Gestational diabetes mellitus tracking follow-up visit regulation and control device

Abstract

The invention discloses a gestational diabetes mellitus tracking follow-up visit regulation and control device, and belongs to the technical field of medical treatment. A gestational diabetes mellitus tracking follow-up visit regulation and control device comprises a data acquisition module, a feature extraction module, a model prediction module and an auxiliary decision making module. The problems that existing diabetes mellitus follow-up visit is high in cost and insufficient in patient subjective initiative are solved, tracking follow-up visit regulation and control are conducted on gestational diabetes mellitus, a large amount of historical blood glucose data of pregnant women are collected and preprocessed, then feature extraction is conducted, a blood glucose prediction model is constructed, the data and the deep learning technology are combined, and the gestational diabetes mellitus tracking follow-up visit regulation and control method is provided for the gestational diabetes mellitus. By means of the method, data analysis and prediction can be better achieved, early-stage discovery and intervention of diseases are improved, pregnant woman blood glucose data collected in follow-up visit are analyzed through the blood glucose prediction model, doctors can be helped to judge the blood glucose condition of a pregnant woman timely and accurately, a more reasonable follow-up visit plan and a more reasonable treatment scheme are made, and the accuracy of follow-up visit is improved. Therefore, risks of pregnant women and fetuses can be reduced.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and particularly to a tracking, follow-up and regulation device for gestational diabetes mellitus. Background Art

[0002] Diabetes is a common metabolic disease that affects the patient's body and health. With the changes in modern lifestyle and eating habits, the incidence of gestational diabetes has been increasing year by year. During pregnancy, diabetic patients need to pay special attention to their blood sugar levels to avoid adverse effects on the mother and fetus, so the tracking and follow-up of gestational diabetes becomes particularly important. The traditional diabetes management methods mainly include doctors regularly conducting physical examinations, blood sugar tests and medication guidance for patients, etc., but this method has some problems, such as high time cost, multiple trips to the hospital are required, and the patient's subjective initiative is insufficient, making it difficult to adhere to a healthy lifestyle for a long time; therefore, it does not meet the existing needs, and for this reason, we propose a tracking, follow-up and regulation device for gestational diabetes. Summary of the Invention

[0003] The purpose of the present invention is to provide a tracking, follow-up and regulation device for gestational diabetes. By collecting and preprocessing a large amount of historical blood sugar data of pregnant women, then extracting features and constructing a blood sugar prediction model, it can analyze the blood sugar data of pregnant women collected during follow-up to predict the blood sugar status of pregnant women, and the prediction results are displayed in the form of charts or graphs, providing an important basis for doctors to evaluate the condition, make a diagnosis and formulate subsequent treatment plans, and solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solutions: A tracking, follow-up and regulation device for gestational diabetes, comprising:

[0005] A data acquisition module, used for:

[0006] Formulating a follow-up plan, collecting the blood sugar data of pregnant women according to the formulated follow-up plan, and at the same time, collecting a large amount of historical blood sugar data of pregnant women, and respectively preprocessing the collected blood sugar data and historical blood sugar data;

[0007] A feature extraction module, used for:

[0008] Extracting features from the preprocessed historical blood sugar data, and dividing the historical blood sugar data into a training set and a test set;

[0009] A model prediction module, used for:

[0010] Using the historical blood sugar data processed by the data acquisition module, constructing a blood sugar prediction model by deep learning;

[0011] Train a blood glucose prediction model using the data in the training set and evaluate the performance of the blood glucose prediction model using the data in the test set;

[0012] Deploy the trained blood glucose prediction model to actual applications, and analyze the blood glucose data collected during follow-up through the blood glucose prediction model to predict the blood glucose situation of pregnant women;

[0013] An auxiliary decision-making module, used for:

[0014] Convert the prediction results of the model prediction module into charts or graphs for display;

[0015] The displayed prediction results are provided for doctors to evaluate the condition, make a diagnosis, and formulate subsequent treatment plans.

[0016] Furthermore, the data acquisition module includes:

[0017] A receipt collection module, used for:

[0018] Formulate a follow-up plan, including the time and frequency of follow-up;

[0019] According to the formulated follow-up plan, real-time monitor the blood glucose information of pregnant women through wearable devices or sensors, and collect the blood glucose data of pregnant women;

[0020] At the same time, collect a large amount of historical blood glucose data of pregnant women;

[0021] A data processing module, used for:

[0022] Preprocess the collected blood glucose data and historical blood glucose data respectively, including data cleaning, format conversion, and data reduction;

[0023] A data storage module, used for:

[0024] Upload the blood glucose data collected during follow-up to the cloud database for storage;

[0025] Among them, after training the blood glucose prediction model, the model prediction module retrieves the blood glucose data from the cloud database for prediction.

[0026] Furthermore, the preprocessing is specifically:

[0027] Data cleaning: Identify and correct missing values and outliers in the blood glucose data and historical blood glucose data;

[0028] For missing values, either delete the data containing missing values or fill in the missing values with the mean and median;

[0029] For outliers, directly delete the data containing outliers;

[0030] Format conversion: Convert the blood glucose data and historical blood glucose data into a unified format;

[0031] Data reduction: Delete the duplicate, invalid, and redundant data in the blood glucose data and historical blood glucose data.

[0032] Further, after data reduction, determine whether the sample size of the blood glucose data meets the model training requirements, including:

[0033] Extract the blood glucose data before data reduction;

[0034] Use the blood glucose data before data reduction to obtain the unit observation period duration;

[0035] Among them, the unit observation period duration is obtained through the following formula:

[0036]

[0037] Among them, S represents the unit observation period duration; T cp represents the average time interval of blood glucose data collection; S0 represents the preset unit observation period duration reference value; n represents the number of pregnant women included in the blood glucose data; X bi represents the standard deviation of blood glucose values corresponding to the i-th pregnant woman; T bi represents the standard deviation of blood glucose collection time intervals corresponding to the i-th pregnant woman;

[0038] Extract the blood glucose data after data reduction, and use the unit observation period duration to determine whether the sample size of the blood glucose data meets the model training requirements.

[0039] Further, extract the blood glucose data after data reduction, and use the unit observation period duration to determine whether the sample size of the blood glucose data meets the model training requirements, including:

[0040] Extract the blood glucose data after data reduction;

[0041] Use the blood glucose data after data reduction and the blood glucose data before data reduction to obtain the unit observation period duration adjustment coefficient;

[0042] Among them, the unit observation period duration adjustment coefficient is obtained through the following formula:

[0043]

[0044] Among them, J represents the unit observation period duration adjustment coefficient; X bx represents the standard deviation of the overall blood glucose data before data reduction; X bh represents the standard deviation of the overall blood glucose data after data reduction; T bxIndicates the standard deviation of the overall blood glucose data collection time interval before data reduction; T bh Indicates the standard deviation of the overall blood glucose data collection time interval after data reduction; M h Indicates the sample size of the overall blood glucose data after data reduction; M x Indicates the sample size of the overall blood glucose data before data reduction;

[0045] Adjust the unit observation period duration using the unit observation period duration adjustment coefficient to obtain the adjusted unit observation period duration;

[0046] Among them, the adjusted unit observation period duration is obtained through the following formula:

[0047]

[0048] Among them, S t Indicates the adjusted unit observation period duration; J indicates the unit observation period duration adjustment coefficient; S indicates the unit observation period duration;

[0049] For the blood glucose data after data reduction, extract the sample size of the blood glucose data within each adjusted unit observation period duration, and obtain the average value of the sample size of the blood glucose data corresponding to each adjusted unit observation period duration;

[0050] When the average value of the sample size of the blood glucose data corresponding to each adjusted unit observation period duration is lower than the preset sample size threshold, it is determined that the sample size of the blood glucose data does not meet the model training requirements, and a sample size insufficient warning is issued.

[0051] Furthermore, the cloud database includes the following functions:

[0052] Classification storage function: Classify and store the blood glucose data collected during follow-up in chronological order;

[0053] Regard the blood glucose data collected each time during follow-up as a time entry, including name, date, and timestamp;

[0054] Data backup function: Preset a backup strategy, including backup period and frequency;

[0055] Execute backup operations regularly according to the backup strategy to realize backup and storage of the stored blood glucose data.

[0056] Furthermore, the feature extraction module includes:

[0057] Data feature extraction module, used for:

[0058] Use a variety of feature extraction methods to extract features from the preprocessed historical blood glucose data. Among them, the feature extraction methods include mean feature extraction, median feature extraction, standard deviation feature extraction, and correlation coefficient feature extraction;

[0059] A data partitioning module, for:

[0060] Determine the partitioning ratio, and the partitioning ratio of the training set and the test set is 7:3 or 8:2;

[0061] According to the partitioning ratio, extract corresponding proportion of data samples from the historical blood glucose data as the training set and the test set;

[0062] The training set is used to train the blood glucose prediction model, and the test set is used to evaluate the performance of the blood glucose prediction model.

[0063] Furthermore, the data feature extraction module is specifically:

[0064] Mean feature extraction: Calculate the average value of blood glucose at each time point of the historical blood glucose data as the overall representative of the blood glucose level;

[0065] Median feature extraction: Calculate the median of blood glucose at each time point of the historical blood glucose data to reflect the change trend of the blood glucose level;

[0066] Standard deviation feature extraction: Calculate the standard deviation of blood glucose at each time point of the historical blood glucose data to reflect the degree of dispersion of the blood glucose level;

[0067] Correlation coefficient feature extraction: Calculate the correlation coefficient of blood glucose at adjacent time points of the historical blood glucose data to reflect the synchronization and continuity of blood glucose changes.

[0068] Furthermore, the model prediction module includes:

[0069] A model construction module, for:

[0070] Based on the historical blood glucose data processed by the data acquisition module, use deep learning to construct a blood glucose prediction model;

[0071] Among them, deep learning uses a convolutional neural network model or a recurrent neural network model to construct a blood glucose prediction model;

[0072] A model training module, for:

[0073] Set model parameters, including the size of the model, learning rate, and regularization parameters;

[0074] Use the data in the training set to train the blood glucose prediction model, and continuously update the model parameters during the training process;

[0075] A model optimization module, for:

[0076] Evaluate the performance of the blood glucose prediction model using the data in the test set, including evaluating the accuracy, precision, and recall rate of the blood glucose prediction model;

[0077] If the performance of the blood glucose prediction model is poor, optimize the blood glucose prediction model, including adding or deleting features and adjusting parameters;

[0078] The model deployment module is used for:

[0079] Deploy the trained and optimized blood glucose prediction model into actual applications, predict the blood glucose data collected during follow-up through the blood glucose prediction model, and understand the blood glucose conditions of pregnant women based on the prediction results to discover potential blood glucose risks of pregnant women.

[0080] Furthermore, the auxiliary decision-making module includes:

[0081] The result display module is used for:

[0082] Obtain the prediction results from the model prediction module and display the prediction results in the form of charts or graphs. Among them, the charts include bar charts, pie charts, and line charts, and the graphs include bar charts, pie charts, heat maps, and scatter plots;

[0083] The user interface module is used for:

[0084] Provide a user interface through which doctors can access and view the displayed prediction results;

[0085] The treatment plan recommendation module is used for:

[0086] Doctors conduct disease assessment, diagnosis, and formulation of subsequent treatment plans based on the accessed and viewed prediction results, and recommend the formulated treatment plans to pregnant women via text messages or emails.

[0087] Compared with the prior art, the beneficial effects of the present invention are:

[0088] For the tracking, follow-up, and regulation of gestational diabetes, the present invention can establish an effective diabetes prediction model by collecting a large amount of historical blood glucose data and combining steps such as data preprocessing, feature extraction, and model construction. Combining data with deep learning technology can better realize data analysis and prediction, thereby improving the early detection and intervention of diseases. By analyzing the blood glucose data of pregnant women collected during follow-up through the blood glucose prediction model, doctors can be helped to judge the blood glucose conditions of pregnant women in a timely and accurate manner, and then formulate more reasonable follow-up plans and treatment plans, thereby reducing the risks of pregnant women and fetuses and saving medical resources. Description of the Drawings

[0089] Figure 1This is a schematic structural diagram of the gestational diabetes mellitus tracking, follow-up and regulation device of the present invention. Detailed implementation manners

[0090] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0091] To solve the technical problems that the existing diabetes management methods mainly include doctors regularly conducting physical examinations, blood glucose tests and medication guidance for patients, resulting in high time costs, multiple trips to the hospital, and insufficient subjective initiative of patients, making it difficult to adhere to a healthy lifestyle for a long time, please refer to Figure 1 This embodiment provides the following technical solutions:

[0092] A gestational diabetes mellitus tracking, follow-up and regulation device includes:

[0093] A data acquisition module, which is used for:

[0094] Formulating a follow-up plan, collecting pregnant women's blood glucose data according to the formulated follow-up plan, at the same time, collecting a large amount of historical blood glucose data of pregnant women, and respectively preprocessing the collected blood glucose data and historical blood glucose data;

[0095] A feature extraction module, which is used for:

[0096] Extracting features from the preprocessed historical blood glucose data, and dividing the historical blood glucose data into a training set and a test set;

[0097] A model prediction module, which is used for:

[0098] Using the historical blood glucose data processed by the data acquisition module, constructing a blood glucose prediction model by deep learning;

[0099] Using the data in the training set to train the blood glucose prediction model, and using the data in the test set to evaluate the performance of the blood glucose prediction model;

[0100] Deploying the trained blood glucose prediction model into actual applications, analyzing the blood glucose data collected during follow-up through the blood glucose prediction model, and predicting the blood glucose conditions of pregnant women;

[0101] An auxiliary decision-making module, which is used for:

[0102] Converting the prediction results of the model prediction module into the form of charts or graphs for display;

[0103] The predicted results shown are for doctors to evaluate the condition, make a diagnosis, and formulate subsequent treatment plans.

[0104] The technical effects of the above are as follows: The data acquisition module collects and preprocesses the follow-up blood glucose data of pregnant women and a large amount of historical blood glucose data. The purpose of preprocessing is to improve the data quality, making subsequent model training and prediction more efficient and accurate. Then, through the feature extraction module and the model prediction module, features are extracted from the historical blood glucose data and a blood glucose prediction model is constructed. By combining data with deep learning technology, data analysis and prediction can be better achieved, thereby improving the early detection and intervention of diseases. And by using the constructed blood glucose prediction model to predict the blood glucose data of pregnant women collected during follow-up, the prediction results can be displayed through the auxiliary decision-making module, thus helping doctors timely and accurately judge the blood glucose status of pregnant women, and then formulating a more reasonable follow-up plan and treatment plan to reduce the risks of pregnant women and fetuses.

[0105] The data acquisition module includes:

[0106] The receipt collection module is used for:

[0107] Formulate a follow-up plan, including the time and frequency of follow-up. The follow-up plan should consider factors such as the health status, gestational age, and lifestyle of pregnant women in order to customize a suitable follow-up strategy for pregnant women;

[0108] According to the formulated follow-up plan, the blood glucose information of pregnant women is monitored in real time through wearable devices or sensors, and the blood glucose data of pregnant women is collected;

[0109] At the same time, a large amount of historical blood glucose data of pregnant women is collected. The historical blood glucose data can come from past examinations or self-records;

[0110] The data processing module is used for:

[0111] Preprocess the collected blood glucose data and historical blood glucose data respectively, including data cleaning, format conversion, and data reduction;

[0112] The data storage module is used for:

[0113] Upload the blood glucose data collected during follow-up to the cloud database for storage;

[0114] Among them, after training the blood glucose prediction model, the model prediction module retrieves the blood glucose data from the cloud database for prediction.

[0115] The technical effects of the above content are as follows: The receipt collection module can formulate a follow-up plan in advance to ensure continuous and regular acquisition of pregnant women's blood glucose data. It can monitor pregnant women's blood glucose information in real time through wearable devices or sensors, collect pregnant women's real-time blood glucose data, and at the same time, collect a large amount of historical blood glucose data of pregnant women. The collected blood glucose data and historical blood glucose data provide data support for subsequent model construction and prediction. The data processing module is responsible for preprocessing the collected blood glucose data and historical blood glucose data, which can improve the data quality and make subsequent model training and prediction more efficient and accurate. The data storage module can upload the blood glucose data collected during the follow-up to the cloud database for storage, so as to facilitate the subsequent model prediction module to call.

[0116] Preprocessing, specifically:

[0117] Data cleaning: Identify and correct missing values and outliers in the blood glucose data and historical blood glucose data;

[0118] For missing values, either delete the data containing missing values or fill in the missing values with the mean and median;

[0119] For outliers, directly delete the data containing outliers;

[0120] Format conversion: Convert the blood glucose data and historical blood glucose data into a unified format;

[0121] Data reduction: Delete duplicate, invalid, and redundant data in the blood glucose data and historical blood glucose data.

[0122] The technical effects of the above content are as follows: Data cleaning ensures the quality and integrity of the data by identifying and correcting missing values and outliers in the blood glucose data and historical blood glucose data. For example, for missing values, the data containing missing values can be deleted or the missing values can be filled in with the mean and median to avoid data anomalies caused by missing values. For outliers, the data containing outliers can be directly deleted to prevent the impact of outliers on the model. Format conversion can convert the blood glucose data and historical blood glucose data into a unified format, and the unified format can improve the comparability of the data, which is beneficial to subsequent data analysis and model construction. Data reduction can delete duplicate, invalid, and redundant data in the blood glucose data and historical blood glucose data, reduce the data volume, improve the calculation efficiency, and make the data analysis process more efficient. The purpose of the above three preprocessing methods is to improve the data quality and make subsequent analysis and model construction work more accurate, stable, and efficient.

[0123] Specifically, after data reduction, determine whether the sample size of the blood glucose data meets the model training requirements, including:

[0124] Extract the blood glucose data before data reduction;

[0125] Obtain the duration of a unit observation period using the blood glucose data before the data reduction;

[0126] Wherein, the duration of the unit observation period is obtained through the following formula:

[0127]

[0128] Wherein, S represents the duration of the unit observation period; T cp represents the average value of the time intervals for blood glucose data collection; S0 represents the preset reference value of the duration of the unit observation period; n represents the number of pregnant women included in the blood glucose data; X bi represents the standard deviation of the blood glucose values corresponding to the i-th pregnant woman; T bi represents the standard deviation of the blood glucose collection time intervals corresponding to the i-th pregnant woman;

[0129] Extract the blood glucose data after the data reduction, and determine whether the sample quantity of the blood glucose data meets the model training requirements by using the duration of the unit observation period.

[0130] The technical effects of the above technical solution are as follows: By extracting the blood glucose data before the data reduction and calculating the duration of the unit observation period, this solution can more accurately evaluate the overall characteristics and quality of the data. In particular, by considering the average value of the time intervals for blood glucose data collection, the preset reference value of the duration of the unit observation period, the number of pregnant women involved, and the standard deviation of the blood glucose values and the standard deviation of the blood glucose collection time intervals for each pregnant woman, this solution provides a comprehensive metric to reflect the stability and consistency of the data. Using the calculated duration of the unit observation period, this solution can determine whether the sample quantity of the blood glucose data after the data reduction meets the model training requirements. This helps to ensure that too much key information is not lost during the data reduction process, thereby maintaining the accuracy and effectiveness of model training. Through reasonable data reduction and sample quantity evaluation, this solution helps to reduce the unnecessary computational burden and improve the efficiency of model training. At the same time, ensuring that the sample quantity meets the training requirements also helps to improve the accuracy and generalization ability of the model. The standard deviation of the blood glucose values and the standard deviation of the blood glucose collection time intervals for each pregnant woman are considered in the solution, which reflects the attention to individual differences. By evaluating the data characteristics of each pregnant woman individually, this solution can more accurately reflect the complexity of the real world, thereby helping to build a more robust and accurate model. This solution provides a systematic method for data reduction and model training, which helps to optimize the entire data processing process. Through clear steps and calculation formulas, this solution makes the data processing process more transparent and replicable, facilitating popularization and implementation in practical applications.

[0131] In summary, by comprehensively considering multiple dimensions and characteristics of the data, and personalized consideration of the data of each pregnant woman, this technical solution provides an effective and accurate method for the processing of blood glucose data and model training. This not only helps to improve the efficiency and accuracy of model training, but also helps to optimize the entire data processing process, providing strong support for related research and applications.

[0132] Specifically, extract the blood glucose data after data reduction, and use the unit observation period duration to determine whether the sample quantity of the blood glucose data meets the model training requirements, including:

[0133] Extract the blood glucose data after data reduction;

[0134] Obtain the unit observation period duration adjustment coefficient by using the blood glucose data after data reduction and the blood glucose data before data reduction;

[0135] Among them, the unit observation period duration adjustment coefficient is obtained through the following formula:

[0136]

[0137] Among them, J represents the unit observation period duration adjustment coefficient; X bx represents the standard deviation of the overall blood glucose data before data reduction; X bh represents the standard deviation of the overall blood glucose data after data reduction; T bx represents the standard deviation of the overall blood glucose data collection time interval before data reduction; T bh represents the standard deviation of the overall blood glucose data collection time interval after data reduction; M h represents the sample quantity of the overall blood glucose data after data reduction; M x represents the sample quantity of the overall blood glucose data before data reduction;

[0138] Adjust the unit observation period duration by using the unit observation period duration adjustment coefficient to obtain the adjusted unit observation period duration;

[0139] Among them, the adjusted unit observation period duration is obtained through the following formula:

[0140]

[0141] Among them, S t represents the adjusted unit observation period duration; J represents the unit observation period duration adjustment coefficient; S represents the unit observation period duration;

[0142] Extract the number of samples of blood glucose data within each adjusted unit observation period duration after data reduction, and obtain the average value of the number of samples of blood glucose data corresponding to each adjusted unit observation period duration;

[0143] When the average value of the number of samples of blood glucose data corresponding to each adjusted unit observation period duration is lower than a preset sample number threshold, it is determined that the number of samples of blood glucose data does not meet the model training requirements, and a warning of insufficient sample number is given.

[0144] The technical effects of the above technical solution are as follows: By comparing the standard deviation of blood glucose data and the standard deviation of the acquisition time interval before and after data reduction, the unit observation period duration adjustment coefficient is calculated. This technical solution can evaluate the impact of data reduction on the characteristics of blood glucose data. This helps to ensure that the data reduction process does not overly lose key information and maintains the stability and representativeness of the data. Using the unit observation period duration adjustment coefficient to adjust the unit observation period duration makes the adjusted unit observation period duration better reflect the actual situation after data reduction. This helps to more accurately evaluate whether the number of samples of blood glucose data meets the model training requirements in subsequent analysis. By extracting the number of samples of blood glucose data within each adjusted unit observation period duration and calculating the average value of the number of samples, this technical solution can accurately evaluate whether the number of samples after data reduction is sufficient. This helps to ensure that there are enough samples in the model training process to support the generalization ability and accuracy of the model. When it is found that the average value of the number of samples of blood glucose data corresponding to each adjusted unit observation period duration is lower than the preset sample number threshold, this technical solution can flexibly respond and further determine that the number of samples of blood glucose data does not meet the model training requirements. This helps to adjust the model training strategy according to the data situation in practical applications and ensure the effectiveness and accuracy of model training. This technical solution provides a systematic method for data reduction, unit observation period duration adjustment, and sample number evaluation, which helps to optimize the entire data processing process. Through clear steps and calculation formulas, this solution makes the data processing process more transparent, replicable, and controllable, and is convenient for popularization and implementation in practical applications.

[0145] In summary, this technical solution provides an effective and accurate method for the processing of blood glucose data and model training by comprehensively considering the data characteristics before and after data reduction, adaptively adjusting the unit observation period duration, and accurately evaluating the sample number. This not only helps to improve the efficiency and accuracy of model training, but also helps to optimize the entire data processing process and provides strong support for related research and applications.

[0146] The cloud database includes the following functions:

[0147] Classification storage function: Classify and store the blood glucose data collected during follow-up in chronological order;

[0148] Blood glucose data collected at each follow-up visit were considered as a time entry, including name, date, and time stamp;

[0149] Data backup function: pre-set backup strategy, including backup cycle and frequency;

[0150] Perform backup operations regularly according to the backup strategy to back up and save the stored blood sugar data.

[0151] The technical effect of the above content is: by classifying and storing the blood sugar data collected during follow-up in chronological order, the data can be easily managed and queried, which facilitates long-term tracking and observation of blood sugar changes. By regarding the blood sugar data collected during each follow-up as a time entry, the recording method of the time entry can help us better understand the changing trend of blood sugar. For example, we can view the changes in blood sugar according to the time period of one day, or we can view the changes in blood sugar according to the time period of one week or one month, so as to better understand the actual situation of blood sugar control. By pre-setting the backup strategy, we can ensure the security of the data and prevent data loss. If data loss occurs, it can be restored according to the backup record.

[0152] Feature extraction module, including:

[0153] Data feature extraction module, used for:

[0154] A variety of feature extraction methods are used to extract features from the preprocessed historical blood glucose data. The feature extraction methods include mean feature extraction, median feature extraction, standard deviation feature extraction, and correlation coefficient feature extraction. In practical applications, a single feature extraction method is often difficult to fully capture all the information of the data. Therefore, a combination of multiple feature extraction methods is usually used to obtain a more comprehensive feature representation, which can not only improve the accuracy of the model, but also help to discover potential patterns and laws. Among them:

[0155] Mean feature extraction: Calculate the average blood sugar value at each time point of historical blood sugar data as the overall representative of blood sugar level;

[0156] Median feature extraction: Calculate the median of blood sugar at each time point in the historical blood sugar data to reflect the changing trend of blood sugar levels;

[0157] Standard deviation feature extraction: Calculate the standard deviation of blood sugar at each time point in historical blood sugar data to reflect the degree of dispersion of blood sugar levels;

[0158] Correlation coefficient feature extraction: Calculate the correlation coefficient of blood sugar at adjacent time points of historical blood sugar data to reflect the synchronization and continuity of blood sugar changes;

[0159] A data partitioning module, configured to:

[0160] Determine the partitioning ratio, where the partitioning ratio of the training set to the test set is 7:3 or 8:2;

[0161] Extract corresponding proportion of data samples from the historical blood glucose data as the training set and the test set according to the partitioning ratio;

[0162] The training set is used to train the blood glucose prediction model, and the test set is used to evaluate the performance of the blood glucose prediction model.

[0163] The technical effect of the above content is as follows: The main function of the data feature extraction module is to extract features from the preprocessed historical blood glucose data. The extracted features can more accurately reflect the changes and patterns of blood glucose levels, thereby improving the prediction accuracy of the blood glucose prediction model. Specifically, the feature extraction module uses a variety of feature extraction methods to deeply mine the historical blood glucose data and extracts features that can reflect various aspects such as blood glucose change trends, blood glucose level change ranges, blood glucose change synchronization, and continuity. The extracted features can better depict the overall picture of blood glucose data to improve the prediction ability of the prediction model. The data partitioning module divides the historical blood glucose data into a training set and a test set according to a specific partitioning ratio. In this way, it can avoid the model's over-reliance on the training set data, ensure the generalization ability of the model, and improve the accuracy and reliability of the prediction model.

[0164] A model prediction module, including:

[0165] A model construction module, configured to:

[0166] Based on the historical blood glucose data processed by the data acquisition module, construct a blood glucose prediction model using deep learning;

[0167] Among them, deep learning uses a convolutional neural network model or a recurrent neural network model to construct the blood glucose prediction model;

[0168] A model training module, configured to:

[0169] Set model parameters, including the size of the model, learning rate, and regularization parameters;

[0170] Use the data in the training set to train the blood glucose prediction model, and continuously update the model parameters during the training process;

[0171] A model optimization module, configured to:

[0172] Use the data in the test set to evaluate the performance of the blood glucose prediction model, including evaluating the accuracy, precision, and recall rate of the blood glucose prediction model;

[0173] If the performance of the blood glucose prediction model is poor, optimize the blood glucose prediction model, including adding or deleting features and adjusting parameters;

[0174] A model deployment module, used for:

[0175] Deploy the trained and optimized blood glucose prediction model into actual applications, predict the blood glucose data collected during follow-up through the blood glucose prediction model, and understand the blood glucose conditions of pregnant women based on the prediction results to discover potential blood glucose risks of pregnant women.

[0176] The technical effects of the above content are as follows: The model construction module constructs a model capable of predicting the blood glucose of pregnant women based on historical blood glucose data and using deep learning methods (such as convolutional neural networks or recurrent neural networks). By combining data with deep learning technology, it can effectively predict the changes in the blood glucose of pregnant women, thereby helping doctors and pregnant women better understand and control blood glucose. The model training module sets the parameters of the model (such as the size of the convolutional kernel, learning rate, etc.) and trains the blood glucose prediction model using the data in the training set. During the training process, the blood glucose prediction model continuously updates its parameters to achieve the best prediction effect. When the blood glucose prediction model performs poorly on the training set, the model optimization module is used to change the model structure, adjust parameters, etc. to further improve the prediction ability of the model. After the blood glucose prediction model is trained and optimized, the blood glucose prediction model is deployed into the actual application environment through the model deployment module. By predicting the blood glucose data collected during follow-up through the blood glucose prediction model, the blood glucose conditions of pregnant women can be understood, and potential blood glucose risks can be predicted, providing data support for doctors' diagnosis and treatment.

[0177] An auxiliary decision-making module, including:

[0178] A result display module, used for:

[0179] Obtain the prediction results from the model prediction module and display the prediction results in the form of charts or graphs. Among them, the charts include bar charts, pie charts, and line charts, and the graphs include bar charts, pie charts, heat maps, and scatter plots;

[0180] A user interface module, used for:

[0181] Provide a user interface through which doctors can access and view the displayed prediction results;

[0182] A solution recommendation module, used for:

[0183] Doctors conduct disease assessment, diagnosis, and formulation of subsequent treatment plans based on the accessed and viewed prediction results, and recommend the formulated treatment plans to pregnant women via text messages or emails.

[0184] The technical effects of the above content are as follows: The result display module obtains the prediction results from the model prediction module and displays them in an intuitive and easy-to-understand chart or graph form. Doctors can access and view the prediction results through the user interface provided by the user interface module, which can help doctors understand the prediction results more clearly and intuitively, thus making it easier to conduct condition assessment and diagnosis. The treatment plan recommendation module allows doctors to formulate the most suitable treatment plan for the patient based on the accessed and viewed prediction results, combined with the specific situation of the patient and the professional judgment of the doctor. At the same time, the treatment plan recommendation module can also notify the pregnant woman of the treatment plan by means of text messages, emails, etc., to help the pregnant woman better execute the doctor's orders.

[0185] Working principle: The data acquisition module collects and preprocesses the follow-up blood glucose data of pregnant women and a large amount of historical blood glucose data. Preprocessing can improve the data quality, making subsequent model training and prediction more efficient and accurate. The feature extraction module extracts features from the preprocessed historical blood glucose data to improve the prediction accuracy of the blood glucose prediction model. The model prediction module can construct a blood glucose prediction model for predicting the blood glucose changes of pregnant women by combining historical blood glucose data with deep learning technology, which can better realize data analysis and prediction. By using the constructed blood glucose prediction model to predict the blood glucose data of pregnant women collected during follow-up, the prediction results can be displayed through the auxiliary decision-making module, which can help doctors timely and accurately judge the blood glucose status of pregnant women, and then formulate a more reasonable follow-up plan and treatment plan to reduce the risks of pregnant women and fetuses.

[0186] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0187] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention.

Claims

1. A device for tracking and following up gestational diabetes, characterized in that: include: Data acquisition module, used to: Formulate a follow-up plan and collect blood sugar data of pregnant women according to the formulated follow-up plan. At the same time, collect a large amount of historical blood sugar data of pregnant women and pre-process the collected blood sugar data and historical blood sugar data respectively; Feature extraction module for: Perform feature extraction on the preprocessed historical blood glucose data, and divide the historical blood glucose data into a training set and a test set; Model prediction module, used for: Using the historical blood sugar data processed by the data acquisition module, deep learning is used to build a blood sugar prediction model; Use the data of the training set to train the blood glucose prediction model, and use the data of the test set to evaluate the performance of the blood glucose prediction model; Deploy the trained blood sugar prediction model to practical applications, analyze the blood sugar data collected during follow-up visits through the blood sugar prediction model, and predict the blood sugar status of pregnant women; Decision support module for: Convert the prediction results of the model prediction module into charts or graphs for display; The displayed prediction results are used by doctors to evaluate the condition, diagnose and formulate subsequent treatment plans.

2. The device for tracking and following up gestational diabetes mellitus according to claim 1, characterized in that: The data acquisition module comprises: A receipt collection module for: Develop a follow-up plan, including the timing and frequency of follow-up visits; According to the established follow-up plan, the pregnant woman's blood sugar information is monitored in real time through wearable devices or sensors, and the pregnant woman's blood sugar data is collected; At the same time, a large amount of historical blood sugar data of pregnant women is collected; Data processing module for: Preprocess the collected blood glucose data and historical blood glucose data separately, including data cleaning, format conversion and data reduction; Data storage module for: Upload the blood glucose data collected during follow-up to the cloud database for storage; Among them, after training the blood glucose prediction model, the model prediction module retrieves blood glucose data from the cloud database for prediction.

3. The device for tracking and following up gestational diabetes mellitus according to claim 2, characterized in that: The preprocessing is specifically as follows: Data cleaning: Identify and correct missing values ​​and outliers in blood glucose data and historical blood glucose data; For missing values, the data containing missing values ​​are deleted, or the mean and median are used to fill the missing values; For outliers, the data containing outliers are directly deleted; Format conversion: convert blood glucose data and historical blood glucose data into a unified format; Data reduction: Delete duplicate, invalid and redundant data in blood glucose data and historical blood glucose data.

4. The device for tracking and following up gestational diabetes mellitus according to claim 3, characterized in that: After data reduction, determine whether the number of blood glucose data samples meets the model training requirements, including: Extract blood glucose data before data reduction; Obtaining the unit observation period duration by using the blood sugar data before data reduction; The unit observation period duration is obtained by the following formula: Among them, S represents the unit observation period length; T cp represents the average time interval of blood glucose data collection; S0 represents the preset unit observation period duration benchmark value; n represents the number of pregnant women included in the blood glucose data; X bi represents the standard deviation of blood sugar value corresponding to the i-th pregnant woman; T bi represents the standard deviation of the blood glucose collection time interval corresponding to the i-th pregnant woman; Extract the blood glucose data after data reduction, and use the unit observation period to determine whether the number of blood glucose data samples meets the model training requirements.

5. The device for tracking and following up gestational diabetes mellitus according to claim 4, characterized in that: Extract the blood sugar data after data reduction, and use the unit observation period to determine whether the number of blood sugar data samples meets the model training requirements, including: Extract blood glucose data after data reduction; Obtaining a unit observation period duration adjustment coefficient using the blood sugar data after the data reduction and the blood sugar data before the data reduction; The unit observation period adjustment coefficient is obtained by the following formula: Among them, J represents the unit observation period length adjustment coefficient; X bx represents the standard deviation of the overall blood glucose data before data reduction; X bh represents the standard deviation of the overall blood glucose data after data reduction; T bx represents the standard deviation of the overall blood glucose data collection time interval before data reduction; T bh represents the standard deviation of the overall blood glucose data collection time interval after data reduction; M h Represents the number of samples of the overall blood glucose data after data reduction; M x Represents the number of samples of the overall blood glucose data before data reduction; The unit observation period duration is adjusted using the unit observation period duration adjustment coefficient to obtain an adjusted unit observation period duration; The adjusted unit observation period duration is obtained by the following formula: Among them, S t represents the adjusted unit observation period; J represents the unit observation period adjustment coefficient; S represents the unit observation period; Extracting the number of samples of blood sugar data within each adjusted unit observation period from the blood sugar data after data reduction, and obtaining the average number of samples of blood sugar data corresponding to each adjusted unit observation period; When the average number of samples of blood glucose data corresponding to each adjusted unit observation period duration is lower than a preset sample number threshold, it is determined that the sample number of blood glucose data does not meet the model training requirements, and an insufficient sample number warning is issued.

6. The device for tracking and following up gestational diabetes mellitus according to claim 2, characterized in that: The cloud database includes the following functions: Classification storage function: Classify and store the blood sugar data collected during follow-up in chronological order; Blood glucose data collected at each follow-up visit were considered as a time entry, including name, date, and time stamp; Data backup function: pre-set backup strategy, including backup cycle and frequency; Perform backup operations regularly according to the backup strategy to back up and save the stored blood sugar data.

7. The device for tracking and following up gestational diabetes mellitus according to claim 1, characterized in that: The feature extraction module comprises: Data feature extraction module, used for: A variety of feature extraction methods are used to extract features from the preprocessed historical blood glucose data, wherein the feature extraction methods include mean feature extraction, median feature extraction, standard deviation feature extraction and correlation coefficient feature extraction; Data partitioning module, used to: Determine the division ratio, the division ratio of the training set and the test set is 7:3 or 8:2; According to the division ratio, extract data samples of corresponding proportion from the historical blood glucose data as training set and test set; The training set is used to train the blood glucose prediction model, and the test set is used to evaluate the performance of the blood glucose prediction model.

8. The device for tracking and following up gestational diabetes mellitus according to claim 7, characterized in that: The data feature extraction module is specifically: Mean feature extraction: Calculate the average blood sugar value at each time point of historical blood sugar data as the overall representative of blood sugar level; Median feature extraction: Calculate the median of blood sugar at each time point in historical blood sugar data to reflect the changing trend of blood sugar levels; Standard deviation feature extraction: Calculate the standard deviation of blood sugar at each time point in historical blood sugar data to reflect the degree of dispersion of blood sugar levels; Correlation coefficient feature extraction: Calculate the correlation coefficient of blood sugar at adjacent time points of historical blood sugar data to reflect the synchronization and continuity of blood sugar changes.

9. The device for tracking and following up gestational diabetes mellitus according to claim 1, characterized in that: The model prediction module comprises: Model building modules for: Based on the historical blood sugar data processed by the data acquisition module, a blood sugar prediction model is constructed using deep learning; Among them, deep learning uses convolutional neural network model or recurrent neural network model to build blood glucose prediction model; Model training module, used to: Set model parameters, including model size, learning rate, and regularization parameters; The blood glucose prediction model is trained using the data from the training set, and the model parameters are continuously updated during the training process; Model optimization module for: Use the test set data to evaluate the performance of the blood glucose prediction model, including the accuracy, precision and recall of the blood glucose prediction model; If the performance of the blood glucose prediction model is poor, the blood glucose prediction model is optimized, including adding or deleting features and adjusting parameters; Model deployment module, used to: The trained and optimized blood glucose prediction model is deployed in practical applications. The blood glucose data collected during follow-up are predicted by the blood glucose prediction model. The blood glucose status of pregnant women is understood based on the prediction results, and the potential blood glucose risks of pregnant women are discovered.

10. The device for tracking and following up gestational diabetes mellitus according to claim 1, characterized in that: The auxiliary decision module comprises: The result display module is used to: Obtain prediction results from the model prediction module and convert the prediction results into charts or graphs for display, wherein the charts include bar charts, pie charts and line charts, and the graphs include bar charts, pie charts, heat maps and scatter plots; User interface module for: providing a user interface through which a physician accesses and views the displayed prediction results; Solution recommendation module, used for: The doctor will assess the condition, diagnose and formulate a follow-up treatment plan based on the predicted results of the visit and review, and recommend the formulated treatment plan to the pregnant woman via text message or email.

Images