model for prediction, and perform incremental learning on the final obstetric complications prediction model with the help of the new feature set and actual results until the prediction accuracy of the final obstetric complications prediction model reaches a preset accuracy threshold; Obtain the latest psychological state data and the latest physiological data of the pregnant woman to be predicted, preprocess and generate features for the latest data, and then input them into the final obstetric complications prediction model for analysis and prediction, obtain the prediction results, and give corresponding obstetric complications prevention measures based on the prediction results and mental health data.

[0007] Preferably, the physiological data includes but is not limited to the basic information of the pregnant woman, her medical history, pregnancy examination data, family genetic history and lifestyle.

[0008] Preferably, the preprocessing of the pregnant woman sample data set comprises the following steps: Missing value processing: processing the null values ​​or missing values ​​in the data set. When there are few missing values ​​and the amount of data is small, linear interpolation or polynomial interpolation is used to fill them. When there are few missing values ​​and the amount of data is large, the entire row containing the missing values ​​is deleted. When the proportion of missing values ​​in a column is very high, the entire column containing the missing values ​​is deleted. Outlier processing: using the Z-Score method or IQR method to identify extreme values ​​in the data set that do not conform to the expected pattern, and processing outliers by deletion or mean replacement; Data standardization: standardize data of different dimensions through minimum-maximum standardization or Z-Score standardization.

[0009] Preferably, the step of generating features includes: Feature extraction: automatically identifying and extracting useful feature representations from raw data, including height, weight, platelet count, and serum aspartate aminotransferase (AST); Feature construction: Based on domain knowledge and experience, manually design and create new features, including body mass index BMI and HELLP syndrome index ; Feature fusion combines features from different sources or different types to form a comprehensive feature representation.

[0010] Preferably, the specific operation steps of feature extraction, feature construction and feature fusion for the effective data set of pregnant women samples are as follows: Extract physiological data features contained in the image through image processing technology, including but not limited to fetal development parameters, amniotic fluid index and placental thickness. For structured data and self-reported data, the corresponding physiological data features can be directly read, including but not limited to age, height, weight and the content of various cells and enzymes in the blood; According to existing medical knowledge, by selecting, combining and transforming relevant physiological data features, including but not limited to height, weight, platelet count and serum aspartate aminotransferase content AST, new physiological data features are constructed, including but not limited to body mass index BMI and HELLP syndrome index. , where the calculation formula for body mass index (BMI) is: ;in represents the weight of the pregnant woman in kilograms, and It indicates the height of the pregnant woman in meters, where the HELLP syndrome index The calculation formula is: ;in represents the hemolysis score, indicates liver enzyme score, Indicates platelet count score, which is given according to whether there is abnormal red blood cell morphology in the hemolysis test. Assign a value of 1 or 0 according to the serum aspartate aminotransferase (AST) content in the blood. Assignment, when When, give Assign 0, when When, give Assign value 1, when When, give Assign 2, according to the platelet count Assignment, when platelet count When, give Assign 0, when Platelet count When, give Assign a value of 1 when the platelet count When, give Assign 2; The newly constructed physiological data features are used to replace the corresponding extracted physiological data features, and the non-replaced and newly constructed physiological data features are fused into initial fused features to be input through a multimodal fusion method.

[0011] Preferably, the training steps of the initial obstetric complications prediction model include: Obtain an initial fusion feature set, and divide the initial fusion feature set into an initial fusion feature training set and an initial fusion feature test set at a ratio of 9:1; The initial fusion feature training set is input into the RNN neural network model for iterative training to obtain the obstetric complications prediction model; Based on the cross-validation method, the internal parameters of the obstetric complications prediction model were optimized to obtain the optimized obstetric complications prediction model; The optimized obstetric complications prediction model is tested based on the initial fusion feature test set. When the test results are qualified, the initial obstetric complications prediction model is obtained.

[0012] Preferably, the specific operation steps of obtaining the psychological state data and constructing the features are as follows: A psychological assessment questionnaire was used to investigate the psychological state of pregnant women, including the depression scale PHQ-9, the anxiety scale GAD-7 and the stress scale PSS. , anxiety scale score and stress scale scores ; Constructing psychological data features: mental health index , and the mental health index The calculation formula is: ;in , and denote the weights of the depression scale, anxiety scale, and stress scale, respectively, and , using this mental health index To assess the overall mental health status of pregnant women.

[0013] Preferably, the initial obstetric complications prediction model is a two-layer RNN model, including an input layer, two hidden layers and an output layer, the input layer is used to receive the initial fusion features, the hidden layer is mainly used to extract deeper nonlinear relationships in the initial fusion features to obtain comprehensive feature representation, and the output layer is used to output the prediction results; An input layer and a fusion layer are added to the two-layer RNN model of the initial obstetric complications prediction model to obtain the final obstetric complications prediction model. The network structure of the final obstetric complications prediction model includes two input layers, a fusion layer, two hidden layers and an output layer. The newly added input layer is used to receive psychological data features, and the newly added fusion layer is mainly used to concatenate and fuse the initial fusion features and the psychological data features to obtain the final fusion features. The subsequent two hidden layers are used to extract deeper nonlinear relationships in the final fusion features.

[0014] Preferably, the obstetric complication prevention measures include but are not limited to adjusting the treatment plan, strengthening the lifestyle management of pregnant women and providing relevant health education information, and the adjustment of the treatment plan includes a drug treatment plan and a psychological treatment plan. The drug treatment plan is set based on the model prediction results, and the psychological treatment plan is based on the mental health index. Make settings.

[0015] The AI-based obstetric complication prediction system includes data acquisition module, data preprocessing module, feature engineering module, complication prediction module, preventive measures module and result display module; Data acquisition module: used to obtain physiological data and psychological state data of pregnant women whose obstetric complications are to be predicted; Data preprocessing module: used to preprocess the acquired physiological data and psychological state data of pregnant women to obtain valid data after preprocessing; Feature engineering module: used to extract, construct and fuse features of preprocessed valid data to obtain the final features after fusion; Complication prediction module: used to input the final fused features into the final obstetric complication prediction model for analysis and prediction, and obtain the prediction results of obstetric complications; Preventive measures module: used to provide corresponding preventive measures for obstetric complications based on the prediction results; Result display module: used to visualize the prediction results of obstetric complications and corresponding preventive measures.

[0016] The present invention has the following advantages: The present invention takes the psychological state data of pregnant women into consideration, that is, takes the psychological state data of pregnant women as part of the model input and training data, so that the obstetric complications prediction model can have a more comprehensive understanding of the patient's health status, thereby helping the model to make more accurate predictions. At the same time, the obstetric complications prediction model is optimized through the incremental learning method, which not only can gradually correct the model by introducing psychological state data to avoid overfitting historical data, thereby making the model more generalized, but also can save computing resources, reduce computing costs and time overhead, and there is no need to save a large amount of historical data, only necessary parameters and statistical information need to be retained, thereby reducing storage requirements.

[0017] The present invention can construct new features according to existing medical knowledge, so that the constructed new features can better capture the potential patterns and relationships in the data than the original features, thereby effectively improving the predictive ability of the obstetric complications prediction model, and the feature construction can also remove or reduce the noise in the data, so that the model can focus more on important information, thereby improving the prediction accuracy, and using the newly constructed features to replace the corresponding extracted features can also integrate and simplify various physiological data, thereby reducing the amount of model calculation and reducing the feature dimension, which not only improves the efficiency of the model, but also reduces the risk of overfitting; and by integrating data from multiple sources such as physiological data and psychological state data through feature fusion, the model can be more robust and generalized. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the structure of the AI-based obstetric complications prediction system adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0020] Example 1, an AI-based obstetric complication prediction method, comprising the following steps: A pregnant woman sample data set is obtained through the hospital's historical database, wherein the pregnant woman sample data set includes the physiological data of pregnant women at various stages of pregnancy and corresponding complication information, the pregnant woman sample data set is preprocessed to obtain a valid pregnant woman sample data set, features are generated for the valid pregnant woman sample data set to obtain an initial fusion feature set, and an obstetric complication prediction model based on the RNN network model is trained with the help of the initial fusion feature set to obtain an initial obstetric complication prediction model; Obtain the psychological state data set and physiological data set of the pregnant woman to be predicted, obtain a new feature set after preprocessing and feature generation of the obtained data, add an input layer and a fusion layer to the network structure of the initial obstetric complication prediction model, obtain the final obstetric complication prediction model, input the new feature set into the final obstetric complication prediction model for prediction, and use the new feature set and actual results to perform incremental learning on the final obstetric complication prediction model until the prediction accuracy of the final obstetric complication prediction model reaches the preset accuracy threshold, so that the psychological state data of the pregnant woman can also be taken into consideration, that is, the psychological state data of the pregnant woman is also used as part of the model input and training data, so that the obstetric complication prediction model can have a more comprehensive understanding of the patient's health status, thereby helping the model to make more accurate predictions, and at the same time optimize the obstetric complication prediction model through the incremental learning method, which can not only gradually correct the model by introducing the psychological state data, avoid overfitting historical data, thereby making the model more generalized, but also save computing resources, reduce computing costs and time overhead, and do not need to save a large amount of historical data, only necessary parameters and statistical information need to be retained, thereby reducing storage requirements; Obtain the latest psychological state data and the latest physiological data of the pregnant woman to be predicted, preprocess and generate features for the latest data, and then input them into the final obstetric complications prediction model for analysis and prediction, obtain the prediction results, and give corresponding obstetric complications prevention measures based on the prediction results and mental health data.

[0021] The physiological data includes but is not limited to the basic information of the pregnant woman, her medical history, pregnancy examination data, family genetic history and lifestyle.

[0022] The preprocessing of the pregnant woman sample data set includes the following steps: Missing value processing: processing the null values ​​or missing values ​​in the data set. When there are few missing values ​​and the amount of data is small, linear interpolation or polynomial interpolation is used to fill them. When there are few missing values ​​and the amount of data is large, the entire row containing the missing values ​​is deleted. When the proportion of missing values ​​in a column is very high, the entire column containing the missing values ​​is deleted. Outlier processing: using the Z-Score method or IQR method to identify extreme values ​​in the data set that do not conform to the expected pattern, and processing outliers by deletion or mean replacement; Data standardization: standardize data of different dimensions through minimum-maximum standardization or Z-Score standardization.

[0023] The steps of feature generation include: Feature extraction: automatically identifying and extracting useful feature representations from raw data, including height, weight, platelet count, and serum aspartate aminotransferase (AST); Feature construction: Based on domain knowledge and experience, manually design and create new features, including body mass index BMI and HELLP syndrome index ; Feature fusion combines features from different sources or different types to form a comprehensive feature representation.

[0024] The specific steps for feature extraction, feature construction and feature fusion of the valid data set of pregnant women samples are as follows: Extract physiological data features contained in the image through image processing technology, including but not limited to fetal development parameters, amniotic fluid index and placental thickness. For structured data and self-reported data, the corresponding physiological data features can be directly read, including but not limited to age, height, weight and the content of various cells and enzymes in the blood; According to existing medical knowledge, by selecting, combining and transforming relevant physiological data features, including but not limited to height, weight, platelet count and serum aspartate aminotransferase content AST, new physiological data features are constructed, including but not limited to body mass index BMI and HELLP syndrome index. , constructing new features according to existing medical knowledge will enable the constructed new features to capture the potential patterns and relationships in the data better than the original features, thereby effectively improving the predictive ability of the obstetric complications prediction model. Feature construction can also remove or reduce noise in the data, allowing the model to focus more on important information, thereby improving prediction accuracy. The calculation formula for body mass index BMI is: ;in represents the weight of the pregnant woman in kilograms, and It indicates the height of the pregnant woman in meters, where the HELLP syndrome index The calculation formula is: ;in represents the hemolysis score, indicates liver enzyme score, Indicates platelet count score, which is given according to whether there is abnormal red blood cell morphology in the hemolysis test. Assign a value of 1 or 0 according to the serum aspartate aminotransferase (AST) content in the blood. Assignment, when When, give Assign 0, when When, give Assign value 1, when When, give Assign 2, according to the platelet count Assignment, when platelet count When, give Assign 0, when Platelet count When, give Assign a value of 1 when the platelet count When, give Assign 2; The newly constructed physiological data features are used to replace the corresponding extracted physiological data features, and the non-replaced and newly constructed physiological data features are fused into the initial fusion features to be input through the multimodal fusion method. The use of the newly constructed features to replace the corresponding extracted features can integrate and simplify various physiological data, thereby reducing the amount of model calculation and reducing the feature dimension. This not only improves the efficiency of the model, but also reduces the risk of overfitting. Integrating data from multiple sources through feature fusion can make the model more robust and generalizable.

[0025] The training steps of the initial obstetric complications prediction model include: Obtain an initial fusion feature set, and divide the initial fusion feature set into an initial fusion feature training set and an initial fusion feature test set at a ratio of 9:1; The initial fusion feature training set is input into the RNN neural network model for iterative training to obtain the obstetric complications prediction model; Based on the cross-validation method, the internal parameters of the obstetric complications prediction model were optimized to obtain the optimized obstetric complications prediction model; The optimized obstetric complications prediction model is tested based on the initial fusion feature test set. When the test results are qualified, the initial obstetric complications prediction model is obtained.

[0026] The specific steps of obtaining the psychological state data and constructing the features are as follows: A psychological assessment questionnaire was used to investigate the psychological state of pregnant women, including the depression scale PHQ-9, the anxiety scale GAD-7 and the stress scale PSS. , anxiety scale score and stress scale scores ; Constructing psychological data features: mental health index , and the mental health index The calculation formula is: ;in , and denote the weights of the depression scale, anxiety scale, and stress scale, respectively, and , using this mental health index To assess the overall mental health status of pregnant women.

[0027] The initial obstetric complications prediction model is a two-layer RNN model, including an input layer, two hidden layers and an output layer. The input layer is used to receive the initial fusion features, the hidden layer is mainly used to extract deeper nonlinear relationships in the initial fusion features and obtain comprehensive feature representations, and the output layer is used to output prediction results. An input layer and a fusion layer are added to the two-layer RNN model of the initial obstetric complications prediction model to obtain the final obstetric complications prediction model. The network structure of the final obstetric complications prediction model includes two input layers, a fusion layer, two hidden layers and an output layer. The newly added input layer is used to receive psychological data features, and the newly added fusion layer is mainly used to concatenate and fuse the initial fusion features and the psychological data features to obtain the final fusion features. The subsequent two hidden layers are used to extract deeper nonlinear relationships in the final fusion features.

[0028] The obstetric complication prevention measures include but are not limited to adjusting the treatment plan, strengthening the lifestyle management of pregnant women and providing relevant health education information, and the adjustment of the treatment plan includes a drug treatment plan and a psychological treatment plan. The drug treatment plan is set based on the model prediction results, and the psychological treatment plan is based on the mental health index. Make settings.

[0029] The above is based on the mental health index The specific psychological treatment plan is as follows: when When the pregnant woman's mental condition is "normal", the corresponding treatment plan is: provide lectures on mental health knowledge during pregnancy through obstetric clinics or online platforms to help pregnant women understand common psychological problems and how to deal with them, organize pregnant women support group activities, and let expectant mothers share experiences and emotions and support each other; when When the pregnant woman's mental condition is "sub-healthy", the corresponding treatment plan is: set up a special psychological counseling clinic, and trained psychological counselors provide short-term psychological counseling or guidance to pregnant women, provide brief intervention of cognitive behavioral therapy, and help pregnant women identify and change negative thinking patterns, especially concerns about pregnancy and childcare; when When the pregnant woman's psychological condition is "warning", the corresponding treatment plan is: the obstetrician and gynecologist and the psychiatrist jointly evaluate the pregnant woman's psychological state, determine the nature and severity of the problem, and then select appropriate psychological treatment methods based on the evaluation results, such as long-term psychological counseling, cognitive behavioral therapy (CBT) and psychodynamic therapy, and arrange regular follow-up. If necessary, the psychiatrist prescribes a small amount of safe drugs; when When the pregnant woman's psychological condition is "crisis", the corresponding treatment plan is: arrange for the pregnant woman to be hospitalized for close monitoring and round-the-clock care, use proven drugs that are relatively safe for pregnant women and fetuses to control symptoms, such as certain antipsychotics, antidepressants, and antianxiety drugs, etc. However, they must be used under the guidance of a professional doctor and the side effects of the drugs must be monitored regularly. At the same time, combined with psychotherapy, help pregnant women cope with crises and rebuild psychological balance.

[0030] In this way, different treatment plans can be provided in stages and on a personalized basis for pregnant women with different psychological conditions, which can not only improve the treatment effect, but also prevent the problem from worsening, enhance the confidence of pregnant women and their families, optimize the use of medical resources, improve patient satisfaction, and ultimately ensure the safety and health of mothers and babies.

[0031] Example 2, based on Example 1, an obstetric complications prediction system based on AI, such as Figure 1 As shown, it includes a data acquisition module, a data preprocessing module, a feature engineering module, a complication prediction module, a preventive measures module and a result display module; Data acquisition module: used to obtain physiological data and psychological state data of pregnant women whose obstetric complications are to be predicted; Data preprocessing module: used to preprocess the acquired physiological data and psychological state data of pregnant women to obtain valid data after preprocessing; Feature engineering module: used to extract, construct and fuse features of preprocessed valid data to obtain the final features after fusion; Complication prediction module: used to input the final fused features into the final obstetric complication prediction model for analysis and prediction, and obtain the prediction results of obstetric complications; Preventive measures module: used to provide corresponding preventive measures for obstetric complications based on the prediction results; Result display module: used to visualize the prediction results of obstetric complications and corresponding preventive measures.

[0032] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. An AI-based obstetric complication prediction method, characterized in that: The following steps are involved: A pregnant woman sample data set is obtained through the hospital's historical database, wherein the pregnant woman sample data set includes the physiological data of pregnant women at various stages of pregnancy and corresponding complication information, the pregnant woman sample data set is preprocessed to obtain a valid pregnant woman sample data set, features are generated for the valid pregnant woman sample data set to obtain an initial fusion feature set, and an obstetric complication prediction model based on the RNN network model is trained with the help of the initial fusion feature set to obtain an initial obstetric complication prediction model; Obtain a psychological state data set and a physiological data set of the pregnant woman to be predicted, obtain a new feature set after preprocessing and feature generation of the obtained data, add an input layer and a fusion layer to the network structure of the initial obstetric complications prediction model, obtain a final obstetric complications prediction model, input the new feature set into the final obstetric complications prediction model for prediction, and perform incremental learning on the final obstetric complications prediction model with the help of the new feature set and actual results until the prediction accuracy of the final obstetric complications prediction model reaches a preset accuracy threshold; Obtain the latest psychological state data and the latest physiological data of the pregnant woman to be predicted, preprocess and generate features for the latest data, and then input them into the final obstetric complications prediction model for analysis and prediction, obtain the prediction results, and give corresponding obstetric complications prevention measures based on the prediction results and mental health data.

2. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The physiological data includes but is not limited to the basic information of the pregnant woman, her medical history, pregnancy examination data, family genetic history and lifestyle.

3. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The preprocessing of the pregnant woman sample data set includes the following steps: Missing value processing: processing the null values ​​or missing values ​​in the data set. When there are few missing values ​​and the amount of data is small, linear interpolation or polynomial interpolation is used to fill them. When there are few missing values ​​and the amount of data is large, the entire row containing the missing values ​​is deleted. When the proportion of missing values ​​in a column is very high, the entire column containing the missing values ​​is deleted. Outlier processing: using the Z-Score method or IQR method to identify extreme values ​​in the data set that do not conform to the expected pattern, and processing outliers by deletion or mean replacement; Data standardization: standardize data of different dimensions through minimum-maximum standardization or Z-Score standardization.

4. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The steps of feature generation include: Feature extraction: automatically identifying and extracting useful feature representations from raw data, including height, weight, platelet count, and serum aspartate aminotransferase (AST); Feature construction: Based on domain knowledge and experience, manually design and create new features, including body mass index BMI and HELLP syndrome index ; Feature fusion combines features from different sources or different types to form a comprehensive feature representation.

5. The AI-based obstetric complication prediction method according to claim 4, characterized in that: The specific steps for feature extraction, feature construction and feature fusion of the valid data set of pregnant women samples are as follows: Extract physiological data features contained in the image through image processing technology, including but not limited to fetal development parameters, amniotic fluid index and placental thickness. For structured data and self-reported data, the corresponding physiological data features can be directly read, including but not limited to age, height, weight and the content of various cells and enzymes in the blood; According to existing medical knowledge, by selecting, combining and transforming relevant physiological data features, including but not limited to height, weight, platelet count and serum aspartate aminotransferase content AST, new physiological data features are constructed, including but not limited to body mass index BMI and HELLP syndrome index. , where the calculation formula for body mass index (BMI) is: ;in represents the weight of the pregnant woman in kilograms, and It indicates the height of the pregnant woman in meters, where the HELLP syndrome index The calculation formula is: ;in represents the hemolysis score, indicates liver enzyme score, Indicates platelet count score, which is given according to whether there is abnormal red blood cell morphology in the hemolysis test. Assign a value of 1 or 0 according to the serum aspartate aminotransferase (AST) content in the blood. Assignment, when When, give Assign 0, when When, give Assign value 1, when When, give Assign 2, according to the platelet count Assignment, when platelet count When, give Assign 0, when Platelet count When, give Assign a value of 1 when the platelet count When, give Assign 2; The newly constructed physiological data features are used to replace the corresponding extracted physiological data features, and the non-replaced and newly constructed physiological data features are fused into initial fused features to be input through a multimodal fusion method.

6. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The training steps of the initial obstetric complications prediction model include: Obtain an initial fusion feature set, and divide the initial fusion feature set into an initial fusion feature training set and an initial fusion feature test set at a ratio of 9:1; The initial fusion feature training set is input into the RNN neural network model for iterative training to obtain the obstetric complications prediction model; Based on the cross-validation method, the internal parameters of the obstetric complications prediction model were optimized to obtain the optimized obstetric complications prediction model; The optimized obstetric complications prediction model is tested based on the initial fusion feature test set. When the test results are qualified, the initial obstetric complications prediction model is obtained.

7. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The specific steps of obtaining the psychological state data and constructing the features are as follows: A psychological assessment questionnaire was used to investigate the psychological state of pregnant women, including the depression scale PHQ-9, the anxiety scale GAD-7 and the stress scale PSS. , anxiety scale score and stress scale scores ; Constructing psychological data features: mental health index , and the mental health index The calculation formula is: ;in , and denote the weights of the depression scale, anxiety scale, and stress scale, respectively, and , using this mental health index To assess the overall mental health status of pregnant women.

8. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The initial obstetric complications prediction model is a two-layer RNN model, including an input layer, two hidden layers and an output layer. The input layer is used to receive the initial fusion features, the hidden layer is mainly used to extract deeper nonlinear relationships in the initial fusion features and obtain comprehensive feature representations, and the output layer is used to output prediction results. An input layer and a fusion layer are added to the two-layer RNN model of the initial obstetric complications prediction model to obtain the final obstetric complications prediction model. The network structure of the final obstetric complications prediction model includes two input layers, a fusion layer, two hidden layers and an output layer. The newly added input layer is used to receive psychological data features, and the newly added fusion layer is mainly used to concatenate and fuse the initial fusion features and the psychological data features to obtain the final fusion features. The subsequent two hidden layers are used to extract deeper nonlinear relationships in the final fusion features.

9. The AI-based obstetric complication prediction method according to claim 1, characterized in that: The obstetric complication prevention measures include but are not limited to adjusting the treatment plan, strengthening the lifestyle management of pregnant women and providing relevant health education information, and the adjustment of the treatment plan includes a drug treatment plan and a psychological treatment plan. The drug treatment plan is set based on the model prediction results, and the psychological treatment plan is based on the mental health index. Make settings.

10. An AI-based obstetric complication prediction system, applied to the AI-based obstetric complication prediction method according to any one of claims 1 to 9, characterized in that: It includes data acquisition module, data preprocessing module, feature engineering module, complication prediction module, preventive measures module and result display module; Data acquisition module: used to obtain physiological data and psychological state data of pregnant women whose obstetric complications are to be predicted; Data preprocessing module: used to preprocess the acquired physiological data and psychological state data of pregnant women to obtain valid data after preprocessing; Feature engineering module: used to extract, construct and fuse features of preprocessed valid data to obtain the final features after fusion; Complication prediction module: used to input the final fused features into the final obstetric complication prediction model for analysis and prediction, and obtain the prediction results of obstetric complications; Preventive measures module: used to provide corresponding preventive measures for obstetric complications based on the prediction results; Result display module: used to visualize the prediction results of obstetric complications and corresponding preventive measures.

Images