Full-course follow-up management system and method for obstetrics and gynecology patients

Abstract

The present invention discloses a full-course follow-up management system and method for obstetrics and gynecology patients, which relates to the field of medical information management, including: when a patient visits a doctor for the first time, the patient's basic information, medical-related information, medical imaging data, genetic testing data, sensor data and patient self-report data are entered; the entered patient data is preprocessed, and the preprocessed patient data is integrated into a personal health record; a health assessment model for the patient is established based on the integrated personal health record, the patient's current health assessment score is output, and a follow-up plan is generated according to the patient's current health assessment score; an obstetric complication prediction model is established based on the personal health record, and the patient's complication prediction value is output. The present invention identifies potential obstetric complication risks in advance, especially when the health status of pregnant women changes, and can dynamically update the prediction results, realize real-time monitoring and early warning of complications, and optimize the safety management of parturients.

Description

Technical Field

[0001] The present invention relates to the field of medical information management, and in particular to a full-course follow-up management system and method for obstetrics and gynecology patients. Background Art

[0002] With the development of medical information technology, especially the progress in data processing and artificial intelligence, intelligent management and decision-making support functions in the medical field have attracted attention. As an important branch of medicine, obstetrics and gynecology often sees rapid changes in the health status of patients, especially during pregnancy and delivery. The health status of pregnant women is affected by many factors, including physiology, genetics, and environment. In order to achieve accurate monitoring of maternal health, traditional obstetrics and gynecology follow-up management methods have gradually introduced technical means such as electronic medical records, wearable device data collection, and genetic testing.

[0003] Existing methods mainly focus on the collection and management of static information, and it is difficult to effectively monitor and warn of dynamic changes in patients' health status. In addition, existing technologies often lack integration when processing multi-source data, and cannot fully utilize patients' dynamic physiological data, genetic test data, and self-reported information to effectively predict potential obstetric complications. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a full-course follow-up management system and method for obstetrics and gynecology patients, which solves the problem of lack of dynamic obstetric complication prediction in the prior art and inability to achieve personalized follow-up management and real-time intervention.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for the full-course follow-up management of obstetrics and gynecology patients, comprising:

[0008] When a patient visits the doctor for the first time, the patient's information data is entered;

[0009] Pre-process the entered patient information data and integrate the pre-processed patient information data into personal health records;

[0010] Establish a patient's health assessment model based on the integrated personal health records, output the patient's current health assessment score, and generate a follow-up plan based on the patient's current health assessment score;

[0011] Establish an obstetric complication prediction model based on personal health records and output the patient's complication prediction value;

[0012] Dynamically adjust the follow-up plan according to the patient's predicted complication value;

[0013] The method of establishing an obstetric complication prediction model based on personal health records and outputting a patient complication prediction value specifically includes the following steps:

[0014] Extract the types and descriptions of obstetric complications to form a multi-level medical concept from high to low

[0015] A dynamic feature update mechanism is used to extract features about the patient's health and form a health feature vector;

[0016] Establish a knowledge graph, and use the patient's basic information, patient health feature vectors, types of obstetric complications, and multi-level medical concepts as nodes of the knowledge graph;

[0017] The edges of the knowledge graph are formed through the relationship between the patient's basic information, the patient's health feature vector and the types of obstetric complications;

[0018] Connect to external medical knowledge bases through external API interfaces to automatically update nodes and edges in the graph;

[0019] Use the graph attention mechanism in the graph neural network to achieve adaptive adjustment of edge weights, and adjust the weight and direction of the edge by calculating the correlation between nodes;

[0020] When each node updates information, it dynamically selects neighbor nodes for aggregation based on the relevance and importance between nodes;

[0021] After multiple layers of transmission, the patient’s comprehensive health characteristics are obtained;

[0022] The comprehensive health features are input into the fully connected layer to further extract nonlinear features;

[0023] The predicted probability value of complications is generated through the output of the fully connected layer.

[0024] As a preferred solution of the full-course follow-up management method for obstetrics and gynecology patients of the present invention, when the patient visits the doctor for the first time, the patient information data is input, which specifically includes the following steps:

[0025] Patient information data includes basic patient information, medical-related information, medical imaging data, genetic testing data, sensor data, and patient self-reported data;

[0026] The patient's basic information includes the patient's name, age, telephone number, ID number, home address, occupation, and marital status;

[0027] The medical-related information includes past medical history, family history of genetic diseases, past pregnancy and childbirth history, current gestational age, date of last menstrual period, and history of drug allergy;

[0028] The medical imaging data includes an image of a fetus;

[0029] The genetic testing data includes maternal venous blood sample analysis data;

[0030] The sensor data includes the patient's heart rate, blood pressure, blood sugar, and fetal heart rate collected via a wearable device;

[0031] The patient self-reported data includes the patient's eating habits, exercise status, mood, psychological state and sleep status.

[0032] As a preferred solution of the whole-course follow-up management method for obstetrics and gynecology patients of the present invention, the input patient information data is preprocessed, and the preprocessed patient information data is integrated into a personal health record, which specifically includes the following steps:

[0033] Standardize the data format of entered patient information data;

[0034] Clean and denoise the patient information data after the data format is standardized;

[0035] Extract fetal development features from medical imaging data using convolutional neural networks;

[0036] Through feature selection methods, we can reduce low-correlation genes and extract risk coefficient features from gene testing data;

[0037] Extract physiological index characteristics from sensor data through time series analysis method;

[0038] The subjective scores in patients’ self-reports were used as subjective health score features;

[0039] The extracted features are fused through a weighted multimodal data fusion algorithm to obtain a multimodal feature vector;

[0040] Integrate the multimodal feature vector with the patient's basic information to generate the patient's personal health record.

[0041] As a preferred solution of the whole-course follow-up management method for obstetrics and gynecology patients described in the present invention, a health assessment model for patients is established based on the integrated personal health records, the current health assessment score of the patient is output, and a follow-up plan is generated according to the current health assessment score of the patient, which specifically includes the following steps:

[0042] Logarithmic transformation of multimodal feature vectors in personal health records;

[0043] Based on the degree of influence on the patient's health assessment, principal component analysis was used to assign weights to the multimodal feature vectors and the patient's basic information;

[0044] Through the generalized linear thinking and based on the personal health record, the patient's current health assessment score is output, and its expression is: ;in, Indicates the patient's current health assessment score. Represents the first Features, Indicates The weight coefficient of the feature, Indicates The weight coefficient of the feature in the denominator, Indicates the patient’s Basic information, Indicates The weight of basic information, represents the number of multimodal feature vectors, Indicates the number of basic information of patients;

[0045] A follow-up plan is generated based on the patient's current health assessment score. The follow-up plan includes a follow-up frequency and a follow-up content. The follow-up frequency and the follow-up content will be determined based on the patient's current health assessment score.

[0046] As a preferred solution of the full-course follow-up management method for obstetrics and gynecology patients of the present invention, the follow-up plan is dynamically adjusted according to the predicted value of complications of the patient, and the specific steps are as follows:

[0047] A four-level threshold for patient complications was established based on the patient's basic information;

[0048] The prediction result of the patient's complications is determined based on the interval of the patient's complication threshold value where the patient's complication prediction value is located, and the patient's health status in different time periods is dynamically evaluated through the patient's sensor data;

[0049] Generate a smart health calendar for the patient, automatically adjusting the follow-up plan based on the patient's complication prediction results and the patient's sensor data.

[0050] In a second aspect, the present invention provides a full-course follow-up management system for obstetrics and gynecology patients, comprising:

[0051] The information entry module collects and enters the patient's basic information, medical-related information, medical imaging data, genetic testing data, sensor data, and patient self-reported data;

[0052] The data processing module pre-processes the entered patient data and integrates it into the patient's personal health record;

[0053] The file management module stores and manages all the patient's health data and builds a comprehensive patient health file as the basis for subsequent health assessment and complication prediction;

[0054] The health assessment module builds a health assessment model for patients based on the integrated personal health records and outputs the current health assessment score to assess the overall health status of the patient;

[0055] Follow-up management module, which generates personalized follow-up plans based on the patient's health assessment scores and dynamically adjusts the frequency and content of follow-up according to complication prediction results;

[0056] The complication prediction module establishes an obstetric complication prediction model based on the patient's personal health record, predicts the patient's complication risk, and outputs specific complication prediction values.

[0057] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for full-course follow-up management of obstetric and gynecological patients as described in the first aspect of the present invention is implemented.

[0058] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for full-course follow-up management of obstetric and gynecological patients as described in the first aspect of the present invention is implemented.

[0059] The beneficial effects of the present invention are as follows: using a weighted multimodal data fusion algorithm, data information obtained from different sources is effectively integrated to generate a unified personal health record. The reliability and availability of the data are improved to ensure that subsequent analysis can be based on high-quality data. Through a dynamic feature update mechanism and knowledge graph technology, the patient's health feature vector is associated with the risk factors of obstetric complications. Through the graph attention mechanism in the graph neural network, the weights between different features can be dynamically adjusted, and complication prediction values ​​can be output based on the patient's real-time health status. Potential obstetric complication risks are identified in advance, especially when the health status of pregnant women changes, the prediction results can be dynamically updated. Through this obstetric complication prediction model, real-time monitoring and early warning of complications are achieved, and the safety management of mothers is optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0061] Figure 1 This is a flow chart of the full-course follow-up management method for obstetrics and gynecology patients in Example 1.

[0062] Figure 2 This is a decision diagram for the full-course follow-up management method for obstetrics and gynecology patients in Example 1. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0066] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for the full-course follow-up management of obstetrics and gynecology patients, comprising the following steps:

[0067] S1: When the patient visits the doctor for the first time, the patient information data is entered.

[0068] S1.1: Enter the patient’s basic information.

[0069] Entering the patient's basic information can not only uniquely identify the patient, but also conduct correlation analysis in combination with other health data. For example, age and marital status can provide early judgments about high-risk pregnant women, and occupation and place of residence may also reflect the potential impact of certain lifestyles and environments on health. By collecting the patient's basic information and establishing a preliminary health risk assessment based on demographic characteristics, it is helpful to formulate personalized management plans. In addition, the accurate entry of basic information can ensure the uniqueness of the patient's file, avoid data confusion, and enhance security and reliability.

[0070] The above information is subject to the user's consent and will be used for legal purposes.

[0071] S1.2: Enter the patient's medical information.

[0072] Entering the patient's medical-related information can help doctors quickly understand the patient's health background. For example, past medical history and family genetic history can reveal potential risks of genetic diseases, and past pregnancy and delivery history can help predict possible complications during future pregnancy and delivery. The number of weeks of pregnancy and the date of the last menstrual period are crucial for assessing fetal development and developing appropriate follow-up plans. A history of drug allergies can help avoid the use of drugs that may cause adverse reactions during treatment.

[0073] S1.3: Enter the patient’s fetal imaging data.

[0074] The fetal imaging data of the patient is key information for evaluating fetal development and maternal health. By collecting and analyzing these imaging data, the fetal development status can be obtained, including growth indicators such as fetal weight, head circumference, and limb length.

[0075] S1.4: Enter the patient’s genetic testing data.

[0076] Entering genetic test data is an important source of information for predicting the risk of genetic diseases and certain complex diseases (such as diabetes, hypertension, etc.). Genetic testing can identify potential genetic abnormalities in the mother or fetus and provide a basis for early intervention for high-risk patients, such as chromosomal abnormalities and risk assessment of certain single-gene genetic diseases.

[0077] S1.5: Enter the patient's sensor data.

[0078] The input of real-time sensor data can continuously monitor the physiological status of pregnant women and detect abnormal fluctuations in time. For example, abnormal increases in heart rate and blood pressure may indicate the occurrence of pregnancy-induced hypertension syndrome, and abnormal fetal heart rate may indicate fetal hypoxia. Blood sugar data helps monitor the management of gestational diabetes.

[0079] S1.6: Enter patient self-reported data.

[0080] Entering patient self-reported data is an important way to understand patients' lifestyles and mental states. These data can reflect patients' quality of life and compliance with health management. For example, eating habits and exercise can help assess patients' weight management and blood sugar control; emotions and mental states are important indicators for assessing maternal mental health, especially in the screening of depression during pregnancy and postpartum; abnormal sleep status may be related to physical fatigue, anxiety and other problems.

[0081] It is further explained that through the comprehensive collection of this series of data, a multi-dimensional, dynamically updated patient health record can be established, providing a solid data foundation for health assessment, follow-up plan formulation and complication prediction. The integration of various types of data can achieve all-round monitoring of patient health, which can not only identify potential risks in advance, but also dynamically adjust follow-up plans and intervention measures according to the patient's real-time status.

[0082] S2: Preprocess the entered patient information data and integrate the preprocessed patient information data into a personal health record.

[0083] S2.1: Standardize the format of the entered patient data. For example, text fields need to be converted to a structured data format, numerical data need to be converted to the same unit, and imaging data need to be standardized to the same resolution or format.

[0084] Further explanation: data standardization ensures that information from different data sources is identified and processed. This step improves the efficiency of data processing and avoids information loss or incorrect analysis due to inconsistent data formats.

[0085] S2.2: Clean and denoise the entered patient data. Specific cleaning steps include deleting duplicate data, filling or eliminating missing values, eliminating abnormal values ​​(such as abnormal data where the fetal heart rate exceeds the reasonable range), etc. For imaging data and genetic testing data, denoising is required through image processing algorithms and data filtering methods.

[0086] It is further explained that data cleaning and denoising significantly improve the quality and accuracy of the data and reduce errors caused by data anomalies. With the support of high-quality data, subsequent health assessment and complication prediction models can be calculated based on real and reliable data, thereby improving the accuracy of prediction and the rationality of follow-up management.

[0087] S2.3: Extract key developmental characteristics of the fetus, such as fetal weight, head circumference, limb length, etc. These developmental characteristics reflect the growth and development of the fetus and can identify whether the fetus is at risk of developmental delay or malformation.

[0088] S2.4: Use the filtering method in the feature selection method to calculate the correlation between each gene and disease risk, and select important genes according to the correlation score. The genes retained are the genes most related to health risks. These genes can be used to generate risk coefficient features, that is, the risk coefficient corresponding to each gene.

[0089] S2.5: Through the time series analysis method, statistical, time domain, frequency domain and dynamic characteristics are extracted from the physiological indicators recorded by the sensors, namely the physiological indicator characteristics, such as average heart rate, blood pressure fluctuation range, blood sugar control status, etc. These characteristics can reflect the patient's overall physiological state.

[0090] S2.6: The patient's self-reported data reflects their subjective health feelings, such as eating habits, emotional state, sleep quality, etc. By quantifying these self-reported data, subjective health score features are generated.

[0091] It is further explained that by extracting the features of various types of data, the patient's health status can be fully understood from different dimensions. Medical imaging features can reflect the development of the fetus, genetic testing features reveal genetic risks, sensor data provides real-time physiological monitoring, and patient self-report data reflects subjective health feelings. Multi-dimensional data features provide comprehensive input for subsequent health assessment models, ensuring that judgments are not only based on a single data type, but also that more accurate health analysis can be achieved through multimodal data fusion.

[0092] S2.7: After extracting various data features, these features are fused through a weighted multimodal data fusion algorithm. The algorithm assigns different weights to each type of data feature according to its influence on health assessment. For example, genetic testing data may have more weight in predicting the risk of genetic diseases, while sensor data is more important for assessing real-time health status. Through this weighted fusion method, a multimodal feature vector is generated, which is expressed as: ;in, represents the multimodal feature vector, represents the weight of fetal development characteristics, represents the weight of the risk factor feature, represents the weight of physiological characteristics, represents the weight of the patient's subjective health score feature, Indicates key indicator characteristics of fetal development (such as fetal weight, fetal head circumference, etc.), Represents risk factor characteristics (e.g., risk factor for chromosome 21 abnormality), Indicates physiological characteristics (such as blood pressure, heart rate, blood sugar and other physiological indicators), Represents the subjective health score features in the patient's self-reported data, such as the subjective health scores of diet, exercise, mental health, etc. recorded by the patient.

[0093] It is further explained that weighted multimodal data fusion effectively solves the problem of different importance of different data types in health assessment. Through reasonable weight distribution, it can ensure that the contribution of various types of data is fully reflected in the health assessment, avoiding the situation where a single data dominates the assessment results. This fusion method improves the comprehensiveness and accuracy of health assessment and provides a more reliable basis for personalized follow-up plans and complication prediction.

[0094] S2.8: After completing the data fusion, the multimodal feature vector is integrated with the patient’s basic information (such as age, medical history, etc.) to generate the patient’s personal health record.

[0095] This file not only records the patient's basic information, but also contains multimodal feature vectors generated by multimodal data fusion. Personal health records are the core data source for subsequent health assessment and complication prediction, and are also the basis for personalized follow-up and intervention for patients.

[0096] S3: Establish a health assessment model for the patient based on the integrated personal health records, output the patient's current health assessment score, and generate a follow-up plan based on the patient's current health assessment score.

[0097] S3.1: Perform logarithmic transformation on multimodal feature vectors in personal health records.

[0098] The multimodal feature vector is obtained by multimodal data fusion, which includes a variety of health-related information such as medical images, genetic testing, sensor data, and patient self-report data. Since these data may have different scales and distributions, in order to reduce the difference between eigenvalues ​​and enhance the stability of the model, the multimodal feature vector is first logarithmically transformed;

[0099] It is further shown that logarithmic transformation can effectively reduce the difference between different eigenvalues ​​and avoid model instability caused by excessive eigenvalues. Through this transformation, extreme values ​​in the data are eliminated, making the subsequent weight distribution more reasonable.

[0100] S3.2: After logarithmic transformation, principal component analysis (PCA) was used to reduce the dimension of multimodal feature vectors and basic patient information and assign weights.

[0101] It is further explained that principal component analysis can reduce redundant information, highlight the features that have the greatest impact on health assessment, and reduce data dimensions, thereby improving model calculation efficiency. By reasonably allocating weights, personalized assessment results can be generated based on the differences in the characteristics of different patients, avoiding inaccurate assessment results caused by the same weight of all features on the health score.

[0102] S3.3: After completing the weight assignment, the patient's health status is evaluated through the generalized linear model (GLM) and the current health assessment score is calculated, which is expressed as: ;in, Indicates the patient's current health assessment score. Represents the first Features, Indicates The weight coefficient of the feature, Indicates The weight coefficient of the feature in the denominator, Indicates the patient’s Basic information, Indicates The weight of basic information, represents the number of multimodal feature vectors, Indicates the number of basic information of the patient.

[0103] S3.4: Generate a follow-up plan based on the patient’s current health assessment scores.

[0104] A follow-up plan is formulated based on the patient's current health assessment score. The follow-up plan includes the follow-up frequency and follow-up content, which will be determined based on the patient's current health assessment score.

[0105] For example, the value range of the patient's current health assessment score is [0,1]. If the patient's current health assessment score is greater than or equal to 0.8, the patient is considered to be in good health and the follow-up frequency is set to once a month. The follow-up content includes routine physical examinations and basic health monitoring, such as blood pressure, heart rate, weight, etc. Based on the patient's health record, routine blood tests, urine tests and other basic screenings are performed. Routine fetal monitoring and prenatal examinations are performed.

[0106] If the patient's current health assessment score is greater than or equal to 0.5 and less than 0.8, the patient's health status is considered to require attention, and the follow-up frequency is set to once every two weeks. The follow-up content includes additional monitoring items, including blood sugar, blood lipids, liver and kidney function and other indicators. Conduct more detailed prenatal examinations such as fetal growth monitoring and amniotic fluid monitoring. According to the patient's personalized risks (such as obesity, diabetes, etc.), add corresponding examination items and guidance. Provide advice on healthy lifestyles, such as diet, exercise, and mental health.

[0107] If the patient's current health assessment score is less than 0.5, the patient is considered to have a high health risk, and the follow-up frequency is set to once a week. The follow-up content includes key monitoring of high-risk indicators, such as blood pressure, blood sugar, ECG monitoring, liver and kidney function, etc. Increase the frequency of fetal monitoring, such as fetal heart monitoring, ultrasound examination, amniotic fluid examination, etc. Develop a detailed intervention plan for the patient's specific risk factors (such as cardiovascular disease, pregnancy complications, etc.), including drug adjustment, diet control, etc. Provide psychological support and lifestyle guidance to help patients cope with health risks.

[0108] It is further explained that personalized health management can be achieved through this follow-up plan generation mechanism based on health assessment points. Patients with different health conditions will receive different follow-up frequencies and contents, thus avoiding a "one-size-fits-all" follow-up arrangement and improving the efficiency and pertinence of follow-up management. This dynamic follow-up plan adjustment mechanism helps to identify high-risk patients in a timely manner, and by increasing the frequency and content of follow-up, provide early intervention and reduce the risk of complications.

[0109] S4: Establish an obstetric complication prediction model based on personal health records and output the patient's complication prediction value.

[0110] S4.1: Through searching the medical literature database and hospital case database, extract the types of obstetric complications and their descriptions, and organize the collected medical concepts into levels to form multi-level medical concepts from high to low.

[0111] S4.2: Use a dynamic feature update mechanism to extract features about the patient's health from the personal health record to form a health feature vector that is dynamically adjusted according to the patient's health changes at different stages.

[0112] The health changes of the patient at different stages refer to the impact of historical health data on current health characteristics, and the dynamic change rates of different characteristics may be different;

[0113] For example, the update of feature vectors should not only depend on current health data, but also consider the impact of historical health data on current health features;

[0114] Different features may have different dynamic change rates because weight may change slowly, while blood sugar levels may change quickly. Therefore, the update mechanism of the feature vector can take into account the change rate of each feature;

[0115] The expression of the health feature vector that is dynamically adjusted according to the health changes of patients at different stages is: ;in, Indicates at time Patient at all times The health feature vector of represents a function for selecting and extracting health-related features from a person's health records, Indicates at time Health data at all times, Indicates time lag, Indicates time point The previous Health data at each moment, represents the time decay weight, Indicates a point in time, Indicates the length of historical data, express The change rate coefficient of each health characteristic, Indicates Health characteristics The rate of change over time can capture the dynamic changes of health characteristics by calculating the derivative, reflecting the changing trend of the characteristic at the current moment. Represents the total number of health characteristics.

[0116] It is further explained that according to the patient's health status and the rate of change of characteristics, a personalized health feature vector is generated to improve the adaptability of the model, which can capture the changing trend of health characteristics, generate more accurate health characteristics, and support more precise health assessment and prediction.

[0117] S4.3: Build a knowledge graph.

[0118] Knowledge graph is a knowledge representation method that uses a graph structure to represent entities (nodes) and their relationships (edges).

[0119] The nodes refer to the patient's basic information, the patient's health feature vector, the types of obstetric complications and the multi-level medical concepts, and the edges refer to the relationship between the patient's basic information, the health feature vector and the types of obstetric complications; the relationship between the patient's health feature vector and the multi-level medical concepts; and the relationship between the types of obstetric complications and the multi-level medical concepts.

[0120] To further illustrate, multi-level medical concepts (such as disease definition, etiology, symptoms, treatment plans, etc.) are not direct individual patient attributes or dynamic health characteristics, but descriptive and background information about obstetric complications and related knowledge. They are more like knowledge background, providing semantic interpretation and medical knowledge support for other nodes (patient information, health characteristics, complication types). Their edges are indirect, explanatory associations generated through other nodes, rather than directly interacting with the patient's individual information.

[0121] The relationship between the patient's basic information, health feature vectors, and types of obstetric complications refers to the relationship between these nodes, which is represented by the edges between them. For example, an older pregnant woman may be associated with the risk of gestational diabetes or pregnancy-induced hypertension; or abnormal blood sugar levels (a feature in the health feature vector) may be associated with complications of gestational diabetes.

[0122] The edge between the patient's health feature vector and the multi-level medical concept represents the connection between the patient's health features and existing medical knowledge. For example, a patient's blood pressure features and blood sugar features may correspond to the diagnostic criteria for hypertension and diabetes in medical knowledge.

[0123] The edges between the types of obstetric complications and multi-level medical concepts represent the connections between different obstetric complications and medical knowledge (such as treatment plans, diagnostic criteria, and preventive measures). For example, the relationship between gestational hypertension and specific drug treatment plans or monitoring measures can be represented by these edges.

[0124] S4.4: Obtain the latest medical information from external medical knowledge bases through APIs, including diseases, drugs, treatment plans, diagnostic criteria, laboratory indicators, etc., and set up regular weekly queries to external medical knowledge bases to obtain the latest medical knowledge.

[0125] Automatically update nodes in the graph based on the latest data obtained from the outside. For example, newly added diseases or treatment plans can be added to the graph as new nodes.

[0126] By analyzing the associations between various medical concepts in external data, the edges in the graph are updated. For example, if the latest research shows that a certain drug is effective for a specific disease, the edge between the drug and the disease will be automatically added or updated.

[0127] S4.5: Use the graph attention mechanism in the graph neural network to achieve adaptive adjustment of edge weights, and adjust the weight and direction of the edge by calculating the correlation between nodes.

[0128] By calculating the feature similarity and correlation between adjacent nodes, the graph attention mechanism can adaptively assign different weights to each edge. For the dynamic changes of health features over time, the weight of the edge can be adjusted in real time according to the feature differences between nodes.

[0129] Given two adjacent nodes in a graph and , whose eigenvectors are and , the expression of edge weight is: ;in, Representation Node For Node The weight of Representation Node In time The feature vector at time, represents the transpose operation, Representation Node In time The feature vector at time is the node The neighbor nodes of Representation Node The set of neighbor nodes is all nodes with Directly connected nodes, Representation Node Another neighbor node of , Represents a learnable weight matrix, which is used to linearly transform the feature vector of the node and map the feature vectors of different nodes into the same space, so that similarity calculation can be performed.

[0130] S4.6: After calculating the weights between nodes, dynamically select neighbor nodes with high correlation with the current node features for aggregation.

[0131] Further explanation: by calculating the weight of the edge, the model can automatically select which neighbor nodes have a greater impact on the current node. This means that the aggregation of information is no longer a simple average, but is weighted according to the similarity and correlation between nodes. By giving a larger weight to useful information, the model can more effectively aggregate neighbor information that is helpful to the current node and reduce the interference of noise nodes.

[0132] S4.7: After multi-layer propagation, the patient's comprehensive health characteristics are obtained, and the comprehensive health characteristics are input into the fully connected layer to further extract nonlinear features. Through the output of the fully connected layer, the predicted probability value of complications is generated.

[0133] Through the multi-layer propagation mechanism, the model propagates and aggregates the patient's health characteristics and the information of related nodes multiple times to obtain a comprehensive health feature vector, which is input into the first layer of the fully connected neural network. The fully connected neural network contains several hidden layers. The nodes in each layer are connected to all the nodes in the previous layer and linearly transformed, and then the results of the linear transformation are processed nonlinearly through nonlinear activation. The Softmax function is used to convert the output of the last layer into an easy-to-understand probability distribution of multiple complication categories, which is the predicted probability value of the patient's obstetric complications. The expression is: ;in, Indicates that when given an input, the category The probability of Represents the comprehensive health feature vector obtained after multi-layer propagation, represents the total number of categories, Indicates the specific category of complications. Indicates category The unnormalized score of Indicates The scores corresponding to the categories.

[0134] Further explanation: The multi-layer propagation mechanism allows the model to not only capture the health characteristics of the patient itself, but also effectively integrate the information of nodes related to the patient (such as other patients, medical knowledge nodes, feature nodes, etc.). This helps to capture complex relationships and interaction information, thereby improving the prediction accuracy of the model, especially when dealing with medical data with complex relationships.

[0135] S5: Dynamically adjust the follow-up plan based on the patient's predicted complication value.

[0136] S5.1: Establish a four-level threshold for patient complications based on the patient's basic information.

[0137] Based on different categories of complications (such as gestational diabetes, gestational hypertension, etc.), patient complication thresholds are established, including the first threshold (low risk), the second threshold (medium-low risk), the third threshold (medium-high risk) and the fourth threshold (urgent risk);

[0138] The first-level threshold (low risk) is 0.2, that is, when the patient's complication prediction value is greater than 0 and less than or equal to 0.2, it indicates that the patient is in good health and the possibility of complications is low.

[0139] The secondary threshold (medium-low risk) is 0.5, that is, when the patient's complication prediction value is greater than 0.2 and less than 0.5, it indicates that the patient has a slight risk of complications and requires regular monitoring.

[0140] The third-level threshold (medium-high risk) is 0.8, that is, when the patient's complication prediction value is greater than or equal to 0.5 and less than 0.8, it means that the patient has a significant risk of complications and needs to strengthen health monitoring and possibly take preventive measures.

[0141] The fourth-level threshold (urgent risk) is 1, that is, when the patient's complication prediction value is greater than or equal to 0.8 and less than or equal to 1, it indicates that the patient is at high risk of complications and requires immediate intervention measures and frequent follow-up and treatment.

[0142] S5.2: Determine the prediction result of the patient's complications based on the interval of the patient's complication threshold value where the patient's complication prediction value is located;

[0143] For example, the model predicts that a patient may develop three common obstetric complications (such as gestational diabetes, pregnancy-induced hypertension, and premature birth). At the output layer, the model's Softmax function may output the following probability distribution, This means that the probability of this patient developing gestational diabetes is 20%, and the probability of this patient developing gestational diabetes is 20%. According to the previously established complication threshold, this prediction value is within the first-level threshold (low risk).

[0144] The probability of gestational hypertension is 30%, and the patient has a 30% chance of developing gestational hypertension. This is within the secondary threshold (medium-low risk), indicating that the patient has a certain risk of complications and needs regular monitoring.

[0145] The probability of premature birth is 50%, and the patient has a 50% chance of having a premature birth. This is within the level 3 threshold (medium-high risk), indicating that the patient has a high risk of premature birth and needs to strengthen health monitoring and possibly take preventive measures.

[0146] The patient's health status will also be dynamically assessed in combination with the sensor data worn by the patient (such as blood sugar, blood pressure, fetal heart rate monitoring, etc.). If the sensor data indicates that the patient's health status is deteriorating, the patient's risk level will be automatically raised and the follow-up plan will be adjusted.

[0147] For example, if a patient's blood pressure rises sharply in a short period of time, even if the current model predicts a 30% risk of gestational hypertension (low to medium risk), the follow-up frequency will be adjusted in a timely manner based on the latest sensor data, and the patient may be advised to seek medical attention immediately or undergo more frequent monitoring.

[0148] S5.3: Automatically generate a personalized smart health calendar for the patient based on the patient's complication prediction value, risk level, and sensor data. The calendar includes follow-up dates, important health reminders (such as the time to measure blood sugar or blood pressure), and health management suggestions (such as diet adjustments or exercise plans).

[0149] If the model predicts that a patient’s risk of complications is increasing, or if sensor data indicates that the patient’s health is deteriorating, follow-up dates are automatically adjusted.

[0150] For example, if a patient's risk of gestational hypertension increases from 30% (low to intermediate risk) to 50% (high to intermediate risk), the follow-up date is moved forward and the physician and patient are notified.

[0151] If the patient's health improves, the follow-up interval may be extended to avoid excessive follow-up.

[0152] It is further explained that by dynamically adjusting the follow-up plan, it can ensure that high-risk patients receive timely intervention, reduce the incidence and severity of complications, and the smart health calendar and automatic reminder mechanism can help patients better manage their health and conduct necessary tests and follow-ups in a timely manner, thereby increasing patients' participation in health management.

[0153] When the model predicts that the patient's complication risk value is greater than or equal to 0.8, it means that the patient is in a fourth-level emergency risk state and may face serious health problems such as a sharp deterioration of gestational diabetes, pregnancy-induced hypertension, or premature birth.

[0154] This means that the patient needs immediate medical intervention to avoid further complications or life-threatening consequences.

[0155] When in an emergency risk state, emergency notifications will be automatically sent to the patient's attending physician, medical team, and relevant medical emergency departments;

[0156] The notification content includes the patient's current health status, real-time monitoring data (such as blood sugar, blood pressure, heart rate, etc.), historical medical records, and complication prediction results, ensuring that doctors can quickly understand the patient's health status and make necessary medical decisions.

[0157] At the same time, a notification containing a health risk assessment is automatically sent to the patient, indicating the current emergency risk status and describing in detail the predicted type of complications (such as gestational diabetes, pregnancy-induced hypertension, premature birth, etc.).

[0158] For example, the notification might state: "Your health monitoring shows that you currently have an 85% chance of developing gestational hypertension. Please contact your doctor immediately and follow the following suggestions for initial treatment," along with preliminary recommendations including diet control, rest, and temporary care measures.

[0159] The dietary control includes advising the patient to immediately stop consuming a high-sugar or high-salt diet.

[0160] The rest includes advising the patient to rest in bed and avoid strenuous activities.

[0161] Interim care measures include advising patients to measure their blood pressure or blood glucose meters immediately and upload the latest data if they have them. In addition, some home care advice may be provided, such as how to properly use medications or how to contact emergency services.

[0162] A personalized intervention plan is developed based on the patient's personal health records (including age, medical history, genetic background, etc.), recent follow-up records (such as the results of the most recent physical examination, doctor's advice, etc.), and the patient's living environment (such as whether he or she lives in a remote area, whether he or she is accompanied by family members, etc.).

[0163] For example, if a patient's medical history shows that she has a family history of hypertension, more attention will be paid to changes in blood pressure and relevant intervention measures will be prioritized.

[0164] If the patient lives far from the hospital, home medical services or telemedicine consultations may be recommended to ensure that the patient can receive timely medical support.

[0165] During the implementation of the intervention program, the patient's health data will be continuously collected through the patient's wearable devices (such as blood glucose meters, blood pressure monitors, heart rate monitors, etc.).

[0166] These data will be uploaded in real time and the intervention plan will be adjusted according to the changes in the patient's health data. For example, if the blood sugar level does not improve significantly after the use of the drug, the doctor will be prompted to adjust the drug dosage or change the medication regimen.

[0167] After the emergency intervention is completed, a follow-up plan is automatically generated based on the patient’s latest health status. This plan may include regular physical examinations, remote health monitoring, and doctor’s visits when necessary.

[0168] The frequency of follow-up visits is dynamically adjusted based on the patient’s health risk level. For example, for patients who have just left an emergency state, the follow-up interval may be once a week, while for patients whose health conditions are gradually stabilizing, the follow-up interval may be extended to once a month.

[0169] Patients’ health status will continue to be tracked through wearable devices and other data sources (such as regularly uploaded physical examination reports). If abnormal changes occur, doctors and patients will be reminded to take new intervention measures in a timely manner.

[0170] Further explanation: through real-time risk prediction and automated notification mechanism, doctors and medical teams can be quickly notified when patients are at emergency risk, ensuring that medical intervention can be carried out in the shortest possible time. This automated process greatly shortens the medical response time, especially for emergency treatment in high-risk situations. Personalized intervention plans are based on patients' health records, follow-up records and real-time data, and can best meet the individual needs of patients. Compared with traditional standardized treatment plans, personalized plans are more targeted, which helps to improve the effectiveness of treatment and reduce side effects.

[0171] This embodiment also provides a full-course follow-up management system for obstetrics and gynecology patients, including:

[0172] The information entry module collects and enters the patient's basic information, medical-related information, medical imaging data, genetic testing data, sensor data, and patient self-reported data;

[0173] The data processing module pre-processes the entered patient data and integrates it into the patient's personal health record;

[0174] The file management module stores and manages all the patient's health data and builds a comprehensive patient health file as the basis for subsequent health assessment and complication prediction;

[0175] The health assessment module builds a health assessment model for patients based on the integrated personal health records and outputs the current health assessment score to assess the overall health status of the patient;

[0176] Follow-up management module, which generates personalized follow-up plans based on the patient's health assessment scores and dynamically adjusts the frequency and content of follow-up according to complication prediction results;

[0177] The complication prediction module establishes an obstetric complication prediction model based on the patient's personal health record, predicts the patient's complication risk, and outputs specific complication prediction values.

[0178] This embodiment also provides a computer device, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to implement the full-course follow-up management method for obstetrics and gynecology patients proposed in the above embodiment when executing the computer executable instructions.

[0179] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0180] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the full-course follow-up management of obstetric and gynecological patients proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memor, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0181] In summary, the present invention effectively integrates data information obtained from different sources by utilizing a weighted multimodal data fusion algorithm to generate a unified personal health record. The reliability and availability of the data are improved to ensure that subsequent analysis can be based on high-quality data. Through the dynamic feature update mechanism and knowledge graph technology, the patient's health feature vector is associated with the risk factors of obstetric complications. Through the graph attention mechanism in the graph neural network, the weights between different features can be dynamically adjusted, and the complication prediction value can be output based on the patient's real-time health status. Potential obstetric complication risks are identified in advance, especially when the health status of pregnant women changes, the prediction results can be dynamically updated. Through this obstetric complication prediction model, real-time monitoring and early warning of complications are achieved, and the safety management of mothers is optimized.

[0182] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for the full-course follow-up management of obstetrics and gynecology patients, characterized in that: include: When a patient visits the doctor for the first time, the patient's information data is entered; Pre-process the entered patient information data and integrate the pre-processed patient information data into personal health records; Establish a patient's health assessment model based on the integrated personal health records, output the patient's current health assessment score, and generate a follow-up plan based on the patient's current health assessment score; Establish an obstetric complication prediction model based on personal health records and output the patient's complication prediction value; Dynamically adjust the follow-up plan according to the patient's predicted complication value; The method of establishing an obstetric complication prediction model based on personal health records and outputting a patient complication prediction value specifically includes the following steps: Extract the types of obstetric complications and their descriptions to form a multi-level medical concept from high to low; A dynamic feature update mechanism is used to extract features about the patient's health and form a health feature vector; Establish a knowledge graph, and use the patient's basic information, patient health feature vectors, types of obstetric complications, and multi-level medical concepts as nodes of the knowledge graph; The edges of the knowledge graph are formed through the relationship between the patient's basic information, the patient's health feature vector and the types of obstetric complications; Connect to external medical knowledge bases through external API interfaces to automatically update nodes and edges in the graph; Use the graph attention mechanism in the graph neural network to achieve adaptive adjustment of edge weights, and adjust the weight and direction of the edge by calculating the correlation between nodes; When each node updates information, it dynamically selects neighbor nodes for aggregation based on the relevance and importance between nodes; After multiple layers of transmission, the patient’s comprehensive health characteristics are obtained; The comprehensive health features are input into the fully connected layer to further extract nonlinear features; The predicted probability value of complications is generated through the output of the fully connected layer.

2. The method for full-course follow-up management of obstetrics and gynecology patients according to claim 1, characterized in that: When a patient visits a doctor for the first time, the patient information data is entered, which includes the following steps: Patient information data includes basic patient information, medical-related information, medical imaging data, genetic testing data, sensor data, and patient self-reported data; The patient's basic information includes the patient's name, age, telephone number, ID number, home address, occupation, and marital status; The medical-related information includes past medical history, family history of genetic diseases, past pregnancy and childbirth history, current gestational age, date of last menstrual period, and history of drug allergy; The medical imaging data includes an image of a fetus; The genetic testing data includes maternal venous blood sample analysis data; The sensor data includes the patient's heart rate, blood pressure, blood sugar, and fetal heart rate collected via a wearable device; The patient self-reported data includes the patient's eating habits, exercise status, mood, psychological state and sleep status.

3. The method for full-course follow-up management of obstetrics and gynecology patients as claimed in claim 2, characterized in that: The input patient information data is preprocessed and the preprocessed patient information data is integrated into a personal health record, which specifically includes the following steps: Standardize the data format of entered patient information data; Clean and denoise the patient information data after the data format is standardized; Extract fetal development features from medical imaging data using convolutional neural networks; Through feature selection methods, we can reduce low-correlation genes and extract risk coefficient features from gene testing data; Extract physiological index characteristics from sensor data through time series analysis method; The subjective scores in patients’ self-reports were used as subjective health score features; The extracted features are fused through a weighted multimodal data fusion algorithm to obtain a multimodal feature vector; Integrate the multimodal feature vector with the patient's basic information to generate the patient's personal health record.

4. The method for full-course follow-up management of obstetrics and gynecology patients as claimed in claim 3, characterized in that: Based on the integrated personal health records, a health assessment model for the patient is established, the patient's current health assessment score is output, and a follow-up plan is generated based on the patient's current health assessment score. The specific steps include: Logarithmic transformation of multimodal feature vectors in personal health records; Based on the degree of influence on the patient's health assessment, principal component analysis was used to assign weights to the multimodal feature vectors and the patient's basic information; Through the generalized linear thinking, and based on the personal health record, the patient's current health assessment score is output, and its expression is: ;in, Indicates the patient's current health assessment score. Represents the first Features, Indicates The weight coefficient of the feature, Indicates The weight coefficient of the feature in the denominator, Indicates the patient’s Basic information, Indicates The weight of basic information, represents the number of multimodal feature vectors, Indicates the number of basic information of patients; A follow-up plan is generated based on the patient's current health assessment score, wherein the follow-up plan includes the follow-up frequency and follow-up content.

5. The method for full-course follow-up management of obstetrics and gynecology patients according to claim 4, characterized in that: The follow-up plan is dynamically adjusted according to the patient's predicted complication value. The specific steps are as follows: A four-level threshold for patient complications was established based on the patient's basic information; The prediction result of the patient's complications is determined based on the interval of the patient's complication threshold value where the patient's complication prediction value is located, and the patient's health status in different time periods is dynamically evaluated through the patient's sensor data; Generate a smart health calendar for the patient, automatically adjusting the follow-up plan based on the patient's complication prediction results and the patient's sensor data.

6. A full-course follow-up management system for obstetrics and gynecology patients, based on the full-course follow-up management method for obstetrics and gynecology patients according to any one of claims 1 to 5, characterized in that: include: The information entry module collects and enters the patient's basic information, medical-related information, medical imaging data, genetic testing data, sensor data, and patient self-reported data; The data processing module pre-processes the entered patient data and integrates it into the patient's personal health record; The file management module stores and manages all the patient's health data and builds a comprehensive patient health file as the basis for subsequent health assessment and complication prediction; The health assessment module builds a health assessment model for patients based on the integrated personal health records and outputs the current health assessment score to assess the overall health status of the patient; Follow-up management module, which generates personalized follow-up plans based on the patient's health assessment scores and dynamically adjusts the frequency and content of follow-up according to complication prediction results; The complication prediction module establishes an obstetric complication prediction model based on the patient's personal health record, predicts the patient's complication risk, and outputs specific complication prediction values.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the full-course follow-up management method for obstetrics and gynecology patients described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the full-course follow-up management method for obstetrics and gynecology patients described in any one of claims 1 to 5 are implemented.

Images