Intelligent prediction system and method for gestational diabetes mellitus

Abstract

The present invention discloses an intelligent prediction system and method for gestational diabetes, and relates to the field of medical artificial intelligence technology. The system includes a collection module for collecting pregnancy index data, preprocessing the index data, and extracting the preprocessed data features; an evaluation module for performing synchronization and rhythmicity analysis based on the extracted features, establishing a physiological cycle correlation model, and evaluating the synchronization and disorder of physiological rhythms; a prediction module for constructing a risk prediction model based on the analysis results of the physiological cycle correlation model to predict the risk level of physiological status. The present invention significantly improves the early recognition rate of gestational diabetes, helps to reduce the incidence of the disease and the risk of complications, reduces the consumption of medical resources, and has important clinical application value and social benefits. And parameter optimization and model adjustment can be performed according to the characteristics of different populations, providing strong support for personalized medicine and precision prevention.

Description

Technical Field

[0001] The present invention relates to the field of medical artificial intelligence technology, and in particular to an intelligent prediction system and method for gestational diabetes mellitus. Background Art

[0002] Gestational diabetes mellitus (GDM) is one of the most common metabolic diseases during pregnancy, with its incidence showing an increasing trend annually. According to research, the global incidence of GDM is approximately 5% to 15%, reaching as high as 17.5% in parts of my country. This disease not only severely impacts maternal health but can also lead to complications such as fetal developmental abnormalities, macrosomia, and neonatal hypoglycemia, placing a heavy burden on families and society. Currently, the prediction and diagnosis of GDM primarily utilize traditional risk factor assessment methods, single biochemical marker monitoring methods, machine learning prediction models, and continuous blood glucose monitoring systems. Traditional risk factor assessment methods primarily rely on static risk factors such as age, BMI, and family history, but these methods offer low predictive accuracy, are unable to conduct dynamic risk assessments, and fail to fully account for individual differences. Single physiological marker monitoring methods primarily rely on fasting and postprandial blood glucose levels for prediction, but these results are susceptible to multiple factors, lack an overall assessment of systemic status, and provide delayed warnings, making early intervention difficult. While machine learning prediction models can comprehensively analyze multiple risk factors, they require high data volume and quality, have poor model interpretability, and are computationally complex. Continuous glucose monitoring systems use implantable sensors to monitor blood glucose levels in real time, but they are invasive, costly, and require extensive maintenance. Furthermore, they focus solely on blood glucose levels, ignoring other important physiological indicators.

[0003] Existing technologies generally have problems such as insufficient predictive performance, large application limitations, and obvious technical bottlenecks. Specifically, they are weak early warning capabilities, high rates of missed diagnosis and misdiagnosis, low degree of individualization, complex operations, high costs, and poor compliance. At the same time, due to technical bottlenecks such as the limitations of a single indicator, insufficient system evaluation capabilities, and difficulties in dynamic monitoring, the accuracy and practicality of existing prediction models need to be improved. In clinical applications, problems such as the difficulty of real-time monitoring, insufficient early intervention guidance, lack of personalized plans, and poor long-term management effects also need to be urgently addressed. Therefore, there is an urgent need for a new prediction method that can achieve early warning of gestational diabetes, provide accurate risk assessment, support personalized intervention, and have good clinical practicality. Summary of the Invention

[0004] In view of the problems existing in the existing intelligent prediction systems and methods for gestational diabetes, the present invention proposes an intelligent prediction system and method for gestational diabetes.

[0005] Therefore, the problem to be solved by the present invention is how to achieve early warning of gestational diabetes, provide accurate risk assessment, and achieve good clinical practicality.

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

[0007] In a first aspect, an embodiment of the present invention provides an intelligent prediction system for gestational diabetes, which includes a collection module for collecting gestational indicator data, preprocessing the indicator data, and extracting data features after preprocessing; an evaluation module for extracting time domain features and frequency domain features based on the data features to perform synchronization and rhythmicity analysis, quantify the synchronization between multiple physiological signals through cross-correlation, establish a physiological cycle correlation model, and evaluate the synchronization and disorder of physiological rhythms; a prediction module for analyzing relevant risk factors based on the analysis results of the physiological cycle correlation model, constructing a risk prediction model, and predicting the risk level of the physiological state; the relevant risk factors include age-standardized score, pre-pregnancy BMI-standardized score, fasting blood glucose-standardized score, diabetes family history indicator, previous pregnancy history risk score, lifestyle score, gestational weight gain rate-standardized score, and blood pressure-standardized score.

[0008] As a preferred solution of the intelligent prediction system for gestational diabetes described in the present invention, the collection module collects pregnancy indicator data, performs preliminary preprocessing on the data, obtains a high-quality preprocessed data base, and extracts data features; the evaluation module analyzes the time domain features and frequency domain features of the data based on the extracted data features, constructs a physiological cycle correlation model, evaluates the synchronization and disorder degree of physiological rhythms, and determines whether the individual's physiological rhythm is within the normal range; the prediction module constructs a risk prediction model based on the results of the physiological cycle correlation model, predicts the physiological state risk level, predicts the health status, and feeds back the health risk prediction results to the collection module to adjust the focus of future data collection.

[0009] In a second aspect, an embodiment of the present invention provides an intelligent prediction method for gestational diabetes, which includes: collecting pregnancy indicator data, preprocessing the indicator data, and extracting data features after preprocessing; performing synchronization and rhythmicity analysis based on the extracted data features, establishing a physiological cycle correlation model, and evaluating the synchronization and disorder of physiological rhythms; based on the analysis results of the physiological cycle correlation model, constructing a risk prediction model to predict the risk level of the physiological state.

[0010] As a preferred embodiment of the intelligent prediction method for gestational diabetes described in the present invention, the indicator data includes sleep quality data, hormonal cycle data, and circadian rhythm data; the sleep quality data includes sleep duration, sleep onset time, sleep efficiency, and sleep stage data; the sleep stage data is divided into deep sleep, light sleep, and REM stage; the hormonal cycle data includes basal body temperature, estrogen level, progesterone level, and gonadotropin level; and the circadian rhythm data includes heart rate variability, cortisol level, melatonin secretion curve, and activity level changes.

[0011] As a preferred solution of the intelligent prediction method for gestational diabetes mellitus of the present invention, wherein: preprocessing is performed based on the index data, including outlier processing, missing value supplementation, data standardization and time series alignment; the outlier processing detects and identifies unreasonable values ​​in a single data stream by establishing association rules or multimodal data fusion models; the collected time series is set as heart rate data and activity data using the Pearson correlation coefficient Calculate the correlation between heart rate and activity level. When the correlation coefficient is lower than the set threshold, it is considered abnormal, expressed as:

[0012] ;

[0013] in, represents the length of the observation window, Indicates time Heart rate, For time The amount of activity during is the mean of the heart rate in the window, is the mean of the activity within the window; the missing values ​​are supplemented by predicting the possible values ​​of the missing data through other known data sources, and linear interpolation is used to infer the values ​​of the missing data, which is expressed as:

[0014] ;

[0015] in, and is the missing value level of adjacent valid data points, The time series alignment adopts the multi-scale time window method, which is to use a small window for high-frequency data and a large window for low-frequency data, and align them using the average value or interpolation method in each window, and set the window length to and ,but:

[0016] ;

[0017] ;

[0018] in, is the window length of high-frequency data, is the window length of low-frequency data, and are the average and interpolation values ​​within their respective windows, and Respectively represent heart rate data and activity data at time and The value at , , Indicates the number of sampling points within the high-frequency data window length, Indicates the number of sampling points within the low-frequency data window length.

[0019] As a preferred embodiment of the intelligent prediction method for gestational diabetes mellitus of the present invention, the construction of the physiological cycle correlation model includes: extracting time domain features and frequency domain features based on pre-processed data features to identify physiological regularities and anomalies; the time domain features reflect the stability and volatility of physiological data by extracting the overall trend and dispersion of the mean and variance statistical data of the pre-processed data, and analyzing the time series. The mean calculation of is expressed as:

[0020] ;

[0021] in, Representing time series The mean of express The observed value at time, represents the total length of the time series; the variance is expressed as:

[0022] ;

[0023] in, Representing time series Variance; Based on the auto-correlation function, identify the periodic signal, analyze the periodicity in the data, obtain the periodic characteristics, for time series , the autocorrelation function is expressed as:

[0024] ;

[0025] in, is the autocorrelation function, which indicates the number of time series lags The autocorrelation under represents the number of time lag steps, Represents the relative position in the time series Backward lag The frequency domain features are based on the data of time domain features, and the frequency spectrum information of the data is analyzed by Fourier transform to clarify the distribution of different frequency components in the physiological signal. , whose discrete Fourier transform Expressed as:

[0026] ;

[0027] in, express The observed value at time, represents a frequency variable, Indicates the total length of the time series data, is a complex exponential term, where is an imaginary number; based on the Fourier transform, the power spectrum density is obtained , expressed as:

[0028] ;

[0029] in, Indicates the signal frequency The amplitude of the lower Fourier transform result, Frequency The corresponding power.

[0030] As a preferred embodiment of the intelligent prediction method for gestational diabetes mellitus of the present invention, a physiological cycle correlation model is constructed based on the data characteristics to evaluate the synchronization and disorder of physiological rhythms, which is expressed as:

[0031] ;

[0032] in, is the physiological cycle synchronization assessment index, is the total length of the observation period, is a time variable, indicating the observation time point; Represents the instantaneous phase difference function, which is used to calculate the time difference between two physiological signals The absolute value of the phase difference at the moment; is the phase difference attenuation coefficient; represents the normalized cross-correlation variance, reflecting the stability of signal synchronization; represents the mutual information entropy, Represents the mutual information entropy benchmark value; sets the physiological cycle synchronization evaluation index The threshold value is determined according to the output of the model, and the threshold range is determined to analyze and evaluate the synchronization and disorder of physiological rhythms.

[0033] As a preferred embodiment of the intelligent prediction method for gestational diabetes mellitus of the present invention, a risk prediction model is constructed based on the physiological cycle synchronization assessment index, which is expressed as:

[0034] ;

[0035] in, is the risk prediction index, is the risk benchmark coefficient used to standardize the risk value, is the physiological cycle synchronization assessment index, For the The weight coefficient of each risk factor, For the Risk factor values, , for The 24-hour change in value, is the fluctuation sensitivity coefficient, for are the maximum and minimum values ​​within the observation period, respectively. value, is the average value during the observation period value, is the number of risk factors considered; according to the results of the prediction model, it is divided into the first threshold, the second threshold and the third threshold, and the first threshold < the second threshold < the third threshold; if the risk prediction index ≤ the first threshold, indicating the physiological cycle synchronization assessment index In the highest range, If the diurnal fluctuation amplitude is within the baseline fluctuation range and all risk factor scores are within the normal reference range, the basic physiological index monitoring frequency should be maintained and basic health data should be recorded regularly; if the first threshold < risk prediction index <The second threshold value indicates the physiological cycle synchronization assessment index Below the highest range but above the warning range, If the fluctuation range exceeds the baseline range but does not reach the warning line, and any risk factor score exceeds the normal reference range but does not reach the warning value, the blood glucose monitoring program will be strengthened, the monitoring time points will be increased, and the health monitoring frequency will be increased; if the second threshold is ≤ the risk prediction index <The third threshold value indicates the physiological cycle synchronization assessment index Falling to the warning range, If the fluctuation amplitude reaches the warning line, multiple risk factor scores reach the warning value, and the risk index exceeds the fluctuation threshold in a short period of time, the blood glucose monitoring frequency will be increased to the basic frequency. times, implement a strict diet control program; if the third threshold ≤ risk prediction index , represents the physiological cycle synchronization assessment index Drop to the lowest range, If the fluctuation range exceeds the warning line, the key risk factors exceed the alert value, and the system stability drops to the lowest level, all-weather monitoring will be implemented, a real-time warning mechanism will be established, and all-round health management will be implemented.

[0036] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned intelligent prediction method for gestational diabetes is implemented.

[0037] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned intelligent prediction method for gestational diabetes is implemented.

[0038] The beneficial effects of the present invention include: It enables intelligent prediction of gestational diabetes risk, demonstrating significant technological innovation and practical application value. It not only provides a new approach to disease prediction but also establishes a comprehensive risk management system, which is of great significance for improving the prevention and control of gestational diabetes. It can significantly increase the early identification rate of gestational diabetes, help reduce the incidence of the disease and the risk of complications, and reduce the consumption of medical resources, thus possessing significant clinical and social benefits. Furthermore, it can optimize parameters and adjust models based on the characteristics of different populations, providing strong support for personalized medicine and precision prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0040] Figure 1 This is a flow chart of the intelligent prediction method for gestational diabetes mellitus. DETAILED DESCRIPTION

[0041] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0042] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0044] Example 1: Reference Figure 1 , which is the first embodiment of the present invention, provides an intelligent prediction method for gestational diabetes, comprising:

[0045] S1: Collecting pregnancy indicator data, preprocessing the indicator data, and extracting features of the preprocessed data.

[0046] Collects a variety of physiological and behavioral data from users to obtain comprehensive health indicators. The collected data includes: sleep quality data, hormonal cycle data, and circadian rhythm data.

[0047] By monitoring heart rate and body movement frequency, it can infer sleep status and stages, perceive pressure changes, identify changes in body position and number of body movements, and further assist in sleep quality analysis to obtain sleep quality data.

[0048] Sleep quality data includes sleep duration, time to fall asleep, sleep efficiency and sleep stage data; sleep stage data is divided into deep sleep, light sleep and REM stage (rapid phase sleep).

[0049] By recording the basal body temperature curve daily, capturing temperature changes with the cycle, regularly testing hormone levels such as estrogen and progesterone, providing accurate menstrual cycle data, and monitoring hormone levels in urine as auxiliary information for menstrual cycle changes, hormone cycle data can be obtained.

[0050] Hormonal cycle data includes basal body temperature, estrogen levels, progesterone levels, gonadotropin levels, etc. to cover the different hormonal fluctuations in the menstrual cycle.

[0051] Heart rate variability data can be used to reveal the activity of the sympathetic / parasympathetic nerves and detect fluctuations in cortisol and melatonin levels to understand the physiological manifestations of circadian rhythms and obtain circadian rhythm data.

[0052] Circadian rhythm data include heart rate variability (HRV), cortisol levels, melatonin secretion curve, activity level changes, etc. to describe the circadian rhythm situation.

[0053] Preprocessing is performed on the collected raw data to ensure data consistency and high quality. The preprocessing includes outlier processing, missing value filling, data standardization, and time series alignment.

[0054] Outlier processing detects and identifies unreasonable values ​​in a single data stream by establishing association rules or multimodal data fusion models between multiple physiological signal acquisition sources.

[0055] Assuming the collected time series is heart rate data and activity data, a correlation formula is constructed between the two to detect abnormalities, using the Pearson correlation coefficient. Calculate the correlation between heart rate and activity level. When the correlation coefficient is lower than the set threshold, it is considered abnormal, expressed as:

[0056] ;

[0057] in, represents the length of the observation window, Indicates time Heart rate, For time The amount of activity during is the mean of the heart rate in the window, is the mean activity within the window.

[0058] like Below the set threshold , it is preliminarily determined that there is an anomaly between heart rate and activity level, marked as an abnormal point, and then the abnormal data is eliminated.

[0059] Furthermore, when there are missing data from multiple sources, the possible values ​​of the missing data are predicted by other known data sources to maintain data integrity. For missing physiological data, such as cortisol levels, linear interpolation is used to infer the missing values, which is expressed as:

[0060] ;

[0061] in, and is the cortisol level of the adjacent valid data point, is the interpolated cortisol value.

[0062] Furthermore, during the real-time data normalization process, a sliding window is used to update the data mean and standard deviation in real time to eliminate the influence of long-term trends and sudden changes. When the window length is If the standardization is performed within a time period, the standardization formula is expressed as:

[0063] ;

[0064] in, represents the normalized heart rate data, Indicates time The mean heart rate in the sliding window at time , Indicates time The standard deviation of heart rate within the sliding window at time .

[0065] Further,

[0066] ;

[0067] ;

[0068] in, is the length of the observation window, Indicates heart rate data at time The value at .

[0069] Furthermore, for data sources with different frequencies, the multi-scale time window method is used to retain the details of high-frequency data while reducing the computational burden of low-frequency data. (high-frequency sampling) and activity data (Low frequency sampling) Align by different windows. During the alignment process, a small window is used for high frequency data and a large window is used for low frequency data. The average value or interpolation method is used to align each window. Set the window length to and ,but:

[0070] ;

[0071] ;

[0072] in, is the window length of high-frequency data, is the window length of low-frequency data, and are the average and interpolation values ​​within their respective windows, and Respectively represent heart rate data and activity data at time and The value at , , Indicates the number of sampling points within the high-frequency data window length, Indicates the number of sampling points within the low-frequency data window length.

[0073] S2: Based on the preprocessed data features, time domain features and frequency domain features are extracted to perform synchronization and rhythmicity analysis. The synchronization between multiple physiological signals is quantified through cross-correlation, and a physiological cycle correlation model is established to evaluate the synchronization and disorder of physiological rhythms.

[0074] Feature extraction based on preprocessed data helps to identify the time domain features and frequency domain features of physiological laws and abnormalities.

[0075] The time domain features reflect the stability and volatility of physiological data by extracting the mean and variance of the preprocessed data, the overall trend and discreteness of the statistical data, revealing the normal or abnormal fluctuations of physiological signals within a specific time range, and preliminarily judging the periodicity of the data by identifying the lag steps of the autocorrelation function.

[0076] Specifically, the mean of the time domain feature represents the central trend of the data. The mean calculation of is expressed as:

[0077] ;

[0078] in, Representing time series The mean of express The observed value at time, Indicates the total length of the time series.

[0079] Variance is used to measure data volatility and reflects the degree of dispersion of data, which is expressed as:

[0080] ;

[0081] in, Representing time series The variance of .

[0082] Furthermore, the periodic signal is identified through the autocorrelation function, the periodic pattern in the data is analyzed, and the periodic characteristics are obtained. , the autocorrelation function is expressed as:

[0083] ;

[0084] in, is the autocorrelation function, which indicates the number of time series lags The autocorrelation under represents the number of time lag steps, Represents the relative position in the time series Backward lag The data value at each time point is obtained; the periodicity of the data is preliminarily judged by finding the number of lag steps after the peak appears, and the time domain characteristics of the data are obtained.

[0085] Frequency domain features are used to reveal the frequency components of signals and analyze physiological rhythms. Based on time domain features, frequency domain features are extracted and further analyzed using Fourier transform to analyze the spectrum information of data.

[0086] Fourier transform converts the time domain signal into the frequency domain to clarify the distribution of different frequency components in the physiological signal. , whose discrete Fourier transform Expressed as:

[0087] ;

[0088] in, express The observed value at time, represents a frequency variable, Indicates the total length of the time series data, is a complex exponential term, where Is an imaginary number. The complex exponential term is used to convert the time domain signal into the frequency domain by calculating the signal value at all times. Perform weighted summation to construct the spectrum.

[0089] The power spectrum density PSD is used to describe the power distribution of the signal at different frequencies, and the contribution of different frequency components to the signal energy is determined. The power spectrum density is obtained by Fourier transform. , expressed as:

[0090] ;

[0091] in, Indicates frequency The energy density of the signal at Indicates the signal frequency The amplitude of the lower Fourier transform result, Frequency The corresponding power is obtained by analyzing the power spectral density , determine significant frequencies in the data and thus identify underlying physiological rhythms.

[0092] Based on the extraction results of time domain features and frequency domain features, physiological cycle synchronization analysis is performed. The purpose is to evaluate the coordination and synchronization of different physiological rhythms through the correlation and phase relationship of multidimensional physiological signals. The synchronization between multiple physiological signals is quantified by cross-correlation, and the lag effect between two signals is evaluated by cross-correlation analysis. Given two signals and , then the cross-correlation function is expressed as:

[0093] ;

[0094] in, Indicates the time lag step. By calculating different time lag steps The correlation under the condition of lag is used to determine the synchronization and lag relationship between the two signals.

[0095] With the data support provided by time domain and frequency domain features, physiological cycle synchronization analysis can more accurately focus on the periodic changes of each signal, find correlations, and evaluate the coordination of physiological rhythms.

[0096] Based on the above evaluation and analysis, a physiological cycle correlation model was constructed to evaluate the synchronization and disorder of physiological rhythms, which is expressed as:

[0097] ;

[0098] in, is the physiological cycle synchronization assessment index, is the total length of the observation period, is a time variable, indicating the observation time point; Represents the instantaneous phase difference function, which is used to calculate the time difference between two physiological signals The absolute value of the phase difference at the moment; is the phase difference attenuation coefficient; represents the normalized cross-correlation variance, reflecting the stability of signal synchronization; Represents mutual information entropy, which is used to quantify the degree of mutual dependence between two signals. Represents the mutual information entropy benchmark value.

[0099] Setting the menstrual cycle synchronization assessment index The threshold value is determined according to the output of the model, and the threshold range is determined to analyze and evaluate the synchronization and disorder of physiological rhythms.

[0100] S3: Based on the analysis results of the physiological cycle correlation model, analyze relevant risk factors, build a risk prediction model, and predict the risk level of physiological status.

[0101] According to the physiological cycle synchronization assessment index The feature analysis found The smaller the value, the higher the degree of system disorder, and the degree of disorder is positively correlated with the risk of disease, so (1- ) as the basic risk measure.

[0102] When physiological disorders occur, they are often associated with multiple risk factors. We then further analyze the relevant risk factors and integrate them into the prediction model to obtain a risk prediction model, which is expressed as:

[0103] ;

[0104] in, is the risk prediction index, is the risk benchmark coefficient, which is used to standardize the risk value. is the physiological cycle synchronization assessment index, For the The weight coefficient of each risk factor, For the Risk factor values, , for The 24-hour change in value, is the fluctuation sensitivity coefficient, and are the maximum and minimum values ​​within the observation period, respectively. value, is the average value during the observation period value, is the number of risk factors considered.

[0105] Furthermore, risk factors Including: age-standardized score, pre-pregnancy BMI-standardized score, fasting blood glucose-standardized score, diabetes family history index, previous pregnancy history risk score, lifestyle score, gestational weight gain-standardized score, and blood pressure-standardized score.

[0106] According to the result of the prediction model, it is divided into a first threshold, a second threshold and a third threshold, and the first threshold < the second threshold < the third threshold.

[0107] If the risk prediction index ≤ the first threshold, indicating the physiological cycle synchronization assessment index In the highest range, If the diurnal fluctuation amplitude is within the baseline fluctuation range and all risk factor scores are within the normal reference range, maintain the monitoring frequency of basic physiological indicators and record basic health data regularly.

[0108] If the first threshold < risk prediction index <The second threshold value indicates the physiological cycle synchronization assessment index Below the highest range but above the warning range, If the fluctuation range exceeds the baseline range but does not reach the warning line, and any risk factor score exceeds the normal reference range but does not reach the warning value, the frequency of prenatal examinations will be increased to twice the basic frequency, the blood glucose monitoring plan will be strengthened, the monitoring time points will be increased, and the frequency of health monitoring will be increased.

[0109] If the second threshold ≤ risk prediction index <The third threshold value indicates the physiological cycle synchronization assessment index Falling to the warning range, If the fluctuation range reaches the warning line, the scores of multiple risk factors reach the warning value, and the risk index exceeds the fluctuation threshold in the short term, the follow-up frequency will be increased to 4 times the basic frequency, the blood glucose monitoring frequency will be increased to 4 times the basic frequency, a strict diet control plan will be implemented, and a comprehensive health management system will be established.

[0110] If the third threshold ≤ risk prediction index , represents the physiological cycle synchronization assessment index Drop to the lowest range, If the fluctuation range exceeds the warning line, the key risk factors exceed the alert value, and the system stability drops to the lowest level, round-the-clock monitoring will be implemented, a real-time early warning mechanism will be established, and drug intervention will be considered to implement all-round health management.

[0111] Further, the judgment of drug intervention effect evaluation, including risk prediction index Whether it falls back to a lower range, whether the synchronization of physiological cycles returns to a higher level, whether the system stability is improved to an ideal range, whether various risk factors return to the normal reference range, and whether physiological indicators remain in a stable range.

[0112] In summary, this invention enables intelligent prediction of gestational diabetes risk, demonstrating significant technological innovation and practical application value. It not only provides a new approach to disease prediction but also establishes a comprehensive risk management system, which is of great significance for improving the prevention and control of gestational diabetes. It can significantly improve the early identification rate of gestational diabetes, help reduce the incidence of the disease and the risk of complications, and reduce the consumption of medical resources, thus possessing important clinical application value and social benefits. Furthermore, it can optimize parameters and adjust models based on the characteristics of different populations, providing strong support for personalized medicine and precision prevention.

[0113] This embodiment further provides an intelligent prediction system for gestational diabetes mellitus, which applies the intelligent prediction method for gestational diabetes mellitus described in the above embodiment, including:

[0114] A collection module is used to collect pregnancy indicator data, preprocess the indicator data, and extract features of the preprocessed data;

[0115] An evaluation module is used to extract time domain features and frequency domain features based on the data features to perform synchronization and rhythmicity analysis, quantify the synchronization between multiple physiological signals through cross-correlation, establish a physiological cycle correlation model, and evaluate the synchronization and disorder of physiological rhythms;

[0116] The prediction module is configured to analyze relevant risk factors based on the analysis results of the physiological cycle correlation model, construct a risk prediction model, and predict the physiological state risk level. This embodiment also provides a computer device comprising a memory and a processor; the memory is configured to store computer-executable instructions, and the processor is configured to execute the computer-executable instructions to implement the intelligent prediction method for gestational diabetes as proposed in the above embodiment.

[0117] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0118] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for intelligently predicting gestational diabetes mellitus as proposed in the above embodiment is implemented.

[0119] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. Technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0120] Example 2: This example provides an intelligent prediction method for gestational diabetes. In order to verify the beneficial effects of the present invention, a scientific demonstration is conducted through simulation experiments.

[0121] In this example, 1000 pregnant women aged 25-35 years, with a gestational period ranging from 12 to 28 weeks, were randomly divided into an experimental group and a control group, with 500 cases in each group. The comparison method included:

[0122] Method A: The physiological cycle synchronization analysis method of the present invention;

[0123] Method B: Traditional risk factor assessment method;

[0124] Method C: Single physiological indicator monitoring method;

[0125] Method D: Machine learning prediction model.

[0126] Initial collection of experimental data, including:

[0127] Basic information: age, BMI, gestational age, family history, etc.;

[0128] Physiological indicators: heart rate, blood pressure, body temperature, activity level, etc.;

[0129] Metabolic indicators: fasting blood glucose, postprandial blood glucose, glycosylated hemoglobin, etc.;

[0130] Monitoring period: 12 consecutive weeks;

[0131] Sampling frequency: Physiological indicators once every 5 minutes.

[0132] The entire trial cycle is divided into three stages:

[0133] The first phase is data collection (12 weeks), which involves establishing a standardized data collection process, synchronously collecting data on various indicators, recording clinical diagnosis results, and implementing quality control measures.

[0134] The second stage is model application (4 weeks), which involves applying four forecasting methods, recording forecast results, compiling early warning statistics, and evaluating forecast effectiveness.

[0135] The third phase is follow-up verification (24 weeks), tracking pregnancy outcomes, recording morbidity, collecting complication data, and evaluating prediction accuracy.

[0136] A preliminary prediction is made for the second stage, and the prediction results are shown in Table 1:

[0137] Table 1 Comparison of prediction performance indicators

[0138] Evaluation Metrics Method (A) of the present invention Traditional scoring method (B) Single indicator method (C) Machine Learning (D) Sensitivity 92.5% 78.4% 71.2% 85.6% Specificity 89.3% 75.6% 73.8% 83.2% ROC area 0.91 0.77 0.72 0.84 Early identification lead time 4-6 weeks 2-3 weeks 1-2 weeks 3-4 weeks Measurement accuracy 90.8% 76.5% 72.4% 84.3%

[0139] According to the prediction results, after 12 months of follow-up observation, the experimental results are shown in Table 2:

[0140] Table 2 Comparison of clinical follow-up results

[0141] Evaluation Metrics Method (A) of the present invention Traditional scoring method (B) Single indicator method (C) Machine Learning (D) Incidence rate in high-risk groups 85.6% 65.3% 58.9% 78.4% Early intervention is effective 82.3% 60.8% 55.4% 75.2% Complication rate 8.5% 18.6% 22.4% 12.8% Follow-up completion rate 95.8% 85.4% 88.6% 90.2%

[0142] Analysis of the predictive performance indicators in Table 1 demonstrates that the method of the present invention exhibits significant performance advantages over the other three methods, with a sensitivity of 92.5%, a specificity of 89.3%, and a receiver operating characteristic (ROC) area of ​​0.91. It also enables early identification 4 to 6 weeks in advance, with a prediction accuracy of 90.8%. These data demonstrate that the method of the present invention exhibits significant advantages in key performance indicators such as predictive ability, accuracy, and early warning time. In particular, compared to traditional scoring methods and single-indicator methods, the performance improvement reaches 15 to 20 percentage points.

[0143] In terms of clinical follow-up results, the data in Table 2 show that the method of the present invention exhibited excellent clinical application effects during the 12-month follow-up period. The actual incidence rate of the high-risk population reached 85.6%, confirming the accuracy of the prediction; the early intervention efficacy reached 82.3%, significantly higher than that of other methods; the complication rate was only 8.5%, 10-14 percentage points lower than that of other methods; and the follow-up completion rate reached 95.8%. These data fully reflect that the method of the present invention has significant practical value and good application effects in actual clinical applications, which not only improves the prediction accuracy, but also improves the clinical intervention effect, while optimizing the efficiency of medical resource utilization.

[0144] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. The intelligent prediction system for gestational diabetes mellitus is characterized by: include: A collection module, used to collect pregnancy index data, pre-process the index data, and extract features of the pre-processed data; An evaluation module, for extracting time domain features and frequency domain features according to the data features to perform synchronization and rhythmicity analysis, quantifying the synchronization between multiple physiological signals through cross-correlation, establishing a physiological cycle correlation model, and evaluating the synchronization and disorder of physiological rhythms; A prediction module, used to analyze relevant risk factors according to the analysis results of the physiological cycle correlation model, build a risk prediction model, and predict the risk level of the physiological state; The relevant risk factors include age-standardized score, pre-pregnancy BMI-standardized score, fasting blood glucose-standardized score, diabetes family history index, previous pregnancy history risk score, lifestyle score, gestational weight gain rate-standardized score, and blood pressure-standardized score; The indicator data includes sleep quality data, hormone cycle data and circadian rhythm data; The sleep quality data includes sleep duration, sleep onset time, sleep efficiency and sleep stage data; the sleep stage data is divided into deep sleep, light sleep and REM stage; The hormonal cycle data include basal body temperature, estrogen level, progesterone level and gonadotropin level; The circadian rhythm data include heart rate variability, cortisol level, melatonin secretion curve and activity level changes; Based on the data features, a physiological cycle correlation model is constructed to evaluate the synchronization and disorder of physiological rhythms, which is expressed as: ； in, is the physiological cycle synchronization assessment index, is the total length of the observation period, is a time variable, indicating the observation time point; Represents the instantaneous phase difference function, which is used to calculate the time difference between two physiological signals The absolute value of the phase difference at the moment; is the phase difference attenuation coefficient; represents the normalized cross-correlation variance, reflecting the stability of signal synchronization; represents the mutual information entropy, Represents the mutual information entropy benchmark value; Setting the menstrual cycle synchronization assessment index The threshold value is determined according to the output of the model, and the threshold range is determined to analyze and evaluate the synchronization and disorder of physiological rhythms; A risk prediction model is constructed based on the physiological cycle synchronization evaluation index, which is expressed as: ； in, is the risk prediction index, is the risk benchmark coefficient, which is used to standardize the risk value. is the physiological cycle synchronization assessment index, For the The weight coefficient of each risk factor is For the Risk factor values, , for The 24-hour change in value, is the fluctuation sensitivity coefficient, and are the maximum and minimum values ​​within the observation period, respectively. value, is the average value within the observation period value, is the number of risk factors considered; According to the result of the prediction model, it is divided into a first threshold, a second threshold and a third threshold, and the first threshold < the second threshold < the third threshold; If the risk prediction index ≤ the first threshold, indicating the physiological cycle synchronization assessment index Located in the highest range, If the diurnal fluctuation range is within the baseline fluctuation range and all risk factor scores are within the normal reference range, the basic physiological index monitoring frequency should be maintained and basic health data should be recorded regularly; If the first threshold < risk prediction index <The second threshold, indicating the physiological cycle synchronization assessment index Below the highest range but above the warning range, If the fluctuation range exceeds the baseline range but does not reach the warning line, and any risk factor score exceeds the normal reference range but does not reach the warning value, the blood sugar monitoring plan will be strengthened, the monitoring time points will be increased, and the frequency of health monitoring will be increased; If the second threshold ≤ risk prediction index <The third threshold, indicating the physiological cycle synchronization assessment index Down to the warning range, If the fluctuation amplitude reaches the warning line, the scores of multiple risk factors reach the warning value, and the risk index exceeds the fluctuation threshold in a short period of time, the blood glucose monitoring frequency will be increased to the basic frequency. times, implement a strict diet control program; If the third threshold ≤ risk prediction index , represents the physiological cycle synchronization assessment index To the lowest range, If the fluctuation range exceeds the warning line, the key risk factor exceeds the alert value, and the system stability drops to the lowest level, we will implement round-the-clock monitoring, establish a real-time early warning mechanism, and implement all-round health management.

2. The intelligent prediction system for gestational diabetes mellitus according to claim 1, characterized in that: Based on the collection module, the pregnancy index data are collected, and the data are preliminarily preprocessed to obtain a high-quality preprocessed data basis and extract data features; The evaluation module analyzes the time domain characteristics and frequency domain characteristics of the data according to the extracted data characteristics, constructs a physiological cycle correlation model, evaluates the synchronization and disorder degree of physiological rhythm, and determines whether the individual's physiological rhythm is within the normal range. The prediction module constructs a risk prediction model according to the results of the physiological cycle correlation model, predicts the risk level of the physiological state, predicts the health status, and feeds back the health risk prediction results to the collection module to adjust the focus of future data collection.

3. An intelligent prediction method for gestational diabetes, based on the intelligent prediction system for gestational diabetes according to claim 1, characterized in that: include: Collecting pregnancy indicator data, preprocessing the indicator data, and extracting features of the preprocessed data; Based on the extracted data features, synchronization and rhythmicity analysis are performed, a physiological cycle correlation model is established, and the synchronization and disorder of physiological rhythms are evaluated; Based on the analysis results of the physiological cycle correlation model, a risk prediction model is constructed to predict the risk level of the physiological state.

4. The intelligent prediction method for gestational diabetes mellitus according to claim 3, characterized in that: Preprocessing is performed based on the indicator data, including outlier processing, missing value supplementation, data standardization, and time series alignment; The outlier processing is to detect and identify unreasonable values ​​in a single data stream by establishing association rules or multimodal data fusion models; the collected time series is set as heart rate data and activity data, and the Pearson correlation coefficient is used. The correlation between heart rate and activity is calculated. When the correlation coefficient is lower than the set threshold, it is considered abnormal, expressed as: ； in, represents the length of the observation window, Indicates time Heart rate at For time The amount of activity during is the mean of the heart rate in the window, is the mean of the activity in the window; The missing value supplement predicts the possible value of the missing data through other known data sources, and uses linear interpolation to infer the value of the missing data, which is expressed as: ； in, and is the missing value level of adjacent valid data points, is the missing value after interpolation; The time series alignment adopts the multi-scale time window method. The multi-scale time window method uses a small window for high-frequency data and a large window for low-frequency data, and uses the average value or interpolation method to align in each window. The window length is set to and ,but: ； ； in, is the window length of high-frequency data, is the window length of low-frequency data, and are the average and interpolation values ​​in their respective windows, and Respectively represent heart rate data and activity data at time and The value at , , Represents the number of sampling points within the high-frequency data window length, Indicates the number of sampling points within the low-frequency data window length.

5. The intelligent prediction method for gestational diabetes mellitus according to claim 4, characterized in that: The construction of the physiological cycle correlation model includes: extracting time domain features and frequency domain features for identifying physiological regularities and abnormalities based on preprocessed data features; The time domain features reflect the stability and volatility of physiological data by extracting the mean and variance of the preprocessed data, the overall trend and dispersion of the statistical data, and the time series The mean calculation of is expressed as: ； in, Representing time series The mean of express The observed value at time, Indicates the total length of the time series; The variance is expressed as: ； in, Representing time series The variance of Based on the autocorrelation function, periodic signals are identified, periodic patterns in data are analyzed, and periodic features are obtained. , the autocorrelation function is expressed as: ； in, is the autocorrelation function, which indicates the number of time series lags in time The autocorrelation under represents the number of time lag steps, Represents the relative position in the time series Backward lag The data value at a time point; The frequency domain feature is based on the data of the time domain feature, and uses Fourier transform to analyze the spectrum information of the data to clarify the distribution of different frequency components in the physiological signal. Its discrete Fourier transform It is expressed as: ； in, express The observed value at time, represents a frequency variable, Indicates the total length of the time series data, is a complex exponential term, where is an imaginary number; ； in, Indicates that the signal is at frequency The amplitude of the Fourier transform result is, For frequency The corresponding power.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent prediction method for gestational diabetes mellitus according to any one of claims 3 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent prediction method for gestational diabetes mellitus according to any one of claims 3 to 5 are implemented.

Images