GDM sugar control nutrition list recommendation method and system based on big data

Abstract

The invention relates to a big data-based GDM sugar-controlled nutrition list recommendation method and system, and the method comprises the steps: obtaining historical clinical data, priori knowledge and historical pregnancy index data of a plurality of pregnant women, and constructing a pregnant woman state portrait database and a pregnant woman nutrition list database; setting a mapping relation between the pregnant woman state portrait and the pregnant woman nutrition sheet based on the historical clinical data, establishing a first knowledge graph of the pregnant woman state portrait-pregnant woman nutrition sheet according to the mapping relation, and outputting a basic sugar control nutrition sheet of the pregnant woman through the first knowledge graph; acquiring actual pregnancy index data of the pregnant woman, analyzing the historical pregnancy index data and the actual pregnancy index data to obtain a pregnancy state change index, and outputting a target nutritional demand value of the pregnant woman based on the prediction neural network according to the pregnancy state change index; and updating the first knowledge graph according to the output of the prediction neural network to obtain a second knowledge graph, and outputting a target sugar control nutrition list of the pregnant woman through the second knowledge graph.

Description

Technical Field

[0001] This invention relates to the field of pregnant women's health management technology, specifically to a method and system for recommending glucose-controlled nutrition plans based on big data. Background Technology

[0002] Gestational Diabetes Mellitus (GDM) is a common complication during pregnancy, characterized by impaired glucose tolerance. Without proper prenatal intervention, GDM increases the risk of maternal pregnancy complications and fetal morbidity, and also increases the risk of developing other metabolic disorders in the mother (such as obesity or type 2 diabetes), resulting in long-term adverse effects on both the mother and fetus. Dietary management can be chosen as a prenatal care approach to intervene and treat gestational diabetes in mothers.

[0003] For example, the invention patent with publication number CN116631640A discloses a method and platform for generating personalized needs solutions for pregnant women, including a medical database preparation process, a pregnant woman data collection process, and a data matching process. Based on the comparison and matching of the pregnant woman's nutritional profile and the medical nutritional profile, the most suitable personalized needs solution for the pregnant woman is formulated and pushed to the pregnant woman.

[0004] For example, the invention patent with publication number CN118588230A discloses a decision-making system and method for individualized health plans for pregnant women with gestational hyperglycemia. It includes a pregnant woman information acquisition module, a diet plan calorie module, a monitoring and execution module, and a diet plan optimization module. It acquires pregnant woman information index data and dietary behavior data to generate a personalized diet management plan for meal preparation. It obtains dietary feedback at each dietary node of meal preparation and iteratively optimizes the diet management plan to generate a variety of standardized and individualized diet plans that meet specific needs, so that pregnant women receive scientific and reasonable dietary management during the delivery cycle.

[0005] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems:

[0006] In existing technologies, personalized prenatal care dietary management scenarios need to consider the individual differences in the conversion rate of nutrients absorbed by pregnant women in order to address the problem that dietary management plans may not be well matched to individuals with different differences. Simply combining standard indicator data of pregnant women's information, medical data and nutritional data to generate prenatal care dietary management plans results in insufficient effectiveness and adaptability of standard plans for the actual prenatal care of individuals. Summary of the Invention

[0007] The primary objective of this application is to address at least one of the aforementioned problems by providing a method and system for recommending GDM (Glucose Digestive Management) nutritional plans based on big data.

[0008] To achieve the various objectives of this application, the following technical solution is adopted:

[0009] One of the purposes of this application provides a big data-based method for recommending GDM (Glycemic Control and Nutrition) nutritional plans, comprising the following steps:

[0010] Historical clinical data, prior knowledge, and historical pregnancy indicator data from multiple pregnant women were acquired to construct a pregnant woman status profile database and a pregnant woman nutrition record database. A mapping relationship between the pregnant woman status profiles and the pregnant woman nutrition records was established based on the historical clinical data. A first knowledge graph of the pregnant woman status profile and the pregnant woman nutrition record was built based on this mapping relationship, and a basic blood sugar control nutrition record for the pregnant woman was output through the first knowledge graph. Actual pregnancy indicator data of the pregnant women was collected, and the historical and actual pregnancy indicator data were analyzed to obtain pregnancy status change indicators. Based on these indicators, a predictive neural network was used to output the pregnant woman's target nutritional requirements. The entities, attributes, and relationships in the first knowledge graph were updated based on the output of the predictive neural network to obtain a second knowledge graph. The target blood sugar control nutrition record for the pregnant woman was then output through the second knowledge graph.

[0011] Furthermore, for pregnant women with gestational diabetes, historical clinical data and historical gestational indicator data from prenatal testing are collected through a medical information system; prior knowledge is obtained through publicly available data sources, including publicly available research reports, prior knowledge bases, publicly available experimental data, publicly available nutritional reports, nutritional information databases, and publicly available experimental data related to gestational diabetes; the historical clinical data is first descriptive text data, which is used to reflect the degree of impact of gestational diabetes on the pregnant woman's weight status, fetal growth status, pregnant woman's blood sugar status, and pregnant woman's dietary choices;

[0012] The prior data is the second descriptive text data, and the first descriptive text includes information reflecting the impact of dietary choices by pregnant women with gestational diabetes on the pregnant woman's weight status, fetal growth status, and pregnant woman's blood sugar status; the historical pregnancy index data includes historical data on pregnant woman's weight, blood sugar, amniotic fluid volume, fetal growth status, and historical dietary choices.

[0013] Furthermore, based on the impact of gestational diabetes on pregnancy indicators in the first descriptive text, a correlation analysis was performed on historical pregnancy indicators and historical clinical data to obtain pregnant woman status characterization data that defines the pregnant woman status profile. A pregnant woman status profile database was constructed using this data. The feasibility of pregnant women's dietary choices was cross-validated based on the first and second descriptive texts to obtain feasible nutritional plan data for pregnant women's dietary choices. A pregnant woman nutrition plan database was constructed using this feasible nutritional plan data.

[0014] Furthermore, prior relationships among pregnant women's dietary choices, weight status, fetal growth status, and blood glucose status under gestational diabetes conditions are obtained from historical clinical data through entity recognition and relation extraction, and a prior relationship threshold is set. Entity recognition and relation extraction are performed on the prior knowledge to obtain a first mapping relationship between pregnant women's status representation data and feasible nutrition plans. Entity recognition and relation extraction are performed on historical pregnant women's indicator data to obtain a second mapping relationship between pregnant women's status representation data and feasible nutrition plans. Candidate pregnant women's status representation data that meet the prior relationship threshold in the pregnant women's status profile dataset are selected, and candidate feasible nutrition plans that meet the prior relationship threshold in the pregnant women's nutrition plan dataset are selected. A unified coding method is used to integrate the mapping relationships between the candidate pregnant women's status representation data and the candidate feasible nutrition plan data. Entity relationships in the pregnant women's status profile dataset and the pregnant women's nutrition plan dataset are determined based on the mapping relationships. A first knowledge graph is constructed based on the entities in the pregnant women's status profile dataset, the entities in the pregnant women's nutrition plan dataset, and the entity relationships between the pregnant women's status profile dataset and the pregnant women's nutrition plan dataset.

[0015] Furthermore, actual pregnancy indicator data of pregnant women are collected through a medical information system. Based on prior knowledge, a preset time length for analyzing changes in pregnancy status is determined. The actual pregnancy indicator data specifically includes pregnant woman's weight, blood glucose, amniotic fluid volume, fetal growth status, and actual dietary choices during the actual pregnancy. Based on the preset time length for analyzing changes in pregnancy status, the rates of change in pregnant woman's weight, blood glucose, maternal weight, and fetal weight are calculated from historical and actual pregnancy indicator data. A dataset of status change indicators related to dietary choices is generated based on a first knowledge graph. Based on the degree of influence of dietary choices on changes in pregnant woman's weight, maternal weight, and fetal growth status, pregnancy status change assessment indicators are determined. These indicators include indicators for changes in pregnant woman's weight, maternal weight, and fetal weight. Based on the degree of influence of amniotic fluid volume on changes in fetal growth status and the degree of influence of dietary choices on changes in pregnant woman's blood glucose, pregnancy status change impact indicators are determined. These impact indicators are set based on amniotic fluid volume change indicators, pregnant woman's blood glucose change indicators, pregnant woman's weight change indicators, and fetal weight change indicators.

[0016] Furthermore, the pregnancy status change assessment indicators and pregnancy status change impact indicators are input into a preset predictive neural network to output the pregnant woman's target nutritional requirements. This predictive neural network is trained using a training dataset that includes multiple historical clinical data, historical pregnancy indicator data, and corresponding actual nutritional requirement annotations. The predictive neural network is a composite network, specifically including a requirement screening network and a requirement prediction network. The requirement screening network includes a long short-term network model, a classifier model, and a discriminant screening model. The long short-term network model is used to predict the first nutritional requirement for each change indicator in the input pregnancy status change assessment indicators. The classifier model is used to predict first candidate nutritional requirements based on multiple first nutritional requirements and compare the predicted first candidate nutritional requirements with the actual nutritional requirements. Similarity is used to classify the pregnant woman's status profile for the first candidate nutritional requirement value. The first nutritional requirement includes the maternal nutritional conversion rate and the fetal nutritional conversion rate. The first candidate nutritional requirement includes the total nutritional conversion rate of the pregnant woman's intake. The discriminant screening model is used to update the amniotic fluid volume screening threshold and blood glucose change screening threshold based on the input pregnancy status change influence indicators. The first candidate nutritional requirement value of the pregnant woman's status profile is then screened to obtain the nutritional requirement value that simultaneously meets the amniotic fluid volume screening threshold and the blood glucose change screening threshold as the second nutritional requirement input to the requirement prediction network. The requirement prediction network is used to predict the pregnant woman's status profile classification input to the classifier model and the second nutritional requirement output by the discriminant screening model to obtain the pregnant woman's target nutritional requirement value.

[0017] Furthermore, based on the pregnant woman's status profile output by the classifier model, entities for the pregnant woman's status profile classified according to nutrient conversion rate are added to the first knowledge graph, and the entity attributes of the pregnant woman's status profile are updated. Based on the target nutrient requirement value of the pregnant woman's status profile under reclassification, the dietary selection combinations of feasible nutrition plans in the pregnant woman's nutrition plan dataset are updated, as are the entity attributes of the pregnant woman's nutrition plan and the entity relationships between the entities in the pregnant woman's status profile and the entities in the pregnant woman's nutrition plan. The first knowledge graph is dynamically updated to the second knowledge graph. Based on the actual pregnancy index data of the target pregnant woman, the pregnant woman's status profile is determined. Based on the entity relationships between the entities in the pregnant woman's status profile and the entities in the pregnant woman's nutrition plan in the second knowledge graph, the target blood sugar control nutrition plan required by the target pregnant woman is output.

[0018] On the other hand, a big data-based GDM (Glycemic Control and Nutritional Recommendation) system for blood sugar control, provided to meet one of the purposes of this application, includes...

[0019] On another front, a big data-based GDM (Glucose Digestion Management) nutrition recommendation device, provided to meet one of the purposes of this application, includes a central processing unit and a memory. The central processing unit is used to call and run a computer program stored in the memory to execute the steps of the big data-based GDM nutrition recommendation method described in this application.

[0020] In another aspect, a computer-readable storage medium is provided to suit one of the purposes of this application, the computer-readable storage medium storing computer-executable instructions for causing a computer to perform the big data-based GDM blood sugar control nutrition recommendation method as disclosed in any of the first aspects of this invention.

[0021] The technical solution of this application has many advantages, including but not limited to the following aspects:

[0022] This application first constructs a database of pregnant women's status profiles and nutritional plans by using historical clinical data, prior knowledge, and pregnancy indicator data. By matching the profiles with nutritional plans that have high accuracy and suitability, it provides data support for the accurate recommendation of nutritional plans for blood sugar control in pregnant women with gestational diabetes. Furthermore, it helps pregnant women reduce the risk of gestational diabetes in both the mother and fetus through prenatal care using dietary management methods.

[0023] Secondly, by constructing a knowledge graph through the mapping relationship between pregnant women's status profiles and the pregnant women's nutrition plan database, and using the prior relation threshold obtained from historical clinical data, clinical data that matches the actual clinical situation of pregnant women can be filtered out. This can further improve the accuracy of the knowledge graph in recommending blood sugar control nutrition plans for pregnant women. The mapping relationship more realistically reflects the pregnancy status of pregnant women at different historical periods, and determines feasible nutrition plan data under different pregnant women's status profiles in the prior knowledge. This can provide a more scientific basis for pregnant women to make feasible dietary choices, making the blood sugar control nutrition plans recommended based on the pregnant women's status profiles more accurate and appropriate.

[0024] Next, a predictive neural network is used to predict nutritional needs represented by different indicators. Then, based on the pregnant woman's status profile under the individual differences of multiple nutritional needs, the corresponding candidate nutritional needs prediction results are output. At the same time, the influencing factors of amniotic fluid volume and blood glucose changes are introduced to further screen the prediction results of candidate nutritional needs. The total nutritional needs required by each pregnant woman are output. Then, the individual nutritional absorption differences are eliminated by profile classification, and the target nutritional needs value is output for the pregnant woman to accurately predict the actual nutritional needs of the pregnant woman. This helps to improve the efficiency of the accuracy of nutritional needs prediction based on the prediction results of multiple neural networks and joint judgment, and provides accurate data reference for recommending blood sugar control nutrition plans, thereby improving the matching accuracy and suitability of blood sugar control nutrition plans for different pregnant women.

[0025] Finally, based on the first knowledge graph, the selection of foods in the adjusted target blood sugar control nutrition plan can be quickly determined by the target nutritional requirement value, without the need for complex updates to the knowledge graph. Moreover, the dynamically updated knowledge graph can continuously provide new profile entities, further refining the granularity of the pregnant woman profile classification based on individual differences, thereby improving the accuracy and suitability of the recommended nutrition plan for pregnant women with individual differences. It has high application value in the field of pregnant woman health management technology. Attached Figure Description

[0026] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0027] Figure 1 This is a schematic diagram of the user interface of a mobile application for a blood sugar control recommendation form, which is an example of this application.

[0028] Figure 2 A schematic diagram of the user interface for outputting a nutrition report in a mobile application for a blood sugar control recommendation form, which is an example of this application.

[0029] Figure 3 A flowchart illustrating one embodiment of the GDM (Glucose Digestion Management) nutrition recommendation method based on big data according to this application;

[0030] Figure 4This is a schematic diagram of the GDM blood sugar control nutrition recommendation system based on big data used in this application. Detailed Implementation

[0031] The technical solution of this application is applicable to the field of maternal health management technology, and is particularly suitable for the scenario of dietary management for pregnant women with gestational diabetes. In this context, the technical solution of this application can be applied to a typical mobile application, such as... Figure 1 and Figure 2 As shown, users can input the pregnant woman's current health status into the user interface, and the system will provide personalized nutritional prescription recommendations based on the pregnant woman's actual condition.

[0032] In existing technologies, personalized prenatal care dietary management scenarios need to consider the impact of the individual pregnant woman's nutrient absorption and conversion rate on changes in the pregnant woman's condition during pregnancy. This is to address the potential problem of insufficient effectiveness of prenatal care dietary management plans. Simply combining standard indicator data from the pregnant woman's information, medical data, and nutritional data to generate a prenatal care dietary management plan results in insufficient effectiveness and adaptability of the standard plan to the actual prenatal care of individuals.

[0033] To address the shortcomings in the effectiveness and suitability of the aforementioned prenatal care dietary management programs, the embodiments of this application provide the following general approach:

[0034] To better understand the above technical solutions, a detailed explanation will be provided below with reference to the accompanying drawings and specific implementation methods.

[0035] See Figure 3 This application discloses a method for recommending a glucose control nutrition plan based on big data for gestational diabetes mellitus (GDM). In its typical prenatal care embodiment for pregnant women with gestational diabetes mellitus, it includes the following steps:

[0036] Step 1100: Obtain historical clinical data, prior knowledge, and historical pregnancy indicator data of multiple pregnant women to construct a pregnant woman status profile database and a pregnant woman nutrition record database, respectively.

[0037] It should be noted that building a pregnant woman profile database allows for the basic classification of pregnant women with gestational diabetes who require blood sugar control nutrition plans. Compared to pregnant women without the condition, those with gestational diabetes require consideration of more indicators in prenatal care, such as the impact of blood sugar fluctuations on maternal health and the fetal health and growth. Therefore, building a pregnant woman profile database for basic classification is essential. It can identify profiles of pregnant women with gestational diabetes under different basic categories, and by matching these profiles with accurate and suitable maternal nutrition plans, precise recommendations for blood sugar control nutrition prescriptions can be made. Ultimately, through dietary management in prenatal care, it helps pregnant women reduce the risk of disease for both mother and fetus during gestational diabetes.

[0038] Furthermore, constructing a pregnant woman's nutrition record database can provide decision-making data for recommending blood sugar control nutrition prescriptions. For example, for pregnant women with gestational diabetes, it is necessary to consider not only the impact of carbohydrate intake in a single meal on blood sugar changes, but also whether the total energy intake in that meal meets the energy requirements of the current pregnancy. Therefore, it is necessary to select a blood sugar control nutrition prescription that meets the requirements of low carbohydrate intake and high energy intake. It is also necessary to consider factors including, but not limited to, protein intake, fat intake, and the pregnant woman's dietary preferences. Therefore, a pregnant woman's nutrition record database is essential for generating blood sugar control nutrition prescriptions. After determining the pregnant woman's profile, the recommended energy, protein, carbohydrate, and fat requirements for blood sugar control nutrition prescriptions can be determined. Then, based on the intake requirements, the available foods can be arranged and combined to determine multiple optional blood sugar control nutrition prescriptions. Finally, the target blood sugar control nutrition prescription can be selected based on the pregnant woman's dietary preferences. In this way, prenatal care for pregnant women with gestational diabetes can be achieved through dietary management, helping pregnant women reduce the risk of disease for both the mother and the fetus during gestational diabetes.

[0039] It is understandable that there are differences in nutrient absorption among individual pregnant women. For example, both pregnant women A and B have profiles labeled "advanced maternal age," "normal weight during pregnancy," and "gestational diabetes." Based on these profiles, both women received the same blood sugar control recommendations. However, because their nutrient intake conversion rates differed during pregnancy, their weight change rates during the same gestational period also differed. It is worth noting that fetal growth status also needs to be considered. This can be measured through weight changes during fetal development, and fetal weight can be estimated using fetal imaging. Individually, some pregnant women may have total weight gain within the normal range (e.g., a normal weight gain during pregnancy is between 11kg and 16kg), but their total weight gain or rate of weight change is low. In cases where the total weight gain or rate of weight change in the fetus is high, the actual total nutritional intake of the pregnant woman's diet reflects the overall weight of the pregnant woman during pregnancy. Further analysis of the changes in the mother's own weight gain and the changes in the fetal growth and weight is needed. A comprehensive assessment of the health of both the mother and fetus during pregnancy is required, especially for pregnant women with gestational diabetes. At different gestational weeks, there are different types of gestational diabetes, such as rapid fetal growth, excessively rapid maternal weight gain, slow fetal development, low fetal weight, and a preference for certain types of maternal weight gain. More precise nutritional prescriptions for blood sugar control are needed to reduce the health risks to both the mother and fetus caused by fluctuations in blood sugar, maternal weight, and fetal weight. This provides a reasonable and precise support plan for the product-assisted delivery process.

[0040] In practice, for pregnant women with gestational diabetes, historical clinical data and historical gestational indicators from prenatal testing are collected through a medical information system. Prior knowledge is obtained through publicly available data sources. Among these, historical clinical data can effectively reflect the impact of different dietary choices during pregnancy on blood sugar, maternal weight, and fetal weight, facilitating data-supported decision-making for blood sugar control prescriptions. Pregnant women typically undergo periodic or phased prenatal testing to assess the health of both the fetus and the mother, which can serve as reference data for categorizing pregnant women into different groups.

[0041] Optionally, publicly available data sources include publicly available research reports, prior knowledge bases, publicly available experimental data, publicly available nutritional plans research reports, nutritional plan information databases, and publicly available nutritional plan experimental data related to gestational diabetes. The content of these publicly available data sources may include domestic and international medical guidelines and literature related to gestational diabetes, empirically validated data on gestational diabetes, medical guidelines and literature related to dietary choices and nutritional management during pregnancy, experimentally validated dietary choices for pregnant women with gestational diabetes, the nutritional and energy requirements for pregnant women with gestational diabetes at different stages of pregnancy, and the nutritional and energy requirements for pregnant women with gestational diabetes at different weight levels. The relevant prior knowledge obtained from publicly available data sources can be deduplicated and noise-reduced through screening and evaluation.

[0042] Optionally, historical clinical data is the first descriptive text data, which reflects the impact of gestational diabetes on the pregnant woman's weight status, fetal growth status, maternal blood glucose status, and the pregnant woman's dietary choices. Natural language processing technology can be used to perform semantic recognition and relation extraction on the text in the historical clinical data that reflects the impact of gestational diabetes on the pregnant woman's weight status, fetal growth status, maternal blood glucose status, and the pregnant woman's dietary choices. When collecting and organizing historical clinical data, the text can be converted into a computer-processable form, identifying important descriptive text information to facilitate the storage and management of historical clinical data. It is understood that the first descriptive text data describes the actual clinical symptoms of pregnant women as reflected in the dietary management of blood glucose control nutrition prescriptions during historical midwifery processes. It can serve as a basis for decision-making regarding the selection of different blood glucose control nutrition prescriptions or for cross-validation with clinical symptoms represented by other data sources.

[0043] Optionally, the prior data is a second descriptive text data. The first descriptive text includes information reflecting the impact of dietary choices by pregnant women with gestational diabetes on the pregnant woman's weight, fetal growth, and blood glucose levels. The prior data can be preprocessed by deduplication and noise reduction, and then semantically recognized and relationally extracted using natural language processing (NLP) technology. This yields the second descriptive text. It is understood that the prior data, being report data, may contain errors or deviate from reality. Therefore, when assessing the impact of dietary choices on the pregnant woman's weight, fetal growth, and blood glucose levels, the first and second descriptive text data can be cross-validated to determine the authenticity of the impact of dietary choices on these factors, providing data support for subsequent recommendations of blood glucose control nutrition plans.

[0044] Optionally, the historical pregnancy data includes maternal weight data, maternal blood glucose data, maternal amniotic fluid volume data, fetal growth status data, and historical dietary choices data within a historical period.

[0045] In practice, based on the impact of gestational diabetes on pregnancy indicators in the first descriptive text, a correlation analysis of historical pregnancy indicator data and historical clinical data is performed to obtain pregnancy status characterization data that defines the pregnancy status profile. A pregnancy status profile database is then constructed using this data. Historical pregnancy indicator data can be obtained from tests conducted on pregnant women in different regions, hospitals, or under the care of different doctors. Pregnant woman weight data can be used to analyze samples of pregnant women with different weights and different rates of weight change. Blood glucose data can be used to analyze samples of pregnant women with different blood glucose changes. Amniotic fluid volume data can be used to analyze amniotic fluid levels at different stages of pregnancy. The impact of pregnancy on fetal growth and development, as well as amniotic fluid volume data, can be used to denoise the analysis of maternal weight changes. Fetal growth status data is mainly obtained through fetal imaging. In addition to estimating weight, the influence of placental and umbilical cord positions on fetal growth can also be considered. Historical dietary choice data can be used to analyze whether the pregnant woman's dietary choices are reasonable and whether the match between the pregnant woman's profile and the nutrition plan is accurate. By using historical pregnancy index data, a comprehensive characterization of the pregnant woman's status during pregnancy can be achieved, thereby obtaining an accurate profile of the pregnant woman's status in different situations. This leads to the construction of a pregnant woman status profile database, providing data support for dietary management of prenatal care for pregnant women.

[0046] Optionally, when constructing the pregnant woman status profile database, basic prenatal information, pregnancy history, and medical history can also be considered. Basic information includes, but is not limited to, the pregnant woman's name, age, height, weight, blood type, and number of fetuses. Medical history includes, but is not limited to, pre-gestational diabetes, history of gestational diabetes, polycystic ovary syndrome, hypertension, and thyroid-related diseases. This information can be obtained through user input on the user's end. Figure 1 As shown, in application scenarios, pregnant women can fill in information on a mobile app during their first medical visit to generate an electronic medical record for the pregnant woman.

[0047] In practice, the feasibility of pregnant women's dietary choices is cross-validated based on the first and second descriptive texts to obtain feasible nutritional data for pregnant women's dietary choices. A nutritional data database for pregnant women is then constructed based on the feasible nutritional data.

[0048] It is evident that cross-validation of two descriptive texts can determine the impact of real dietary choices on pregnant women, thereby obtaining accurate nutritional lists describing the needs of different pregnant women. This leads to the construction of a pregnant women's nutritional list database, providing data support for the combination of dietary choices in prenatal care for pregnant women.

[0049] Step 2100: Based on historical clinical data, establish a mapping relationship between the pregnant woman's status profile and the pregnant woman's nutrition plan. Based on the mapping relationship, establish a first knowledge graph of the pregnant woman's status profile and the pregnant woman's nutrition plan. Output the basic blood sugar control nutrition plan for the pregnant woman through the first knowledge graph. Among them, the historical clinical data is the real data reflecting the pregnant woman's dietary choices, which can be used to build the association between the pregnant woman's status profile and the pregnant woman's nutrition plan. Thus, the structured pregnant woman's nutrition plan data and the pregnant woman's status profile are constructed in the form of a knowledge graph, which can be used to subsequently formulate personalized pregnant woman's nutrition plans based on the pregnant woman's profile.

[0050] In practice, prior relationships among pregnant women's dietary choices, weight status, fetal growth, and blood glucose levels under gestational diabetes conditions are obtained from historical clinical data through entity recognition and relation extraction. Prior relationship thresholds are set. For example, dietary choices have a certain influence on changes in pregnant women's weight, fetal growth, and blood glucose levels. This influence can be obtained through semantic recognition based on the content recorded in historical clinical data within a certain period of time. Entity recognition and relation extraction then use this influence as a prior relationship. For each prior relationship, a relation threshold is set based on human experience. Meeting the relation threshold proves that there is definitely an influence between the entity identified in the historical clinical data and other entities. For example, large changes in pregnant women's weight affect fetal growth and development.

[0051] As can be seen, the prior relationship threshold obtained from historical clinical data in the above embodiments can be used to screen clinical data that matches the actual clinical situation of pregnant women, which can further improve the accuracy of the knowledge graph in recommending blood sugar control nutrition plans for pregnant women.

[0052] Next, entity recognition and relation extraction are performed on prior knowledge to obtain the first mapping relationship between the pregnant woman's status representation data and the feasible nutrition plan data. Entity recognition and relation extraction are then performed on historical pregnant woman indicator data to obtain the second mapping relationship between the pregnant woman's status representation data and the feasible nutrition plan data. Specifically, the pregnant woman's status representation, obtained from the dietary choices extracted from prior knowledge, serves as the mapping relationship between the pregnant woman's profile and the available nutrition plans. While prior knowledge has high accuracy, it is necessary to further consider the matching of the pregnant woman's status representation with the feasible nutrition plan under real clinical conditions for different pregnant woman samples. Therefore, historical pregnant woman indicator data is extracted from the historical pregnant woman indicator data. The mapping relationship between actual dietary choices and corresponding historical pregnant woman indicator data is used. The first mapping relationship is used to determine the pregnant woman status representation data in the historical pregnant woman indicator data, thereby obtaining data support for constructing a pregnant woman status profile. This can more realistically reflect the pregnancy status of pregnant women in different historical periods and further refine the granularity of the pregnant woman status profile, effectively distinguishing pregnant woman profiles with individual differences. The second mapping relationship is used to determine the feasible nutrition plan data under different pregnant woman status profiles based on prior knowledge. This can provide a more scientific basis for pregnant women to make feasible dietary choices, making the blood sugar control nutrition plan recommended based on the pregnant woman status profile more accurate and appropriate.

[0053] Finally, candidate pregnant woman status representation data that satisfy the prior relation threshold in the first mapping relationship are selected from the pregnant woman status profile dataset, and candidate feasible nutrition plan data that satisfy the prior relation threshold in the pregnant woman nutrition plan dataset are selected from the second mapping relationship. The mapping relationship between the candidate pregnant woman status representation data and the candidate feasible nutrition plan data is integrated using unified coding. The entity relationship in the pregnant woman status profile dataset and the pregnant woman nutrition plan dataset is determined according to the mapping relationship. The first knowledge graph is constructed based on the entities in the pregnant woman status profile dataset, the entities in the pregnant woman nutrition plan dataset, and the entity relationship between the pregnant woman status profile dataset and the pregnant woman nutrition plan dataset.

[0054] As can be seen, the above embodiments provide a method for semantic recognition, entity extraction, relationship recognition, and multiple verifications through historical data and prior knowledge, constructing a first knowledge graph to accurately express the relationship between pregnant women with gestational diabetes and the available maternal nutrition plans. This achieves improved accuracy and suitability of nutrition plan recommendations based on scientific theoretical knowledge and real clinical manifestations, provides data reference for confirming the pregnant woman profile and confirming the nutrition plan, and improves the recognition rate of individual differences of pregnant women in recommended nutrition plans.

[0055] Step 3100: Collect actual pregnancy indicator data of pregnant women, analyze historical pregnancy indicator data and actual pregnancy indicator data to obtain pregnancy status change indicators, and output the target nutritional requirements of pregnant women based on the pregnancy status change indicators and a predictive neural network.

[0056] It's important to clarify that both actual pregnancy indicator data and historical pregnancy indicator data are of the same data type. The only difference lies in the timing: historical pregnancy indicator data comprises optional indicators representing the pregnancy status collected from pregnant women before and after their previous pregnancies, while actual pregnancy indicator data is collected from pregnant women currently in their pregnancy to further differentiate individual differences. The only essential difference is the time dimension. Actual pregnancy indicator data allows for further refinement of individual differences among pregnant women. It's understandable that without refining individual differences, the actual conversion rate of nutrients will differ for different pregnant women even with the same total nutrient intake. This difference manifests not only in variations in overall maternal weight but also in variations in maternal and fetal weight changes despite the same nutrient intake. Therefore, for different individuals, the actual nutritional requirements still differ from the total nutritional amount in the basic blood sugar control nutrition plan provided by the first knowledge graph. This difference needs to be eliminated to output more accurate and appropriate target nutritional requirements, providing the data support needed for recommending nutritional prescriptions.

[0057] Optionally, actual pregnancy indicator data of pregnant women can be collected through a medical information system. Based on prior knowledge, a preset time length for analyzing changes in pregnancy status can be determined. The actual pregnancy indicator data specifically includes the pregnant woman's weight, blood glucose, amniotic fluid volume, fetal growth status, and actual dietary choices during the actual pregnancy. As shown in Table 1, the range of weight change varies for different BMI categories and stages of pregnancy. For example, based on prior knowledge, a pregnant woman with a known low BMI category has a total weight gain range of 11kg-16kg, with a weight gain range of 0-2kg in early pregnancy and an average weight gain of 0.46kg in the mid-to-late stages of pregnancy, ranging from 0.37kg to 0.56kg. Therefore, the preset time length for analyzing changes in pregnancy status needs to be determined based on the weight gain range during different stages of pregnancy. For example, the changes in pregnancy status within 5 days of early pregnancy can be specifically set according to the fine-grained requirements for individual difference analysis.

[0058] Table 1 - Range of Weight Gain During Pregnancy

[0059]

[0060] In practical implementation, based on a preset time length for analyzing changes in pregnancy status, the rates of change in maternal weight, maternal blood glucose, maternal body weight, and fetal weight are calculated from historical and actual pregnancy indicator data. A dataset of status change indicators related to dietary choices is generated based on the first knowledge graph. The time endpoints of the change rates can be determined using the preset time length. The difference between the pregnancy indicator data values ​​corresponding to two time endpoints within the preset time length is obtained, and then divided by the preset time length to obtain the rates of change in maternal weight, maternal blood glucose, maternal body weight, and fetal weight. It can be understood that maternal weight can be calculated by subtracting the fetal weight from the total weight of the pregnant woman at the time endpoint. Regarding the acquisition of fetal weight and amniotic fluid volume, it is important to explain that in early pregnancy, amniotic fluid is mainly formed by maternal plasma permeating into the amniotic cavity through the capillaries of the placenta and uterine wall. This stage provides amniotic fluid for the mother. Therefore, the impact of amniotic fluid volume on maternal weight is not considered in the analysis of changes in pregnancy status before 12 weeks of gestation. However, after the mid-pregnancy stage, fetal urine, pulmonary secretions, and skin exudates are the main sources of amniotic fluid. Therefore, when setting the calculation of the rate of change after the mid-pregnancy stage, the interference of amniotic fluid volume on the mother's own weight needs to be considered. It is worth noting that fetal weight can be estimated by measuring fetal abdominal circumference, femur length, etc. through fetal imaging. The difference between the estimated value and the actual value is negligible and does not affect the accuracy of the rate of change.

[0061] Furthermore, after obtaining the data on various rates of change within a preset time period, the dietary choices of pregnant women within that time period can be confirmed based on the time markers corresponding to the preset time. Then, the data on various rates of change can be associated with the basic blood sugar control nutrition chart output by the first knowledge graph corresponding to the dietary choices.

[0062] Furthermore, based on the degree of influence of dietary choices on changes in pregnant women's weight, changes in maternal weight, and changes in fetal growth status, indicators for assessing changes in pregnancy status were determined.

[0063] Optionally, the assessment indicators for changes in pregnancy status include indicators of changes in pregnant women's weight, changes in maternal weight, and changes in fetal weight. The same dietary choice may result in different outcomes for different pregnant women. In order to further study the actual nutritional needs of pregnant women, a comprehensive assessment can be conducted by changing the total weight of the pregnant woman, changing the maternal weight, and changing the fetal weight, so that the above weight changes are within a reasonable range while reconfirming the actual nutritional value obtained by the pregnant woman.

[0064] Furthermore, based on the degree of influence of amniotic fluid volume on changes in fetal growth status and the degree of influence of dietary choices on changes in maternal blood sugar, indicators affecting changes in pregnancy status were determined.

[0065] Optionally, the indicators affecting changes in pregnancy status are set based on changes in amniotic fluid volume, maternal blood glucose levels, maternal weight, and fetal weight. Among these, besides assessing the actual nutritional intake of the pregnant woman through weight change indicators, changes in amniotic fluid volume and blood glucose levels indirectly affect the weight changes of both the pregnant woman and the fetus. Therefore, changes in amniotic fluid volume and blood glucose levels need to be considered in the analysis and can be added as influencing factors to ensure that the analysis results are not affected by changes in amniotic fluid volume and blood glucose levels, thereby improving the accuracy of the analysis and prediction results.

[0066] In practice, the pregnancy status change assessment index and the pregnancy status change impact index are input into a preset prediction neural network to output the pregnant woman's target nutritional needs value. The prediction neural network is trained through a training dataset that includes multiple historical clinical data, historical pregnancy index data and corresponding actual nutritional needs labeled.

[0067] Furthermore, the predictive neural network is a composite network, specifically including a demand screening network and a demand prediction network. The demand screening network includes a long short-term network model, a classifier model, and a discriminant screening model.

[0068] Furthermore, the pregnancy status change assessment indicators and pregnancy status change impact indicators are input into a preset predictive neural network, which outputs the pregnant woman's target nutritional requirements. This predictive neural network is trained using a training dataset that includes multiple historical clinical data, historical pregnancy indicator data, and corresponding actual nutritional requirement annotations. The predictive neural network is a composite network, specifically including a requirement screening network and a requirement prediction network. The requirement screening network includes a long short-term network model, a classifier model, and a discriminant screening model. The long short-term network model is used to predict the first nutritional requirement for each change indicator in the input pregnancy status change assessment indicators, and the classifier model is used to predict the first nutritional requirement based on multiple first nutritional requirements. The first candidate nutritional requirement is determined, and the similarity between the predicted first candidate nutritional requirement and the actual nutritional requirement is compared. Based on the similarity, the pregnant woman's status profile of the first candidate nutritional requirement value is classified. The first nutritional requirement includes the nutritional conversion rate of the pregnant woman's body and the nutritional conversion rate of the fetus. The first candidate nutritional requirement includes the total conversion rate of the pregnant woman's nutrient intake. The discrimination and screening model is used to update the amniotic fluid volume screening threshold and the blood glucose change screening threshold according to the input pregnancy status change influence index. The first candidate nutritional requirement value of the pregnant woman status profile is screened, and the nutritional requirement value that meets both the amniotic fluid volume screening threshold and the blood glucose change screening threshold is obtained as the second nutritional requirement input to the requirement prediction network.

[0069] Among them, the first nutritional requirement is the nutrient conversion rate corresponding to the nutritional requirements for changes in maternal and fetal weight. It can be used to assist in calculating the actual nutritional requirements for changes in maternal or fetal weight based on the pregnant woman's actual dietary choices. Therefore, the total conversion rate of nutrients ingested by the pregnant woman can be predicted by the nutritional requirements for changes in maternal and fetal weight among multiple first nutritional requirements. The total nutritional requirements can be calculated based on the total amount of nutrients corresponding to the actual dietary choices. The amniotic fluid volume screening threshold and blood glucose change screening threshold are used to measure the impact of changes in amniotic fluid volume caused by the current dietary choices on the fetus, and the impact of changes in maternal blood glucose on the fetus and the pregnant woman. If the threshold is not met, the current dietary choices have a certain impact on the pregnant woman and the fetus. If the selection conditions are not met, the impact of changes in amniotic fluid volume and blood glucose on the prediction results can be reduced, and the accuracy of the prediction results can be improved.

[0070] Furthermore, the demand prediction network is used to classify the pregnant woman's status profile input to the classifier model and predict the second nutritional requirement output by the discrimination and screening model to obtain the pregnant woman's target nutritional requirement value.

[0071] As can be seen, the above embodiments define the network type and architecture of the predictive neural network, enabling it to predict nutritional needs represented by different indicators. Then, based on the individual differences in pregnant women's status profiles, it classifies and outputs corresponding candidate nutritional need prediction results. Finally, it introduces the influencing factors of amniotic fluid volume and blood glucose changes to further filter the prediction results of candidate nutritional needs, outputting the total nutritional needs required by each pregnant woman. It then outputs the target nutritional need value for pregnant women under the individual difference classification, accurately predicting the actual nutritional needs of pregnant women. This helps to improve the efficiency of predicting nutritional needs based on the prediction results and joint judgment of multiple neural networks, providing accurate data references for recommending blood sugar control nutrition plans and improving the matching accuracy and suitability of blood sugar control nutrition plans for different pregnant women.

[0072] Step 4100: Update the entities, attributes, and relationships in the first knowledge graph according to the output of the predictive neural network to obtain the second knowledge graph, and output the pregnant woman's target blood sugar control nutrition plan through the second knowledge graph.

[0073] In practice, based on the pregnant woman status profile output by the classifier model, the pregnant woman status profile entity based on the nutrient conversion rate classification in the first knowledge graph is added, and the entity attributes of the pregnant woman status profile are updated.

[0074] Next, based on the target nutritional requirements of the pregnant woman status profile under reclassification, the dietary selection combinations of feasible nutrition plans in the pregnant woman nutrition plan dataset are updated, as well as the entity attributes of the pregnant woman nutrition plan and the entity relationship between the entities in the pregnant woman status profile and the entities in the pregnant woman nutrition plan are updated, and the first knowledge graph is dynamically updated to the second knowledge graph.

[0075] As an example, a newly added pregnant woman status profile entity can be represented as "carbohydrate utilization rate 90%-91%, protein utilization rate 73%-74%, fat utilization rate 94%-95%, energy conversion rate 45%-46%". This allows for the selection and combination of the basic blood sugar control nutrition plan in the first knowledge graph to create a target blood sugar control nutrition plan that meets the actual needs of pregnant women. For the newly added pregnant woman status profile entity, the description of the relationship between the previously determined feasible nutrition plan and the profile entity needs to be modified. Since the description of the user profile in this application already includes the range of adjustable values, adjustments can be made directly based on the original basic blood sugar control nutrition plan. For example, if the original basic blood sugar control nutrition plan has a fixed total amount of nutrients, then the total amount of nutrients is adjusted first based on nutrient conversion. Then, the types of food to be consumed after the adjusted total amount of nutrients are determined. Since the nutrient content of different foods has been constructed into entity relationships in the knowledge graph through prior knowledge, the selection of foods to be consumed in the target blood sugar control nutrition plan after the adjusted total amount can be quickly determined without complex updates to the knowledge graph. Moreover, the dynamically updated knowledge graph can continuously provide new profile entities, further refining the granularity of the pregnant woman profile classification, thereby improving the accuracy and suitability of the recommended nutrition plan for pregnant women with individual differences.

[0076] Finally, based on the actual pregnancy indicators of the target pregnant woman, a pregnancy status profile is determined. Based on the entity relationship between the pregnant woman status profile entity and the pregnant woman nutrition plan entity in the second knowledge graph, the target blood sugar control nutrition plan required by the target pregnant woman is output.

[0077] The unique technical advantage of this application lies in the following: First, by constructing a database of pregnant women's status profiles and nutritional plans based on historical clinical data, prior knowledge, and pregnancy indicators, the profiles are matched with nutritional plans of pregnant women with high accuracy and suitability. This provides data support for the accurate recommendation of nutritional plans for blood sugar control in pregnant women with gestational diabetes. Furthermore, through prenatal care using dietary management methods, it helps pregnant women reduce the risk of gestational diabetes in both the mother and fetus.

[0078] Secondly, by constructing a knowledge graph through the mapping relationship between pregnant women's status profiles and the pregnant women's nutrition plan database, and using the prior relation threshold obtained from historical clinical data, clinical data that matches the actual clinical situation of pregnant women can be filtered out. This can further improve the accuracy of the knowledge graph in recommending blood sugar control nutrition plans for pregnant women. The mapping relationship more realistically reflects the pregnancy status of pregnant women at different historical periods, and determines feasible nutrition plan data under different pregnant women's status profiles in the prior knowledge. This can provide a more scientific basis for pregnant women to make feasible dietary choices, making the blood sugar control nutrition plans recommended based on the pregnant women's status profiles more accurate and appropriate.

[0079] Next, a predictive neural network is used to predict nutritional needs represented by different indicators. Then, based on the pregnant woman's status profile under the individual differences of multiple nutritional needs, the corresponding candidate nutritional needs prediction results are output. At the same time, the influencing factors of amniotic fluid volume and blood glucose changes are introduced to further screen the prediction results of candidate nutritional needs. The total nutritional needs required by each pregnant woman are output. Then, the individual nutritional absorption differences are eliminated by profile classification, and the target nutritional needs value is output for the pregnant woman to accurately predict the actual nutritional needs of the pregnant woman. This helps to improve the efficiency of the accuracy of nutritional needs prediction based on the prediction results of multiple neural networks and joint judgment, and provides accurate data reference for recommending blood sugar control nutrition plans, thereby improving the matching accuracy and suitability of blood sugar control nutrition plans for different pregnant women.

[0080] Finally, based on the first knowledge graph, the selection of foods in the adjusted target blood sugar control nutrition plan can be quickly determined using the target nutritional requirement value, without the need for complex updates to the knowledge graph. Furthermore, the dynamically updated knowledge graph continuously provides new profile entities, further refining the granularity of the pregnant woman profile classification based on individual differences, thereby improving the accuracy and suitability of the recommended nutrition plans for pregnant women with individual differences. Figure 1 and Figure 2 As shown, it has high application value in the field of maternal health management technology.

[0081] Please see Figure 4According to one aspect of this application, a big data-based GDM (Glucose Demand Management) nutrition plan recommendation system is provided. The system includes: a data acquisition module for acquiring historical clinical data, prior knowledge, and historical pregnancy indicator data of multiple pregnant women, respectively constructing a pregnant woman status profile database and a pregnant woman nutrition plan database; a first recommendation module for setting a mapping relationship between the pregnant woman status profile and the pregnant woman nutrition plan based on historical clinical data, establishing a first knowledge graph of the pregnant woman status profile and the pregnant woman nutrition plan based on the mapping relationship, and outputting a basic glucose control nutrition plan for the pregnant woman through the first knowledge graph; a demand analysis module for collecting actual pregnancy indicator data of pregnant women, analyzing historical pregnancy indicator data and actual pregnancy indicator data to obtain pregnancy status change indicators, and outputting the pregnant woman's target nutritional needs value based on a predictive neural network according to the pregnancy status change indicators; and a second recommendation module for updating entities, attributes, and relationships in the first knowledge graph according to the output of the predictive neural network to obtain a second knowledge graph, and outputting the pregnant woman's target glucose control nutrition plan through the second knowledge graph.

[0082] Based on any embodiment of the system in this application, the system further includes: a data collection module, configured to collect historical clinical data and historical pregnancy indicator data from pregnant women with gestational diabetes through a medical information system; to obtain prior knowledge through public data sources, including publicly available research reports, prior knowledge bases, publicly available experimental data, publicly available nutritional reports, nutritional information databases, and publicly available nutritional experimental data related to gestational diabetes; the historical clinical data is first descriptive text data, which reflects the degree of influence of gestational diabetes on the pregnant woman's weight status, fetal growth status, blood glucose status, and dietary choices; the prior data is second descriptive text data, which includes data reflecting the degree of influence of dietary choices by pregnant women with gestational diabetes on the pregnant woman's weight status, fetal growth status, and blood glucose status; the historical pregnancy indicator data includes historical data on pregnant woman's weight, blood glucose, amniotic fluid volume, fetal growth status, and historical dietary choices.

[0083] Based on any embodiment of the system in this application, the system of this application further includes: a database construction module, configured to perform a correlation analysis on historical pregnancy indicator data and historical clinical data on the influence of gestational diabetes on pregnancy indicator data in the first descriptive text, to obtain pregnancy status characterization data that determines the pregnancy status profile, and construct a pregnancy status profile database through the pregnancy status characterization data; and to cross-validate the feasibility of pregnant women's dietary choices based on the first descriptive text and the second descriptive text, to obtain feasible nutrition chart data for pregnant women's dietary choices, and construct a pregnancy nutrition chart database through the feasible nutrition chart data.

[0084] Based on any embodiment of the system in this application, the system further includes: a first knowledge graph construction module, configured to obtain prior relationships between pregnant women's dietary choices, weight status, fetal growth status, and blood glucose status under gestational diabetes conditions from historical clinical data through entity recognition and relation extraction, and set a prior relationship threshold; perform entity recognition and relation extraction on the prior knowledge to obtain a first mapping relationship between pregnant women's status representation data and feasible nutrition plan data, and perform entity recognition and relation extraction on historical pregnant women's indicator data to obtain a second mapping relationship between pregnant women's status representation data and feasible nutrition plan data; filter candidate pregnant women's status representation data whose first mapping relationship meets the prior relationship threshold in the pregnant women's status profile dataset, filter candidate feasible nutrition plan data whose second mapping relationship meets the prior relationship threshold in the pregnant women's nutrition plan dataset, integrate the mapping relationship between candidate pregnant women's status representation data and candidate feasible nutrition plan data using unified coding, determine the entity relationships in the pregnant women's status profile dataset and the pregnant women's nutrition plan dataset according to the mapping relationship, and construct a first knowledge graph based on the entities in the pregnant women's status profile dataset, the entities in the pregnant women's nutrition plan dataset, and the entity relationships between the pregnant women's status profile dataset and the pregnant women's nutrition plan dataset.

[0085] Based on any embodiment of the system in this application, the system of this application further includes: a change index determination module, configured to collect actual pregnancy index data of pregnant women through a medical information system, determine a preset time length for pregnancy status change analysis based on prior knowledge, the actual pregnancy index data specifically including pregnant woman weight data, pregnant woman blood glucose data, pregnant woman amniotic fluid volume data, fetal growth status data, and actual dietary choice data during the actual pregnancy; based on the preset time length for pregnancy status change analysis, calculate the rate of change of pregnant woman weight, rate of change of pregnant woman blood glucose, rate of change of pregnant woman maternal weight, and rate of change of fetal weight in historical pregnancy index data and actual pregnancy index data, based on the first knowledge The system generates a dataset of state change indicators related to dietary choices based on a graph. It then determines assessment indicators for gestational state changes based on the degree of influence of dietary choices on changes in maternal weight, fetal growth status, and other related factors. These assessment indicators include indicators for maternal weight change, maternal weight change, and fetal weight change. Finally, it determines impact indicators for gestational state changes based on the degree of influence of amniotic fluid volume on fetal growth status and the degree of influence of dietary choices on maternal blood glucose levels. These impact indicators are set based on amniotic fluid volume change, maternal blood glucose change, maternal weight change, and fetal weight change indicators.

[0086] Based on any embodiment of the system in this application, the system of this application further includes: a nutritional requirement prediction module, configured to input the pregnancy status change assessment index and the pregnancy status change impact index into a preset prediction neural network to output the pregnant woman's target nutritional requirement value. The prediction neural network is trained using a training dataset including multiple historical clinical data, historical pregnancy index data, and corresponding actual nutritional requirement annotations. The prediction neural network is a composite network, specifically including a requirement screening network and a requirement prediction network. The requirement screening network includes a long short-term network model, a classifier model, and a discriminant screening model. The long short-term network model is used to evaluate the input pregnancy status change assessment index and the pregnancy status change impact index. Each change indicator in the estimation index is used to predict the first nutritional requirement. The classifier model is used to predict the first candidate nutritional requirement based on multiple first nutritional requirements, and compare the similarity between the predicted first candidate nutritional requirement and the actual nutritional requirement. Based on the similarity, the pregnant woman's status profile of the first candidate nutritional requirement value is classified. The first nutritional requirement includes the nutritional conversion rate of the pregnant woman's body and the nutritional conversion rate of the fetus. The first candidate nutritional requirement includes the total conversion rate of the nutrients ingested by the pregnant woman. The requirement prediction network is used to classify the pregnant woman's status profile input by the classifier model and predict the second nutritional requirement output by the discrimination and screening model to obtain the target nutritional requirement value of the pregnant woman.

[0087] Based on any embodiment of the system in this application, the system of this application further includes: a knowledge graph update module, configured to add pregnant woman status portrait entities classified based on nutritional conversion rate in the first knowledge graph according to the pregnant woman status portrait output by the classifier model, and update the entity attributes of the pregnant woman status portrait; based on the target nutritional requirement value of the pregnant woman status portrait under reclassification, update the dietary selection combination of feasible nutrition plans in the pregnant woman nutrition plan dataset, and update the entity attributes of the pregnant woman nutrition plan and the entity relationship between the pregnant woman status portrait entities and the pregnant woman nutrition plan entities, and dynamically update the first knowledge graph to a second knowledge graph.

[0088] Based on any embodiment of the system in this application, the system of this application further includes: an output module, configured to determine the pregnancy status profile of the target pregnant woman based on the actual pregnancy index data of the target pregnant woman collected, and output the target blood sugar control nutrition plan required by the target pregnant woman based on the entity relationship between the pregnant woman status profile entity and the pregnant woman nutrition plan entity in the second knowledge graph.

[0089] Another embodiment of this application provides a big data-based GDM (Geometry, Diabetes, and Nutrition) nutritional recommendation device. This device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable, non-volatile storage medium stores an operating system, a database, and computer-readable instructions. The database stores information sequences. When the processor executes the computer-readable instructions, it enables the processor to implement a big data-based GDM nutritional recommendation method.

[0090] The processor of this big data-based GDM (Geometry, Diabetes, and Nutrition) nutritional recommendation device provides computing and control capabilities, supporting the operation of the entire device. The device's memory can store computer-readable instructions, which, when executed by the processor, cause the processor to perform the big data-based GDM nutritional recommendation method of this application. The network interface of the device is used for communication with a terminal.

[0091] In this embodiment, the processor is used to execute... Figure 4 The system defines the specific functions of each module, and the memory stores the program code and various data required to execute these modules or submodules. The network interface is used to enable data transmission between user terminals or servers.

[0092] The non-volatile readable storage medium in this embodiment stores the program code and data required to execute all modules in the big data-based GDM blood sugar control nutrition recommendation system of this application. The server can call the program code and data of the server to execute the functions of all modules.

[0093] This application also provides a non-volatile readable storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the big data-based GDM blood sugar control nutrition recommendation method of any embodiment of this application.

[0094] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the method described in any embodiment of this application.

Claims

1. A method for recommending a glucose tolerance test (GDM) nutritional plan based on big data, characterized in that, Includes the following steps: We acquired historical clinical data, prior knowledge, and historical pregnancy indicator data from multiple pregnant women to construct a pregnant woman status profile database and a pregnant woman nutrition record database, respectively. Based on historical clinical data, a mapping relationship is established between the pregnant woman's status profile and the pregnant woman's nutrition plan. A first knowledge graph of the pregnant woman's status profile and the pregnant woman's nutrition plan is established according to the mapping relationship. The basic blood sugar control nutrition plan for pregnant women is output through the first knowledge graph. Collect actual pregnancy indicator data of pregnant women, analyze historical pregnancy indicator data and actual pregnancy indicator data to obtain pregnancy status change indicators, and output the target nutritional requirements of pregnant women based on the pregnancy status change indicators and a predictive neural network. The entities, attributes, and relationships in the first knowledge graph are updated based on the output of the predictive neural network to obtain the second knowledge graph. The target blood sugar control nutrition plan for pregnant women is then output through the second knowledge graph.

2. The method for recommending a blood sugar control nutrition plan based on big data according to claim 1, characterized in that, We obtained historical clinical data, prior knowledge, and historical pregnancy indicators from multiple pregnant women, as detailed below: For pregnant women with gestational diabetes, historical clinical data and historical pregnancy indicators from prenatal testing are collected through a medical information system. Prior knowledge is obtained through publicly available data sources, including publicly available research reports related to gestational diabetes, prior knowledge bases, publicly available experimental data, publicly available nutritional reports, nutritional information databases, and publicly available experimental data related to nutritional reports. The historical clinical data is the first descriptive text data, which is used to reflect the degree of impact of gestational diabetes on the pregnant woman's weight status, fetal growth status, pregnant woman's blood sugar status and pregnant woman's dietary choices. The prior data is the second descriptive text data, and the first descriptive text includes information reflecting the degree of influence of dietary choices of pregnant women with gestational diabetes on the pregnant woman's weight status, fetal growth status, and pregnant woman's blood sugar status. The historical pregnancy data includes maternal weight, maternal blood glucose, maternal amniotic fluid volume, fetal growth status, and historical dietary choices during historical periods.

3. The method for recommending a blood sugar control nutrition plan based on big data according to claim 2, characterized in that, A database of pregnant women's condition profiles and a database of pregnant women's nutrition records were constructed, as detailed below: Based on the impact of gestational diabetes on pregnancy indicators in the first descriptive text, a correlation analysis of historical pregnancy indicators and historical clinical data was conducted to obtain pregnancy status characterization data that defines the pregnancy status profile. A pregnancy status profile database was then constructed using the pregnancy status characterization data. The feasibility of pregnant women's dietary choices is cross-validated based on the first and second descriptive texts to obtain feasible nutritional data for pregnant women's dietary choices. A nutritional data database for pregnant women is then constructed based on the feasible nutritional data.

4. The method for recommending a blood sugar control nutrition plan based on big data according to claim 3, characterized in that, Based on historical clinical data, a mapping relationship was established between the pregnant woman's status profile and her nutritional records. A first knowledge graph of the pregnant woman's status profile and her nutritional records was then constructed based on this mapping relationship, as detailed below: Prior relationships between pregnant women's dietary choices, weight status, fetal growth status, and blood glucose status under gestational diabetes conditions are obtained from historical clinical data through entity recognition and relation extraction, and a prior relationship threshold is set. Entity recognition and relation extraction are performed on prior knowledge to obtain the first mapping relationship between pregnant women's status representation data and feasible nutrition plan data. Entity recognition and relation extraction are performed on historical pregnant women's indicator data to obtain the second mapping relationship between pregnant women's status representation data and feasible nutrition plan data. In the pregnant woman status profile dataset, candidate pregnant woman status representation data with a first mapping relationship satisfying the prior relationship threshold are selected. In the pregnant woman nutrition record dataset, candidate feasible nutrition record data with a second mapping relationship satisfying the prior relationship threshold are selected. The mapping relationship between the candidate pregnant woman status representation data and the candidate feasible nutrition record data is integrated using unified coding. The entity relationship in the pregnant woman status profile dataset and the pregnant woman nutrition record dataset is determined according to the mapping relationship. Based on the entities in the pregnant woman status profile dataset, the entities in the pregnant woman nutrition record dataset, and the entity relationship between the pregnant woman status profile dataset and the pregnant woman nutrition record dataset, a first knowledge graph is constructed.

5. The method for recommending a blood sugar control nutrition plan based on big data according to claim 4, characterized in that, We collected actual pregnancy indicator data from pregnant women and analyzed the historical and actual pregnancy indicator data to obtain pregnancy status change indicators, as detailed below: The actual pregnancy indicator data of pregnant women is collected through the medical information system. The preset time length for analyzing changes in pregnancy status is determined based on prior knowledge. The actual pregnancy indicator data specifically includes the pregnant woman's weight data, blood glucose data, amniotic fluid volume data, fetal growth status data, and actual dietary choices data during the actual pregnancy. Based on the preset time length of the pregnancy status change analysis, the changes in maternal weight, maternal blood sugar, maternal body weight and fetal weight are calculated in historical pregnancy index data and actual pregnancy index data. A dataset of status change indicators related to dietary choices is generated based on the first knowledge graph. Based on the degree of influence of dietary choices on changes in pregnant women's weight, changes in maternal weight, and changes in fetal growth status, assessment indicators for changes in pregnancy status are determined. These assessment indicators include indicators for changes in pregnant women's weight, changes in maternal weight, and changes in fetal weight. Based on the degree of influence of amniotic fluid volume on changes in fetal growth status and the degree of influence of dietary choices on changes in maternal blood glucose, indicators affecting changes in pregnancy status are determined. These indicators are set based on amniotic fluid volume change indicators, maternal blood glucose change indicators, maternal weight change indicators, and fetal weight change indicators.

6. The method for recommending a blood sugar control nutrition plan based on big data according to claim 5, characterized in that, Based on the state change index, a predictive neural network outputs the target nutritional requirements for pregnant women, as detailed below: The pregnancy status change assessment index and pregnancy status change impact index are input into a preset prediction neural network to output the pregnant woman's target nutritional needs value. The prediction neural network is trained through a training dataset that includes multiple historical clinical data, historical pregnancy index data and corresponding actual nutritional needs labeled. The prediction neural network is a composite network, which specifically includes a demand screening network and a demand prediction network. The demand screening network includes a long short-term network model, a classifier model and a discriminant screening model. The long short-term network model is used to predict the first nutritional requirement for each of the input pregnancy status change assessment indicators. The classifier model is used to predict the first candidate nutritional requirement based on multiple first nutritional requirements and compare the similarity between the predicted first candidate nutritional requirement and the actual nutritional requirement. Based on the similarity, the pregnant woman status profile of the first candidate nutritional requirement value is classified. The first nutritional requirement includes the nutritional conversion rate of the pregnant woman's body and the nutritional conversion rate of the fetus. The first candidate nutritional requirement includes the total conversion rate of the pregnant woman's nutrient intake. The discriminant screening model is used to update the amniotic fluid volume screening threshold and the blood glucose change screening threshold based on the input pregnancy status change impact indicators. The first candidate nutritional requirement value of the pregnant woman status profile is screened to obtain the nutritional requirement value that simultaneously meets the amniotic fluid volume screening threshold and the blood glucose change screening threshold as the second nutritional requirement input to the requirement prediction network. The demand prediction network is used to predict the second nutritional requirement of the pregnant woman based on the pregnant woman's status profile input to the classifier model and the output of the discrimination and screening model, so as to obtain the target nutritional requirement value of the pregnant woman.

7. The method for recommending a blood sugar control nutrition plan based on big data according to claim 6, characterized in that, The second knowledge graph is obtained by updating the entities, attributes, and relationships in the first knowledge graph based on the output of the predictive neural network. The target blood sugar control nutrition plan for pregnant women is then output through the second knowledge graph, as follows: Based on the pregnant woman status profile output by the classifier model, add the pregnant woman status profile entity based on the nutrient conversion rate classification in the first knowledge graph, and update the entity attributes of the pregnant woman status profile. Based on the target nutritional requirements of the pregnant woman status profile under reclassification, the dietary selection combinations of feasible nutrition plans in the pregnant woman nutrition plan dataset are updated, as well as the entity attributes of the pregnant woman nutrition plan and the entity relationship between the entities in the pregnant woman status profile and the entities in the pregnant woman nutrition plan are updated, and the first knowledge graph is dynamically updated to the second knowledge graph. Based on the actual pregnancy indicators of the target pregnant woman, a pregnancy status profile is determined. Based on the entity relationship between the pregnant woman status profile entity and the pregnant woman nutrition plan entity in the second knowledge graph, the target blood sugar control nutrition plan required by the target pregnant woman is output.

8. A GDM (Glycemic Control Nutrition Recommendation) system based on big data, characterized in that, The system is used to execute the GDM blood sugar control nutrition recommendation method based on big data as described in any one of claims 1-7, the system comprising: The data acquisition module is used to acquire historical clinical data, prior knowledge, and historical pregnancy indicator data of multiple pregnant women to build a pregnant woman status profile database and a pregnant woman nutrition record database, respectively. The first recommendation module is used to set the mapping relationship between the pregnant woman's status profile and the pregnant woman's nutrition plan based on historical clinical data. It establishes a first knowledge graph of the pregnant woman's status profile and the pregnant woman's nutrition plan based on the mapping relationship, and outputs the basic blood sugar control nutrition plan of the pregnant woman through the first knowledge graph. The requirements analysis module is used to collect actual pregnancy indicator data of pregnant women, analyze historical pregnancy indicator data and actual pregnancy indicator data to obtain pregnancy status change indicators, and output the target nutritional requirements of pregnant women based on the pregnancy status change indicators and a predictive neural network. The second recommendation module is used to update the entities, attributes, and relationships in the first knowledge graph based on the output of the prediction neural network, obtain the second knowledge graph, and output the pregnant woman's target blood sugar control nutrition plan through the second knowledge graph.

9. A GDM (Glycemic Control and Nutritional Recommendation) device based on big data, comprising: At least one processor, and, A memory communicatively connected to the at least one processor; characterized in that the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the big data-based GDM blood sugar control nutrition recommendation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 7, which, when invoked by a computer, executes the steps included in the corresponding method.

Images