A pregnancy test prompting method and system based on pregnancy reverse analysis

Abstract

The present application relates to the field of medical information technology, and particularly relates to a pregnancy test prompting method and system based on pregnancy resistance analysis; through collecting daily pregnancy resistance subjective score of pregnant women and objective sign data of wearable devices, a subjective and objective fusion feature is constructed and the constitution type of the pregnant women is identified, so that the physical examination prompting is no longer dependent on a fixed schedule, but is fitted to the actual physiological response and embryo activity change of the individual. A gestational age weight based on fetal arrest probability is introduced, and the sampling frequency is automatically encrypted in the high incidence period; at the same time, based on the necessary degree score of continuous multiple days of pregnancy test, the trend is determined, the early signal of fetal arrest can be captured, the early warning is given in the high-risk window period, and the clinical dilemma of not being discovered due to fetal arrest occurring between two timed pregnancy tests is effectively avoided. For low-risk pregnant women with stable pregnancy resistance, the subjective and objective data are consistent with the reference curve of the constitution, the system gives a lower necessary degree score, and the pregnancy test is not prompted frequently, so as to reduce unnecessary outpatient service occupation, ultrasonic examination and back-and-forth burden of the pregnant women.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically a method and system for pregnancy detection based on pregnancy reversal analysis. Background Technology

[0002] Currently, prenatal checkups for pregnant women generally follow a standardized schedule (such as registration at 6-13 weeks of pregnancy, NT scan at 11-13 weeks of pregnancy, and systematic ultrasound at 20-24 weeks of pregnancy), with intervals of about 4 weeks between checkups. This scheduled prenatal checkup model is designed based on the average risk level of the population, is simple to operate and easy to manage, and has played an important role in routine obstetric practice.

[0003] However, scheduled prenatal checkups have significant drawbacks: they assume that the embryonic development and risk progression are synchronized across all pregnant women, completely ignoring the vast differences in early pregnancy physiological responses among individual women. Specifically, early pregnancy symptoms such as nausea, vomiting, loss of appetite, and fatigue (collectively known as pregnancy reactions) are closely related to fluctuations in maternal human chorionic gonadotropin (HCG) levels, which can, to some extent, indicate embryonic activity. The severity of pregnancy reactions varies greatly among pregnant women with different constitutions: some women experience mild pregnancy reactions yet their fetuses are active, while others experience a sudden reduction in pregnancy reactions, which may be an early sign of miscarriage. However, scheduled prenatal checkups completely fail to consider the dynamic changes in pregnancy reactions as a basis for adjusting checkup schedules, resulting in high-risk individuals not receiving more intensive monitoring, while low-risk individuals may be subjected to unnecessary frequent examinations.

[0004] Fetal demise (embryonic arrest) is a common adverse outcome in early pregnancy. If not detected in time, the prolonged retention of the stillborn fetus in the uterine cavity can lead to serious complications such as intrauterine adhesions and coagulation disorders, causing both physical and psychological trauma to the pregnant woman. Since fetal demise can occur between any scheduled prenatal checkups, especially during the peak period of 6-8 weeks of pregnancy, the 4-week interval between checkups can easily create the predicament of "fetal demise for several weeks without the pregnant woman's knowledge." Clinically, there are numerous cases where pregnant women are only unexpectedly informed of fetal demise during their next ultrasound, having been unaware of it beforehand due to lingering pregnancy symptoms.

[0005] The existing system cannot provide early warnings based on daily pregnancy reaction data uploaded by pregnant women, nor can it dynamically assess the necessity of current prenatal checkups. In other words, the timing of prenatal checkups remains unchanged regardless of whether a pregnant woman's pregnancy reaction suddenly worsens, remains stable, or abnormally disappears. This rigid model essentially imposes population statistical patterns onto individuals, ignoring the reliance of medical decisions on real-time physiological data.

[0006] Therefore, there is an urgent need for a technical solution that can dynamically adjust the physical examination prompts based on the actual pregnancy symptoms of pregnant women, so that the timing of the physical examination can truly match the individual changes in embryonic activity, thereby reducing the risk of missed miscarriage and avoiding unnecessary waste of medical resources. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a pregnancy test prompt method and system based on pregnancy reversal analysis to solve the above-mentioned technical problems.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: The present invention provides a pregnancy detection method based on pregnancy reversal analysis, comprising the following steps: Multiple sets of vital sign data samples and pregnancy reaction data samples are obtained, and the vital sign data and pregnancy reaction data of the current object in multiple sampling time windows are obtained. The vital sign data samples and the vital sign data both include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples and the pregnancy reaction data both include the degree values ​​of various subjective evaluation pregnancy reaction parameters. The frequency of the sampling time windows is predetermined. Based on the multiple sets of vital sign data samples and pregnancy reaction data samples, feature extraction was performed to obtain typical pregnancy reaction characteristics of various pregnancy reaction constitutions. Among them, the typical pregnancy reaction characteristics include subjective and objective fusion characteristics of multiple pregnancy time windows. The vital signs data and pregnancy reaction data from multiple sampling time windows are fused to obtain the current fused features. The current fused features from multiple time windows are then matched with the typical pregnancy reaction features of various pregnancy reaction constitutions to obtain the similarity and the target pregnancy reaction feature with the highest similarity. Reference pregnancy reaction data for the current time window is extracted from the target pregnancy reaction features, and a pregnancy test necessity score is calculated based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital sign data, the reference pregnancy reaction data, the time interval from the last pregnancy test, and the weight of the gestational week in the current time window. The weight of the gestational week is determined based on the fetal demise data sample. The system determines the necessity of pregnancy testing based on multiple consecutive time windows, and outputs a warning message when the determination result indicates that pregnancy testing is necessary in the current time window.

[0009] This application also provides a pregnancy detection alert system based on pregnancy relapse analysis, including: The acquisition module is used to acquire multiple sets of vital sign data samples and pregnancy reaction data samples, and to acquire the vital sign data and pregnancy reaction data of the current object in multiple sampling time windows. The vital sign data samples and the vital sign data both include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples and the pregnancy reaction data both include the degree values ​​of various subjective evaluation pregnancy reaction parameters. The frequency of the sampling time windows is predetermined. The feature extraction module is used to extract features based on the multiple sets of vital sign data samples and pregnancy reaction data samples to obtain typical pregnancy reaction features of various pregnancy reaction constitutions. The typical pregnancy reaction features include subjective and objective fusion features of multiple pregnancy time windows. The matching module is used to fuse the vital signs data and pregnancy reaction data from multiple sampling time windows to obtain the current fused features, and then match the current fused features from multiple time windows with the typical pregnancy reaction features of various pregnancy reaction constitutions to obtain the similarity and the target pregnancy reaction feature with the highest similarity. The scoring module is used to extract reference pregnancy reaction data for the current time window from the target pregnancy reaction features, and calculate the necessity score of pregnancy test based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital sign data, the reference pregnancy reaction data, the time interval since the last pregnancy test, and the weight of the gestational week in the current time window. The weight of the gestational week is determined based on the fetal demise data sample. The prompt module is used to determine the necessity of pregnancy testing based on multiple consecutive time windows, and output a warning message when the determination result indicates that pregnancy testing is necessary in the current time window.

[0010] The beneficial effects of this invention are as follows: This invention provides a pregnancy check-up prompting method and system based on pregnancy relapse analysis. By collecting daily subjective scores of pregnancy relapse and objective vital sign data from wearable devices, it constructs a fusion of subjective and objective features and identifies the pregnant woman's constitution type. This allows check-up prompts to no longer rely on a fixed schedule but rather to align with the individual's actual physiological responses and changes in embryonic activity. It introduces gestational week weighting based on the probability of miscarriage, automatically increasing the sampling frequency during high-risk periods (e.g., 6-8 weeks of gestation). Simultaneously, based on trend judgment using multiple consecutive days of pregnancy check-up necessity scores, it can capture early signs of miscarriage such as "disappearance of abnormal pregnancy relapse" and "continuous deviation of objective vital signs," issuing early warnings during high-risk windows and effectively avoiding the clinical dilemma of miscarriage occurring between scheduled prenatal check-ups without being detected. For low-risk pregnant women with stable pregnancy relapse and subjective and objective data consistent with their constitution reference curve, the system provides a lower necessity score and does not frequently prompt for pregnancy check-ups, thereby reducing unnecessary outpatient visits, ultrasound examinations, and the burden of travel for pregnant women, allowing limited medical resources to be concentrated on truly high-risk individuals. Attached Figure Description

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is an illustration of an application scenario for a pregnancy detection alert method based on pregnancy reversal analysis, as shown in one embodiment of this application. Figure 2 This is a flowchart illustrating a pregnancy test notification method based on pregnancy reversal analysis in one embodiment of this application; Figure 3 This is a schematic diagram of a questionnaire page in one embodiment of this application; Figure 4 This is a schematic diagram of the operating interface in one embodiment of this application; Figure 5This is a structural diagram of a pregnancy detection alert system based on pregnancy reversal analysis, as shown in one embodiment of this application; Figure 6 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0012] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0013] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the layers related to the present invention and are not drawn according to the actual number, shape and size ratio of the layers in the actual implementation. In the actual implementation, the form and number of each layer can be arbitrarily changed, and the layer layout may also be more complex.

[0014] Numerous details are explored in the following description to provide a more thorough explanation of embodiments of the invention; however, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details.

[0015] First, it should be noted that the solution described in this application is not for predicting or diagnosing miscarriage, but rather for inferring the risk of miscarriage by collecting data on the pregnant woman's pregnancy symptoms. The inventors observed that many women experiencing miscarriage only realize the condition after their pregnancy symptoms have suddenly lessened or disappeared for a period of time. By this time, the stillborn fetus has already been attached to the uterus for some time and is highly susceptible to adhesions, making timely detection and treatment difficult for the pregnant woman. Therefore, the following solution was developed.

[0016] The overall concept of this application is as follows: This application collects a large amount of pregnancy reaction data (subjective scores) and physical signs data (objective physiological indicators) from pregnant women, extracts statistical features (mean, standard deviation, slope of change, peak value) from the physical signs data, aligns the physical signs features with the pregnancy reaction data in time and performs windowing, clusters based on objective physical signs features and corrects extreme values ​​for subjective scores within clusters, and constructs a fusion feature vector of subjective and objective factors; then, it connects individuals into time-series trajectories and uses Dynamic Time Warping (DTW) time-series clustering to construct typical pregnancy reaction feature curves (including mean vectors and covariance matrices for each gestational week) representing different physical conditions (high, medium, and low reactions).

[0017] For pregnant women, the system adaptively adjusts the sampling frequency according to their gestational age, acquires and fuses their vital signs and pregnancy reaction data; it matches the fused features with various typical pregnancy reaction features to obtain the reference features most suitable for their physical condition; then, based on similarity, Mahalanobis distance between the measured fused features and the reference features, the time interval since the last prenatal check-up, and gestational age weight (based on the probability of miscarriage), it calculates the necessity score of prenatal check-up; finally, it determines whether to issue a prenatal check-up warning based on the score trend over several consecutive days, thereby achieving personalized and dynamic prenatal check-up prompts for pregnant women with different physical conditions, reducing the risk of missed miscarriage detection and avoiding unnecessary waste of medical resources.

[0018] Figure 1 This is an illustration of an application scenario for a pregnancy detection alert method based on pregnancy reversal analysis, as shown in one embodiment of this application. Figure 1 As shown, this application adopts a front-end and back-end collaborative architecture. The front-end 110 is a user interaction terminal (such as a mobile APP), responsible for pushing pregnancy sickness questionnaires to users according to the sampling frequency set according to gestational week, collecting and uploading subjective pregnancy sickness data such as daily nausea frequency, nausea severity, appetite score, and breast tenderness score; at the same time, it communicates with the smart bracelet 130 via Bluetooth or other means to synchronously collect the user's daily objective vital signs data, including resting heart rate, sleep duration, activity level, and skin conductance response value. The front-end is also responsible for receiving pregnancy test warning information returned by the back-end and outputting warnings to users in a visual manner (such as pop-ups and notifications).

[0019] The smart bracelet 130 is a wearable device worn on the user's wrist to continuously collect the user's objective physiological indicators, including but not limited to resting heart rate, sleep duration, activity level and skin conductance response value, and synchronize the data to the front end 110 via wireless communication.

[0020] The backend, consisting of 120 cloud servers or remote computing centers, is responsible for storing all pregnancy-related symptoms data, vital sign data, and continuous tracking data. During the offline modeling phase, the backend performs tasks such as data preprocessing, vital sign feature extraction (including mean, standard deviation, slope of change, and peak value), clustering based on objective vital signs and correction of subjective score extrema, construction of subjective-objective fusion feature vectors, concatenation of individual time-series trajectories, DTW time-series clustering, and construction and verification of subgroup reference curves. During online service, the backend receives the current pregnant woman's vital sign data and pregnancy-related symptoms data uploaded by the frontend, completes data fusion, constitution matching, similarity calculation, Mahalanobis distance calculation, pregnancy test necessity scoring, and trend determination based on continuous time windows, ultimately generating pregnancy test notification results and pushing them to the frontend. The backend is also responsible for periodically updating model parameters to ensure the clinical accuracy of typical pregnancy-related symptoms.

[0021] Figure 2 This is a flowchart illustrating a pregnancy test notification method based on pregnancy reversal analysis in one embodiment of this application, as shown below. Figure 2 This embodiment illustrates a pregnancy detection and alert method based on pregnancy relapse analysis. This application is divided into a model building stage and a pregnancy detection and alert application stage, including the following steps: (1) Model building stage (1-1) Obtain multiple sets of vital sign data samples and pregnancy reaction data samples. The vital sign data samples include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples include the severity values ​​of various subjective evaluation parameters of pregnancy reaction. The time window described in this application can be either one day or 12 hours.

[0022] This application collected multiple sets of vital sign data and pregnancy reaction data samples through a combination of online / offline questionnaires and wearable devices. Considering that pregnant women in early pregnancy (4-12 weeks) are physically and mentally sensitive and prone to fatigue, and to avoid decreased compliance due to high-frequency long-term follow-up, this application adopted a "limited time window sampling" strategy: for each pregnant woman who participated in the questionnaire, subjective scores of pregnancy reaction and objective vital sign data from wearable devices were collected continuously for 5-7 days after enrollment, serving as a complete sample for that pregnant woman in that gestational week. Each sample contains vital sign data (resting heart rate, sleep duration, activity level, skin conductance response value) and pregnancy reaction data (frequency of nausea, nausea severity, appetite, breast tenderness) from multiple sampling time windows, sufficient to extract time-series statistical characteristics such as the mean, standard deviation, slope of change, and peak value of the vital signs.

[0023] The aforementioned questionnaire survey does not rely on a front-end web / APP, but rather uses manual data collection. During the manual collection process, if users are willing to provide complete data, follow-up questionnaires can be conducted to verify the model fitted in the subsequent process of this application. The specific details are described later.

[0024] (1-2) Based on the multiple sets of physical signs data samples and pregnancy reaction data samples, feature extraction is performed to obtain typical pregnancy reaction characteristics of various pregnancy reaction constitutions, wherein the typical pregnancy reaction characteristics include subjective and objective fusion characteristics of multiple pregnancy time windows; Methods for constructing typical pregnancy nausea characteristics in various pregnancy nausea predispositions include: (1-2-1) Preprocessing is performed on multiple sets of vital sign data samples and pregnancy reaction data samples to obtain preprocessed vital sign data samples and preprocessed pregnancy reaction data samples. The preprocessing includes outlier removal, missing value completion, and parameter normalization.

[0025] The original questionnaire data may contain data entry errors (e.g., nausea frequency >30), some pregnant women may have omitted certain parameters (missing values), and different parameters may have different units of measurement (e.g., nausea frequency is an integer, appetite score is 1-5, heart rate is 60-100 bpm). Outlier removal uses a 3x3 matrix. σThe principle or interquartile range method; missing value imputation using the mean or interpolation of other samples at the same gestational age; normalization maps each parameter to a uniform scale (such as Z-score) to avoid the bias of dimensional differences on subsequent fitting.

[0026] The multiple sets of vital sign data and pregnancy remission data collected in this step are derived from limited time windows of sampling by different pregnant women (5-7 consecutive days of sampling for each pregnant woman), rather than long-term continuous tracking data of the same pregnant woman. Specifically, each pregnant woman participates in a single data collection session lasting several days within a specific gestational week (e.g., 6 weeks + 3 days to 6 weeks + 9 days of gestation), providing evaluation values ​​of pregnancy remission parameters and synchronous objective vital sign data for multiple time windows within that period. By aggregating a large number of independent samples covering each gestational week (≥50 samples per gestational week), a raw dataset is constructed for subsequent feature extraction and clustering. This limited window design avoids the burden and decreased compliance issues caused by high-frequency long-term tracking for pregnant women, while providing sufficient time-series information for each sample to extract statistical features such as the mean, standard deviation, slope of change, and peak value of vital signs.

[0027] (1-2-2) Extract the vital signs features from the preprocessed vital signs data, wherein the vital signs features include the mean value of vital signs, the standard deviation of vital signs, the slope of changes in vital signs, and the peak value of vital signs; From the preprocessed vital signs data, the following statistical features were extracted for each sample within its corresponding time window: Average vital signs: The arithmetic mean of the vital signs data within the window, reflecting the average physiological state level of the pregnant woman at this gestational age. Standard deviation of vital signs: The standard deviation of vital signs data within the window reflects the fluctuation range of physiological indicators. Excessive fluctuation may indicate physiological stress. Slope of changes in vital signs: The slope of the linear fit of vital sign data within the window over time, reflecting the upward or downward trend of physiological indicators; Peak vital signs: The maximum value of vital signs data within the window, reflecting the degree of extreme deviation of physiological indicators.

[0028] The above four statistical features describe the distribution pattern of objective physical characteristics from four dimensions: "centrality", "dispersion", "trend of change" and "extreme state", respectively, providing a comprehensive physiological state description for subsequent clustering.

[0029] (1-2-3) The physical signs and the preprocessed pregnancy reaction data samples are aligned in time sequence, and multiple time window sample combinations are extracted. Each sample combination includes pregnancy reaction data samples and physical signs within a time window (the specific types of the objective physical signs and pregnancy reaction parameters have been defined in step (1-1). The physical signs and the preprocessed pregnancy nausea data samples were time-aligned, and samples from multiple time windows were combined. Each sample combination included pregnancy nausea data samples and physical signs within one time window. The objective physical signs included resting heart rate, sleep duration, activity level, and skin conductance. The pregnancy nausea parameters included the number of nausea episodes, self-assessed nausea severity, appetite score, and breast tenderness score.

[0030] Since there may be a time discrepancy between the completion time of the subjective pregnancy relapse questionnaire and the collection time of objective physical signs, this application takes the completion time of the pregnancy relapse questionnaire as the center, and takes a certain time period (such as ±30 minutes) before and after as a time window. It extracts the statistical characteristics of objective physical signs (mean, standard deviation, slope of change, peak value) within the window, and pairs them with the subjective pregnancy relapse score at that time to form a sample combination of "objective physical signs + subjective pregnancy relapse score".

[0031] (1-2-4) The preprocessed vital signs data samples were divided according to gestational age, and multiple sample combinations were clustered based on the preprocessed vital signs data samples of different gestational ages to obtain multiple clusters; and the extreme values ​​of the preprocessed pregnancy relapse data samples of the sample combinations within the clusters were corrected to obtain the corrected sample combinations. Because objective signs and subjective pregnancy reactions naturally differ significantly at different gestational weeks (e.g., 4 weeks vs. 10 weeks), directly clustering samples from all gestational weeks together would lead to samples from different gestational weeks being incorrectly grouped into the same cluster, rendering subsequent corrections clinically meaningless. Therefore, this application first stratifies the samples by gestational week to ensure that clustering is performed only among samples from the same gestational week.

[0032] Within each gestational week, multiple sample combinations are clustered based on preprocessed vital sign data samples (i.e., a multidimensional feature vector composed of the mean, standard deviation, slope of change, and peak value of each sample). The clustering algorithm uses K-Means (elbow rule to determine the K value) to group samples with similar objective vital sign states into the same cluster. The core logic of this clustering is that pregnant women with similar objective vital sign states should have similar subjective pregnancy relapse scores.

[0033] The correction process includes: (1-2-4-1) Calculate the median of each pregnancy reaction parameter in the preprocessed pregnancy reaction data samples of the sample combination within the cluster. and standard deviation ,in, Represents a cluster index. Indicates the index of the pregnancy reversal parameter; (1-2-4-2) The standard deviation is... Compared with the preset standard deviation threshold Compare; (1-2-4-3) in At that time, the value of the pregnancy parameter for each sample combination within the cluster. After shrinkage treatment, the corrected pregnancy reaction parameters were obtained. And the corrected sample combination, where the corrected pregnancy reaction parameter value The mathematical expression is: The core purpose of shrinkage processing is to eliminate subjective reporting noise while preserving the true differences between individuals.

[0034] Specifically, samples within the same cluster after clustering are highly similar in objective physical characteristics (such as heart rate, sleep duration, and activity level are all at similar levels). According to medical common sense, pregnant women with similar objective physiological states should also have similar subjective pregnancy symptoms. However, due to the arbitrariness of questionnaire completion—some pregnant women may have memory biases, different tolerance thresholds, or careless completion, resulting in subjective scores that significantly deviate from the reasonable range that should exist under these objective conditions—this creates a data contradiction of "similar objective physical characteristics but vastly different subjective scores."

[0035] Using these uncorrected noisy data directly for subsequent modeling will severely interfere with the accuracy of constitution clustering and reference curve construction. A common method is to directly remove outliers, but this leads to sample size loss and may mistakenly remove some truly extreme constitutions (such as individuals with particularly strong reactions). Therefore, this application performs shrinkage processing on extreme values ​​to reduce their interference. In the shrinkage calculation formula above, the subjective scores of all samples within a cluster shrink proportionally around the median. The standard deviation of the shrunken cluster precisely meets the preset threshold requirement, while the cluster median remains unchanged, and the relative ranking between samples remains unchanged.

[0036] (1-2-4-4) in When the subjective scores within the cluster are sufficiently consistent, the sample combination within the cluster is used as the corrected sample combination without any further modification.

[0037] (1-2-5) The modified sample combination is used as the subjective and objective fusion feature vector, and typical pregnancy symptoms of various pregnancy symptoms are extracted based on multiple subjective and objective fusion feature vectors.

[0038] The modified sample combination is used as a subjective-objective fusion feature vector, which includes objective statistical features (mean, standard deviation, slope, peak value) and modified subjective pregnancy sickness scores (number of nausea, degree of nausea, appetite, breast tenderness).

[0039] Based on the extraction of typical pregnancy symptoms for various pregnancy-related constitutions using multiple subjective and objective fusion feature vectors, the specific features include: (1-2-5-1) Multiple subjective and objective fusion feature vectors are concatenated according to the data source individuals to obtain multiple time-series fusion feature vectors, wherein the time-series fusion feature vectors include subjective and objective fusion feature vectors of multiple time windows; Multiple subjective and objective fusion feature vectors are concatenated in ascending order of gestational age based on the data source individual (i.e., pregnant woman ID) to obtain multiple individual time-series fusion feature vectors. The time-series fusion feature vectors include the subjective and objective fusion feature vectors of the individual in multiple time windows.

[0040] (1-2-5-2) The temporal fusion feature vectors of multiple individuals are resampled and interpolated at equal intervals to obtain a standard fusion feature vector of standard length, wherein each gestational week in the standard fusion feature vector contains one and only one subjective and objective fusion feature vector. The temporal fusion feature vectors of multiple individuals are resampled at equal intervals and interpolated to obtain a standard fusion feature vector of standard length. Since the start and end weeks of gestation vary among pregnant women and the sampling time windows are not completely consistent, the temporal data of all individuals need to be unified onto the same gestational week time axis (e.g., using "days" as the granularity, uniformly from week 4 to week 12 of gestation). Each gestational week time window contains one and only one subjective-objective fusion feature vector, and missing values ​​are filled in by linear interpolation.

[0041] (1-2-5-3) Perform hierarchical clustering or K-medoids clustering based on dynamic time regularization distance on the standard fusion feature vectors of multiple individuals to obtain various sub-clusters of pregnancy-reversal constitution; Hierarchical clustering or K-medoids clustering based on dynamic time warping (DTW) distance is performed on the standard fused feature vectors of multiple individuals to obtain various subgroups of pregnancy relapse (usually divided into three subgroups: high response, medium response, and low response). DTW distance allows for slight misalignment of two time-series curves on the time axis (because the peak of pregnancy relapse may occur at 7, 8, or 9 weeks of gestation for different pregnant women), and can cluster curves with similar shapes but slightly different phases into one class, which is more in line with clinical practice.

[0042] (1-2-5-4) For each subgroup of pregnancy relapse, calculate the mean vector and covariance matrix for each gestational week time window, and set the reference curves for each feature parameter in each subgroup; and construct typical pregnancy relapse features based on the reference curves of multiple subgroups with multiple feature parameters, the mean vectors and covariance matrices of multiple gestational week time windows of multiple subgroups.

[0043] For each subgroup of pregnancy relapse, the mean vector and covariance matrix for each gestational week time window are calculated, and reference curves for each feature parameter in each subgroup are collected; and typical pregnancy relapse characteristics are constructed based on reference curves of multiple subgroups with multiple feature parameters, mean vectors and covariance matrices of multiple gestational week time windows of multiple subgroups.

[0044] Specifically, for each constitution cluster , each week of pregnancy Calculate the number of individuals within it at that gestational week: Mean vector : The in-cluster mean of all fusion features (objective physical signs and statistics + adjusted subjective scores) for this gestational week; covariance matrix : Describes the variation and correlation among characteristics at this gestational week.

[0045] Ultimately, the typical characteristics of pregnancy nausea are defined as: This involves three reference curves that dynamically change with gestational age (high, medium, and low responsiveness). Each curve includes a mean trajectory and a covariance envelope. The mean vector describes the typical value of the responsiveness at that gestational age, while the covariance matrix describes the variation and correlation structure among the parameters. The introduction of the covariance matrix allows subsequent comparisons between measured data and typical characteristics to consider the correlation between parameters, improving the sensitivity of anomaly detection. Once established, the typical gestational remission feature database can be used long-term for individualized assessments of different pregnant women.

[0046] (2) Pregnancy checkup reminder application stage (2-1) Obtain the vital signs data and pregnancy reaction data of the current object in multiple sampling time windows, wherein the vital signs data includes the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data includes the degree values ​​of pregnancy reaction parameters evaluated by multiple subjective evaluations, and the frequency of the sampling time windows is predetermined. Users upload pregnancy reaction data via a web / app. The web / app automatically calculates the upload frequency based on the user's gestational age, with higher upload frequencies for more dangerous gestational weeks. The calculation process for frequency and upload time window is as follows: A. Determine the gestational week of the current time window, and determine the weight of the gestational week of the current time window based on a pre-built gestational week-weight correspondence table. The method for constructing the gestational week-weight correspondence table includes: extracting the probability of fetal demise for each gestational week from fetal demise sample data, and using the fetal demise probability as the weight of the gestational week; The risk of miscarriage is not evenly distributed throughout the first trimester. Clinical statistics show that weeks 6-8 of gestation (especially around week 7) are the peak period for miscarriage, while the risk decreases significantly before week 4 or after week 10. To obtain more intensive monitoring data during the period of highest risk, the system first calculates the incidence of miscarriage for each gestational week from historical miscarriage sample data (e.g., the number of miscarriages per 1,000 pregnant women), normalizes this incidence rate, and then uses it as the weight for that gestational week. A higher weight indicates a greater risk of miscarriage at that gestational age, requiring more frequent data sampling.

[0047] Table 1 shows an example of the gestational age-weight correspondence in one embodiment of this application. The constructed gestational age-weight correspondence table is shown in the table below: Table 1. Examples of Gestational Week-Weight Correspondence

[0048] The weights mentioned above have not been normalized, but they can be normalized and then used in subsequent calculations.

[0049] B. Weighting based on the gestational week within the current time window Calculate the sampling frequency for the current gestational week. Wherein, the sampling frequency The mathematical expression is: In the formula, This represents the floor function. Indicates the benchmark weight. Indicates the reference sampling frequency; The system presets a baseline gestational week (usually the gestational week with the lowest risk of miscarriage, such as 4 weeks or 10 weeks or later) and a corresponding baseline weight. and reference sampling frequency (For example, sampling every 3 days). For the current gestational week, the sampling frequency is linearly adjusted according to its weight and the baseline weight: the higher the weight, the higher the sampling frequency. The formula uses a rounding function. Setting the frequency to an integer (such as every day, every 2 days, every 3 days, etc.) makes it easier for users to understand and for the system to schedule.

[0050] In addition, if the calculation result is less than 1, then 1 is taken (i.e., daily sampling).

[0051] If the calculation result is greater than 7, the maximum value can be limited (e.g., no more than once a week), but usually the risk is high in the early stages, so the frequency will not exceed once a day.

[0052] C. Based on the sampling frequency Determine multiple sampling time windows.

[0053] Once the sampling frequency for the current gestational week is obtained (Unit: times / day) The system can then calculate the specific dates and time windows within a future period where users need to upload pregnancy reaction data. A uniform interval strategy is typically used: if... =1 time / day, then a reminder will be sent at a fixed time every day (e.g., 8 pm); if =0.5 times / day, then once every 2 days. The sampling time window will be determined with reference to the user's usage habits (such as avoiding the early morning hours) and will be carried out continuously throughout the pregnancy until the next gestational week, at which point the frequency will be recalculated.

[0054] During data collection, the front-end automatically outputs a prepared prompt page to the user, who can then select the corresponding pregnancy reaction parameters based on the prompt information. Figure 3 This is a schematic diagram of a questionnaire page in one embodiment of this application. An example of questionnaire collection is shown below. Figure 3 As shown, for example: Example 1: Select the number of times you feel nauseous How many times did you feel nauseous today? (excluding dry heaving) Options: ○ 0 times ○ 1-2 times ○ 3-5 times ○ 6-10 times ○ More than 10 times; Example 2: Self-assessment of nausea level "How bad was it each time you felt nauseous? Please select the closest answer." Options: ○ Mild (does not affect normal activities) ○ Moderate (requires brief sitting or lying down) ○ Severe (completely unable to do anything, even feeling nauseous); Example 3: Appetite rating score "Compare your appetite today to before pregnancy:" Slide bar or 1-5: 1 point (very poor, can barely eat anything) → 5 points (very good, just like before pregnancy); Example 4: Breast tenderness score "Gently press your breasts, how much swelling and tenderness do you feel:" Options: ○ 0 points (no pain) ○ 1 point (mild tenderness) ○ 2 points (significant swelling and pain but does not affect clothing) ○ 3 points (severe swelling and pain, painful to the touch) Example 5: Level of fatigue and drowsiness (optional extended parameter) How tired or sleepy did you feel today? Options: ○ 1 point (Full of energy, no different from before pregnancy) ○ 2 points (Slightly fatigued) ○ 3 points (Average) ○ 4 points (Very tired, always want to lie down) ○ 5 points (Unable to get up and move around); The front-end prompt page typically uses a card-style layout, displaying only one parameter at a time. Users can switch between questions by swiping left or right or using "previous / next" buttons, and the page automatically saves the user's selections. Once all parameters are filled in, the data is packaged and uploaded to the backend. This interaction method reduces the user's input burden while ensuring structured data collection.

[0055] By using predefined judgment prompts, subjective feelings can be quantified as objectively as possible to facilitate subsequent evaluation.

[0056] (2-2) The vital signs data and pregnancy reaction data from multiple sampling time windows are fused to obtain the current fused features, and the current fused features from multiple time windows are matched with the typical pregnancy reaction features of various pregnancy reaction constitutions. The matching process includes: (2-2-1) The vital signs data and the pregnancy reaction data are preprocessed to obtain preprocessed vital signs data and preprocessed pregnancy reaction data. The original vital sign data and pregnancy reaction data uploaded by pregnant women are processed using the same outlier removal, missing value imputation, and normalization methods as in the modeling phase to ensure data consistency. This eliminates input noise and avoids matching bias caused by differences in data format or units.

[0057] (2-2-2) Extract the vital signs features from the preprocessed vital signs data and fuse them with the preprocessed pregnancy reaction data to obtain the current fused features. Extract the same statistical features (mean, standard deviation, slope of change, peak value) from the preprocessed vital signs data as those in the modeling stage, and then concatenate these vital signs features with the preprocessed pregnancy and relapse data within the same time window to form the fusion feature vector for the current time window.

[0058] (2-2-3) Extract the value of each feature parameter from the current fusion features of multiple time windows. The numerical sequence of each feature parameter (objective vital signs statistics + corrected subjective scores) is extracted from the fused features of pregnant women across multiple sampling time windows (e.g., daily over the past 7 days). Here, 7 days refers to the matching window during the online inference phase, which differs from the 5-7 day collection window for each sample during the modeling phase. The purposes of the two windows are to construct a group reference curve and to assess the current individual status, respectively.

[0059] (2-2-4) Calculate the residual sequences of the characteristic parameters at multiple time windows and the reference curves of multiple subgroups. For each subgroup's reference curve, the measured fusion characteristic value of the pregnant woman is subtracted from the reference value of the curve at the same gestational week to obtain a residual sequence. The residual reflects the degree of deviation of the pregnant woman from the typical value of that constitution, thereby eliminating the influence of gestational week and making the deviations of different time windows comparable; the mean of the residual sequence can measure the overall direction and magnitude of deviation.

[0060] (2-2-5) Calculate the average value of the residual sequence corresponding to each subgroup, and calculate the similarity based on the average value of the sequence parameters of various characteristic parameters of each subgroup. The mathematical expression for the similarity is:

[0061] In the formula, Indicates the subgroup number, Indicates the sequence number of the pregnancy reaction parameter. Indicates the number of pregnancy reaction parameters. For the first The weights of the anti-pregnancy parameters, Indicates the first The first subgroup The average value of the residual sequence of the pregnancy inversion parameter; By integrating information from multiple parameters and time windows, the matching degree between pregnant women and each body type is quantified; weighting. This can demonstrate the importance of different parameters in distinguishing body types.

[0062] (2-2-6) The typical pregnancy and relapse features of the subgroup that simultaneously satisfies the following conditions: similarity greater than the preset similarity threshold and the most similarity are taken as the target pregnancy and relapse features.

[0063] The similarity score must be greater than a preset threshold (to ensure matching confidence) and be the highest among all subgroups before the typical pregnancy reaction characteristics of that subgroup are considered as targets. If no subgroup meets this requirement, matching can be temporarily suspended or the default constitution can be used. This avoids misjudgments caused by low-confidence matching and ensures that subsequent scoring is based on reliable constitution attribution.

[0064] (2-3) Extract reference pregnancy reaction data for the current time window from the target pregnancy reaction features, and calculate the necessity score of pregnancy test based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital signs data, the reference pregnancy reaction data, the time interval from the last pregnancy test, and the weight of the gestational week in the current time window.

[0065] The calculation process for the necessity score of prenatal testing includes: (2-3-1) Preprocess, extract features, and fuse the pregnancy reaction data and vital sign data uploaded in the current time window to obtain the current fused feature vector; The raw vital signs data uploaded in the current time window and the pregnancy reaction data undergo the same preprocessing, vital sign feature extraction, and fusion operations as in the modeling phase to form the current fused feature vector. The mean vector and covariance matrix of the preprocessed fused features and typical features are on the same dimension, which facilitates the calculation of Mahalanobis distance.

[0066] (2-3-2) Based on the aforementioned similarity The current fused feature vector The reference pregnancy reaction data, the time interval Weight of the current gestational week and the time window Calculate the necessity score of prenatal testing The mathematical expression for the pregnancy test necessity score is as follows: In the formula, and All represent weighting coefficients ( , ), This indicates the gestational week time window in the reference data for pregnancy reversal. The mean vector, This indicates the gestational week time window in the reference data for pregnancy reversal. The covariance matrix, This represents the Mahalanobis distance calculation function. Indicates the time accumulation factor. This indicates that the minimum value is taken. The time factor uses a square root function, which increases rapidly in the early stages and then slows down, without excessively amplifying the score.

[0067] in, It reflects the degree of deviation from long-term physical condition matching (i.e. whether the pregnant woman's recent overall pregnancy reaction pattern is consistent with the typical characteristics of her own physical condition). It reflects the degree of acute abnormality in the current daily data and incorporates gestational risk and cumulative effects over time. Therefore, the necessity score in this application integrates short-term deviation (Mahavir distance at the current point), long-term pattern deviation (similarity), pregnancy-related risks, and cumulative effects over time to comprehensively assess the urgency of prenatal testing.

[0068] (2-4) Based on the pregnancy test necessity score of multiple consecutive time windows, a pregnancy test prompt is determined, and a warning message is output when the determination result indicates that a pregnancy test is necessary in the current time window.

[0069] The determination process includes: (2-4-1) Extract the pregnancy test necessity score of multiple time windows within the current time window, wherein the current time window includes multiple time windows of the target duration before the current time window; The necessity scores for each day within a period preceding the current time window (e.g., 7 days) are used to construct a sequence. This continuous trend helps to suppress false positives caused by random daily fluctuations.

[0070] (2-4-2) Interpolate the missing values ​​of the pregnancy test necessity scores for multiple time windows within the current time window to obtain the sequence to be analyzed; For missing scores within the window due to reasons such as users not uploading, linear interpolation or forward imputation is used to fill in the gaps and form a complete sequence. This ensures the continuity of subsequent trend analysis and avoids interruptions in judgment due to a small number of missing scores.

[0071] (2-4-3) Compare multiple pregnancy test necessity scores in the sequence to be analyzed with preset scoring thresholds, and determine that pregnancy test is necessary in the current time window when there are more than N pregnancy test necessity scores in the sequence to be analyzed that exceed the preset scoring thresholds; The system counts the number of scores exceeding a preset threshold (e.g., 4.0) within a given window. If more than N scores (e.g., 2) are recorded, a pregnancy test is immediately deemed necessary. This system identifies high-risk statuses that persist for several days, avoiding frequent false alarms caused by occasional daily spikes.

[0072] (2-4-4) When there are no more than N pregnancy test necessity scores exceeding the first threshold in the sequence to be analyzed, the sequence to be analyzed is linearly fitted to obtain the slope, the slope is compared with a preset slope threshold, and when the slope is greater than the preset slope threshold and there are any pregnancy test necessity scores exceeding the first threshold in the sequence to be analyzed, it is determined that there is a need for pregnancy test in the current time window.

[0073] When the number of points exceeding the threshold within the window is less than N, a linear fit is performed on the scoring sequence to calculate the slope. If the slope is greater than a preset slope threshold (indicating a rapid upward trend in the score) and at least one point exceeds the threshold, a pregnancy test is deemed necessary. This system can detect situations where the risk has not yet reached the level of exceeding the threshold for several days but is rapidly increasing, enabling early warning, improving sensitivity, and controlling false positives. Through the above matching, scoring, and judgment process, the system achieves a complete closed loop from individual constitution identification to dynamic risk quantification and robust early warning output, balancing sensitivity and specificity, and effectively solving the rigidity problem of scheduled pregnancy tests.

[0074] Figure 4 This is a schematic diagram of the operating interface in one embodiment of this application, such as... Figure 4 As shown in this example, after submitting today's data, the process described above is used to determine the result, and the interface displays: "No special pregnancy test is required at this time".

[0075] This invention provides a pregnancy check-up notification method based on pregnancy relapse analysis. By collecting daily subjective scores of pregnancy relapse and objective vital signs data from wearable devices, it constructs a fusion of subjective and objective features and identifies the pregnant woman's constitution type. This allows the check-up notifications to no longer rely on a fixed schedule but instead be tailored to the individual's actual physiological responses and changes in embryonic activity. A gestational week weighting based on the probability of miscarriage is introduced, automatically increasing the sampling frequency during high-risk periods (e.g., 6-8 weeks of gestation). Simultaneously, trend analysis based on multiple days of pregnancy check-up necessity scores allows for the detection of early signs of miscarriage, such as "disappearance of abnormal pregnancy relapse" and "persistent deviation of objective vital signs," providing early warnings during high-risk windows and effectively avoiding the clinical dilemma of undetected miscarriage occurring between scheduled prenatal check-ups. For low-risk pregnant women with stable pregnancy relapse and subjective and objective data consistent with their constitution reference curve, the system provides a lower necessity score and does not frequently prompt for prenatal check-ups, thereby reducing unnecessary outpatient visits, ultrasound examinations, and the burden of travel for pregnant women, allowing limited medical resources to be concentrated on truly high-risk individuals.

[0076] like Figure 5 As shown, this application discloses a pregnancy detection alert system based on pregnancy relapse analysis, comprising: The acquisition module is used to acquire multiple sets of vital sign data samples and pregnancy reaction data samples, and to acquire the vital sign data and pregnancy reaction data of the current object in multiple sampling time windows. The vital sign data samples and the vital sign data both include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples and the pregnancy reaction data both include the degree values ​​of various subjective evaluation pregnancy reaction parameters. The frequency of the sampling time windows is predetermined. The feature extraction module is used to extract features based on the multiple sets of vital sign data samples and pregnancy reaction data samples to obtain typical pregnancy reaction features of various pregnancy reaction constitutions. The typical pregnancy reaction features include subjective and objective fusion features of multiple pregnancy time windows. The matching module is used to fuse the vital signs data and pregnancy reaction data from multiple sampling time windows to obtain the current fused features, and then match the current fused features from multiple time windows with the typical pregnancy reaction features of various pregnancy reaction constitutions to obtain the similarity and the target pregnancy reaction feature with the highest similarity. The scoring module is used to extract reference pregnancy reaction data for the current time window from the target pregnancy reaction features, and calculate the necessity score of pregnancy test based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital sign data, the reference pregnancy reaction data, the time interval since the last pregnancy test, and the weight of the gestational week in the current time window. The weight of the gestational week is determined based on the fetal demise data sample. The prompt module is used to determine the necessity of pregnancy testing based on multiple consecutive time windows, and output a warning message when the determination result indicates that pregnancy testing is necessary in the current time window.

[0077] This invention discloses a pregnancy check-up reminder system based on pregnancy relapse analysis. By collecting daily subjective scores of pregnancy relapse and objective vital signs data from wearable devices, the system constructs a fusion of subjective and objective features and identifies the pregnant woman's constitution type. This allows check-up reminders to no longer rely on a fixed schedule but instead to align with the individual's actual physiological responses and changes in embryonic activity. A gestational week weighting based on the probability of miscarriage is introduced, automatically increasing the sampling frequency during high-risk periods (e.g., 6-8 weeks of gestation). Simultaneously, trend analysis based on multiple days of pregnancy check-up necessity scores allows the system to detect early signs of miscarriage, such as "disappearance of abnormal pregnancy relapse" and "persistent deviation of objective vital signs," providing early warnings during high-risk windows and effectively avoiding the clinical dilemma of undetected miscarriage occurring between scheduled prenatal check-ups. For low-risk pregnant women with stable pregnancy relapse and whose subjective and objective data align with their constitution reference curve, the system provides a lower necessity score and does not frequently remind them to undergo check-ups, thereby reducing unnecessary outpatient visits, ultrasound examinations, and the burden of travel for pregnant women, allowing limited medical resources to be concentrated on truly high-risk individuals.

[0078] Figure 6 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 6 The computer system of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0079] like Figure 6 As shown, the computer system includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on a program stored in Read-Only Memory (ROM) 602 or a program loaded from storage portion 608 into Random Access Memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An Input / Output (I / O) interface 605 is also connected to the bus 604.

[0080] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0081] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs various functions defined in the system of this application.

[0082] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0083] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0084] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0085] Another aspect of this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a computer's processor, causes the computer to perform the method as described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.

[0086] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various embodiments described above.

[0087] The above embodiments are merely preferred embodiments provided to fully illustrate this application, and the scope of protection of this application is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on this application are all within the scope of protection of this application.

Claims

1. A method for pregnancy detection based on pregnancy reversal analysis, characterized in that, Including the following steps: Multiple sets of vital sign data samples and pregnancy reaction data samples are obtained, and the vital sign data and pregnancy reaction data of the current object in multiple sampling time windows are obtained. The vital sign data samples and the vital sign data both include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples and the pregnancy reaction data both include the degree values ​​of various subjective evaluation pregnancy reaction parameters. The frequency of the sampling time windows is predetermined. Based on the multiple sets of vital sign data samples and pregnancy reaction data samples, feature extraction was performed to obtain typical pregnancy reaction characteristics of various pregnancy reaction constitutions. Among them, the typical pregnancy reaction characteristics include subjective and objective fusion characteristics of multiple pregnancy time windows. The vital signs data and pregnancy reaction data from multiple sampling time windows are fused to obtain the current fused features. The current fused features from multiple time windows are then matched with the typical pregnancy reaction features of various pregnancy reaction constitutions to obtain the similarity and the target pregnancy reaction feature with the highest similarity. Reference pregnancy reaction data for the current time window is extracted from the target pregnancy reaction features, and a pregnancy test necessity score is calculated based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital sign data, the reference pregnancy reaction data, the time interval from the last pregnancy test, and the weight of the gestational week in the current time window. The weight of the gestational week is determined based on the fetal demise data sample. The system determines the necessity of pregnancy testing based on multiple consecutive time windows, and outputs a warning message when the determination result indicates that pregnancy testing is necessary in the current time window.

2. The method for pregnancy detection based on pregnancy reversal analysis according to claim 1, characterized in that, Based on the aforementioned multiple sets of physical sign data samples and pregnancy relapse data samples, feature extraction was performed to obtain typical pregnancy relapse characteristics for various pregnancy relapse constitutions, including: Multiple sets of vital sign data samples and pregnancy reaction data samples are preprocessed to obtain preprocessed vital sign data samples and preprocessed pregnancy reaction data samples. The preprocessing includes outlier removal, missing value imputation, and parameter normalization. Extract the vital signs features from the preprocessed vital signs data, wherein the vital signs features include the mean value of vital signs, the standard deviation of vital signs, the slope of changes in vital signs, and the peak value of vital signs; The physical signs and the preprocessed pregnancy reaction data samples are aligned in time sequence, and multiple time windows are extracted into sample combinations. Each sample combination includes pregnancy reaction data samples and physical signs within a time window. The objective physical signs include resting heart rate, sleep duration, activity level, and skin conductance response value. The pregnancy reaction parameters include the number of nausea episodes, self-assessed nausea level, appetite evaluation score, and breast tenderness score. The preprocessed vital signs data samples were divided according to gestational age, and multiple sample combinations were clustered based on the preprocessed vital signs data samples of different gestational ages to obtain multiple clusters; and extreme value correction was performed on the preprocessed pregnancy relapse data samples of the sample combinations within the clusters to obtain the corrected sample combinations. The modified sample combination is used as a subjective-objective fusion feature vector, and typical pregnancy symptoms of various pregnancy symptoms are extracted based on multiple subjective-objective fusion feature vectors.

3. The method for pregnancy detection based on pregnancy reversal analysis according to claim 2, characterized in that, Extremum correction is performed on the preprocessed pregnancy test data samples within the cluster to obtain the corrected sample combinations, including: Calculate the median of each pregnancy reaction parameter in the preprocessed pregnancy reaction data samples of the sample combinations within the cluster. and standard deviation ,in, Represents a cluster index. Indicates the index of the pregnancy reversal parameter; The standard deviation Compared with the preset standard deviation threshold Compare; exist At that time, the value of the pregnancy parameter for each sample combination within the cluster. After shrinkage treatment, the corrected pregnancy reaction parameters were obtained. And the corrected sample combination, where the corrected pregnancy reaction parameter value The mathematical expression is: exist When the sample combination within the cluster is used as the corrected sample combination, the sample combination within the cluster is used.

4. The method for pregnancy detection based on pregnancy reversal analysis according to claim 2, characterized in that, Based on the extraction of typical pregnancy relapse characteristics of various pregnancy relapse constitutions using multiple subjective and objective fusion feature vectors, including: Multiple subjective and objective fusion feature vectors are concatenated according to the data source individuals to obtain multiple time-series fusion feature vectors, wherein the time-series fusion feature vectors include subjective and objective fusion feature vectors of multiple time windows; The temporal fusion feature vectors of multiple individuals are resampled at equal intervals and interpolated to obtain a standard fusion feature vector of standard length. Each gestational week in the standard fusion feature vector contains one and only one subjective-objective fusion feature vector. Hierarchical clustering or K-medoids clustering based on dynamic time regularization distance is performed on the standard fusion feature vectors of multiple individuals to obtain various sub-clusters of pregnancy-reversal constitution; For each subgroup of pregnancy relapse, the mean vector and covariance matrix for each gestational week time window are calculated, and reference curves for each feature parameter in each subgroup are collected; and typical pregnancy relapse characteristics are constructed based on reference curves of multiple subgroups with multiple feature parameters, mean vectors and covariance matrices of multiple gestational week time windows of multiple subgroups.

5. The method for pregnancy detection based on pregnancy reversal analysis according to claim 1, characterized in that, Methods for determining multiple sampling time windows include: Determine the gestational week of the current time window, and determine the weight of the gestational week of the current time window based on a pre-built gestational week-weight correspondence table. The method for constructing the gestational week-weight correspondence table includes: extracting the probability of fetal demise for each gestational week from fetal demise sample data, and using the fetal demise probability as the weight of the gestational week; Weight based on the gestational week of the current time window Calculate the sampling frequency for the current gestational week. Wherein, the sampling frequency The mathematical expression is: In the formula, This represents the floor function. Indicates the benchmark weight. Indicates the reference sampling frequency; Based on the sampling frequency Determine multiple sampling time windows.

6. The method for pregnancy detection based on pregnancy reversal analysis according to claim 1, characterized in that, The vital signs data and pregnancy nausea data from multiple sampling time windows are fused to obtain the current fused features. These current fused features from multiple time windows are then matched with typical pregnancy nausea features of various pregnancy nausea constitutions to obtain the similarity score and the target pregnancy nausea feature with the highest similarity, including: The vital signs data and the pregnancy reaction data are preprocessed to obtain preprocessed vital signs data and preprocessed pregnancy reaction data; Extract the vital signs features from the preprocessed vital signs data and fuse them with the preprocessed pregnancy reaction data to obtain the current fused features; Extract the value of each feature parameter from the current fused features across multiple time windows; Calculate the residual sequences of the characteristic parameters at multiple time windows and the reference curves of multiple subgroups respectively; Calculate the average residual sequence for each subgroup, and then calculate the similarity based on the average sequence parameters of various feature parameters for each subgroup. , wherein the similarity The mathematical expression is: In the formula, Indicates the subgroup number, Indicates the sequence number of the pregnancy reaction parameter. Indicates the number of pregnancy reaction parameters. For the first The weights of the anti-pregnancy parameters, Indicates the first The first subgroup The average value of the residual sequence of the characteristic parameters; The typical pregnancy and relapse characteristics of the subgroup that simultaneously meets the following criteria are used as the target pregnancy and relapse characteristics: the similarity is greater than the preset similarity threshold and the subgroup with the highest similarity.

7. A method for pregnancy detection based on pregnancy reversal analysis according to claim 6, characterized in that, A pregnancy test necessity score is calculated based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital signs data, the reference pregnancy reaction data, the time interval since the last pregnancy test, and the weight of the gestational week in the current time window. This score includes: The pregnancy reaction data and vital sign data uploaded in the current time window are preprocessed, feature extracted, and fused to obtain the current fused feature vector. ; Based on the similarity The current fused feature vector The reference pregnancy reaction data, the time interval Weight of the current gestational week and the time window Calculate the necessity score of prenatal testing The mathematical expression for the pregnancy test necessity score is as follows: In the formula, and All represent weighting coefficients. This indicates the gestational week time window in the reference data for pregnancy reversal. The mean vector, This indicates the gestational week time window in the reference data for pregnancy reversal. The covariance matrix, This represents the Mahalanobis distance calculation function. Indicates the time accumulation factor. This indicates taking the minimum value.

8. A method for pregnancy detection based on pregnancy reversal analysis according to claim 1, characterized in that, Pregnancy testing necessity is determined based on a score of pregnancy testing necessity over multiple consecutive time windows, including: Extract the pregnancy test necessity score for multiple time windows within the current time window, wherein the current time window includes multiple time windows of the target duration prior to the current time window; Missing value interpolation is performed on the pregnancy test necessity scores of multiple time windows within the current time window to obtain the sequence to be analyzed; The necessity scores of multiple prenatal tests in the sequence to be analyzed are compared with preset scoring thresholds. If there are more than N prenatal test necessity scores in the sequence to be analyzed that exceed the preset scoring thresholds, it is determined that there is a necessity for prenatal tests in the current time window. If there are no more than N pregnancy test necessity scores exceeding the preset scoring threshold in the sequence to be analyzed, the sequence to be analyzed is linearly fitted to obtain a slope. The slope is compared with a preset slope threshold. If the slope is greater than the preset slope threshold, and the endpoint value of the sequence to be analyzed is greater than the sum of the starting value of the sequence to be analyzed and the preset change threshold, then it is determined that pregnancy test is necessary in the current time window; otherwise, it is determined that pregnancy test is not necessary in the current time window.

9. A method for pregnancy detection based on pregnancy reversal analysis according to claim 1, characterized in that, Obtain multiple sets of pregnancy reaction data, including: The survey questionnaire is sent to the respondents based on a predetermined sampling time window; and pregnancy reaction data is collected based on the survey questionnaire.

10. A pregnancy detection alert system based on pregnancy reversal analysis, characterized in that, include: The acquisition module is used to acquire multiple sets of vital sign data samples and pregnancy reaction data samples, and to acquire the vital sign data and pregnancy reaction data of the current object in multiple sampling time windows. The vital sign data samples and the vital sign data both include the values ​​of objective vital signs in multiple sampling time windows, and the pregnancy reaction data samples and the pregnancy reaction data both include the degree values ​​of various subjective evaluation pregnancy reaction parameters. The frequency of the sampling time windows is predetermined. The feature extraction module is used to extract features based on the multiple sets of vital sign data samples and pregnancy reaction data samples to obtain typical pregnancy reaction features of various pregnancy reaction constitutions. The typical pregnancy reaction features include subjective and objective fusion features of multiple pregnancy time windows. The matching module is used to fuse the vital signs data and pregnancy reaction data from multiple sampling time windows to obtain the current fused features, and then match the current fused features from multiple time windows with the typical pregnancy reaction features of various pregnancy reaction constitutions to obtain the similarity and the target pregnancy reaction feature with the highest similarity. The scoring module is used to extract reference pregnancy reaction data for the current time window from the target pregnancy reaction features, and calculate the necessity score of pregnancy test based on the similarity, the pregnancy reaction data uploaded in the current time window, the vital sign data, the reference pregnancy reaction data, the time interval since the last pregnancy test, and the weight of the gestational week in the current time window. The weight of the gestational week is determined based on the fetal demise data sample. The prompt module is used to determine the necessity of pregnancy testing based on multiple consecutive time windows, and output a warning message when the determination result indicates that pregnancy testing is necessary in the current time window.

Images