Nighttime hypoglycemia warning method during pregnancy based on continuous glucose monitoring

Abstract

The present application relates to the technical field of medical health monitoring, in particular to a pregnancy night hypoglycemia early warning method based on continuous blood glucose monitoring, the method comprising: acquiring multi-modal monitoring data of a target object during a night sleep period; constructing blood glucose baselines of each standard gestational week based on a historical database, and determining initial hypoglycemia threshold values of each gestational week; extracting blood glucose fluctuation characteristics of the target object based on the multi-modal monitoring data, and calculating a deviation degree between the blood glucose fluctuation characteristics and standard fluctuation characteristics corresponding to the standard gestational week blood glucose baselines; correcting the initial hypoglycemia threshold values according to the deviation degree, determining an early warning threshold value of the target object, and early warning the target object for night hypoglycemia based on the early warning threshold value. The present application can improve the effect of pregnancy night hypoglycemia early warning.

Description

Technical Field

[0001] This invention relates to the field of medical and health monitoring technology, specifically to a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring. Background Technology

[0002] Gestational diabetes mellitus (GDM) is a common metabolic complication of pregnancy, and strict blood glucose control is the core means to ensure maternal and infant outcomes. However, intensive clinical intervention to lower blood glucose can significantly increase the risk of hypoglycemia. Nighttime is a special physiological period with the longest fasting time and higher insulin sensitivity, making it a high-risk window for gestational hypoglycemia. Nighttime hypoglycemia is highly insidious; pregnant women cannot detect abnormalities while asleep. If it is not identified and intervened in time, it can easily lead to rebound hyperglycemia, exacerbate blood glucose fluctuations, and in severe cases, cause seizures or coma in the pregnant woman, seriously endangering her health and the normal development of the fetus.

[0003] Current clinical technologies for nocturnal hypoglycemia early warning during pregnancy still have significant shortcomings. Existing solutions mostly use a uniform and fixed blood glucose threshold for early warning, failing to adapt to the physiological and metabolic differences among pregnant women at different gestational weeks. Furthermore, nocturnal blood glucose fluctuations are regulated by multiple physiological parameters, including sleep state and autonomic nervous system function, while traditional early warning methods rely solely on a single blood glucose value, without incorporating related physiological indicators such as heart rate. In addition, individual physiological characteristics vary among pregnant women, and the influence of multiple physiological parameters on blood glucose fluctuations at night varies from person to person. Current technologies cannot accurately capture individualized blood glucose fluctuation characteristics, making it difficult to achieve differentiated and individualized hypoglycemia early warning.

[0004] In other words, current technologies for early warning of nocturnal hypoglycemia during pregnancy are not very effective. Summary of the Invention

[0005] To address the poor performance of existing nocturnal hypoglycemia early warning systems during pregnancy, this invention aims to provide a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous glucose monitoring. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring, the method comprising: Acquire multimodal monitoring data of the target subject during nighttime sleep. The multimodal monitoring data includes glucose time-series data characterizing changes in interstitial fluid glucose concentration, sleep physiological signals characterizing sleep stage status, and heart rate variability parameters characterizing the degree of autonomic nervous system excitation. Based on historical databases, standard gestational blood glucose baselines were constructed for each gestational week, and initial hypoglycemia thresholds for each gestational week were determined. The standard gestational blood glucose baselines were used to characterize the average nocturnal blood glucose metabolism level of the reference group at the same gestational week, and the initial hypoglycemia thresholds were used to characterize the safe threshold for nocturnal blood glucose metabolism of the reference group as the gestational week changes dynamically. Based on multimodal monitoring data, the blood glucose fluctuation characteristics of the target object are extracted, and the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational blood glucose baseline is calculated. The blood glucose fluctuation characteristics include multidimensional feature vectors of the target object in different sleep stages at night, and the deviation is used to characterize the degree of abnormality of the target object's nighttime glucose metabolism state relative to the standard physiological pattern. The initial hypoglycemia threshold is corrected based on the deviation, the warning threshold for the target object is determined, and a nighttime hypoglycemia warning is issued for the target object based on the warning threshold.

[0006] In one embodiment, the step of constructing standard gestational week blood glucose baselines based on a historical database includes: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point. For each sampling time point, extract a set of multiple glucose concentration values ​​for the reference population at each sampling time point; The mean value of each glucose concentration value set is calculated one by one, and the calculated mean value is determined as the baseline blood glucose value at each sampling time point. The baseline blood glucose values ​​corresponding to all sampling time points were fitted to generate a standard gestational week blood glucose baseline that changes over time.

[0007] In one embodiment, the step of extracting blood glucose fluctuation characteristics of the target object based on multimodal monitoring data includes: Preprocessing of glucose time-series data from multimodal monitoring data yields standardized glucose time-series data, standardized sleep physiological signals, and standardized heart rate variability parameters. Sleep stage status during nighttime monitoring periods is identified based on standardized sleep physiological signals; The standardized glucose time-series data were segmented according to sleep stages, and the statistical features, extreme features, and trend features of blood glucose were extracted for each sleep stage. The statistical features of blood glucose included the nighttime average blood glucose value, blood glucose standard deviation, and coefficient of variation. The extreme features of blood glucose included the highest and lowest blood glucose values ​​at night and the corresponding time points. The trend features of blood glucose included the rate of blood glucose decline and the duration of hypoglycemia. The statistical characteristics, extreme values, and trends of blood glucose in different sleep stages are fused to obtain the blood glucose fluctuation characteristics of the target subject.

[0008] In one embodiment, the step of calculating the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational age blood glucose baseline includes: Based on the standard gestational week blood glucose baseline, a time window matching the nighttime monitoring period of the target subjects was extracted, and the baseline mean blood glucose, baseline standard deviation blood glucose, and baseline blood glucose trend slope within the time window were calculated to form the standard fluctuation characteristics. The blood glucose statistical characteristics, blood glucose trend characteristics, and baseline blood glucose mean, baseline blood glucose standard deviation, and baseline blood glucose trend slope in the blood glucose fluctuation characteristics of the target object are calculated to obtain blood glucose mean deviation, blood glucose fluctuation amplitude deviation, and blood glucose change trend deviation. Based on the statistical characteristics, extreme values, and trends of blood glucose in different sleep stages at night, the weight coefficients for each feature are determined. The weight coefficients are positively correlated with the depth of sleep stages. The deviation of the mean blood glucose deviation, the blood glucose fluctuation amplitude deviation, and the blood glucose change trend deviation are weighted and summed based on the weighting coefficients to obtain the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational week blood glucose baseline.

[0009] In one embodiment, the step of determining the initial hypoglycemia threshold for each gestational week includes: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point. For each sampling time point, extract a set of multiple glucose concentration values ​​for the reference population at each sampling time point; The preset percentile values ​​of each glucose concentration value set are calculated one by one, and the calculated preset percentile values ​​are fitted to generate an initial hypoglycemia warning threshold curve that changes over time. The values ​​corresponding to each sampling time point on the initial hypoglycemia warning threshold curve are determined as the initial hypoglycemia threshold.

[0010] In one embodiment, the step of correcting the initial hypoglycemia threshold based on the deviation and determining the warning threshold for the target object includes: The deviation is combined with a preset correction coefficient to obtain the threshold adjustment amount; The initial hypoglycemia threshold is corrected based on the threshold adjustment amount to obtain the warning threshold for the target object.

[0011] In one embodiment, the step of providing a nighttime hypoglycemia warning to the target object based on a warning threshold includes: Real-time acquisition of glucose concentration values ​​of the target object; A hypoglycemia warning signal is generated when the detected glucose concentration is lower than the warning threshold and the duration reaches the preset duration threshold. The preset output device is controlled to perform the warning operation corresponding to the low blood sugar warning signal. The warning operation includes at least one of the following: audible and visual alarm, vibration reminder, and sending a warning message to the associated terminal.

[0012] In one embodiment, the step of fusing the statistical characteristics, extreme values, and trends of blood glucose in each sleep stage to obtain the blood glucose fluctuation characteristics of the target object includes: The statistical characteristics, extreme values, and trend characteristics of blood glucose were normalized to obtain the normalized values ​​of blood glucose statistics, extreme values, and trend. Obtain the feature weight matrix, which includes the weight coefficients corresponding to blood glucose statistical features, the weight coefficients corresponding to blood glucose extreme value features, and the weight coefficients corresponding to blood glucose trend features. Based on the feature weight matrix, the blood glucose statistical normalization value, blood glucose extreme value normalization value, and blood glucose trend normalization value are weighted and fused to obtain the blood glucose fluctuation characteristics of the target object.

[0013] In one embodiment, the step of constructing the feature weight matrix includes: Construct a basic weight mapping table, which includes the baseline weight coefficients corresponding to each feature dimension. The feature dimensions include at least blood glucose statistical features, blood glucose extreme value features, and blood glucose trend features. The sleep stage status corresponding to each sampling point is obtained, and the baseline weight coefficients in the basic weight mapping table are corrected according to the preset physiological rhythm adjustment rules to obtain the feature weight matrix. The preset physiological rhythm adjustment rules include: when the sleep stage status indicates deep sleep, the weight coefficients associated with extreme blood glucose features and blood glucose trend features are adjusted to be greater than their corresponding baseline weight coefficients; when the sleep stage status indicates REM sleep, the weight coefficients associated with blood glucose statistical features are adjusted to be greater than their corresponding baseline weight coefficients.

[0014] In one embodiment, the method further includes: Obtain the gestational cycle parameters of each candidate individual in the historical sample set, and divide the continuous gestational cycles into multiple discrete time windows according to the preset time span rules. The multimodal monitoring data of each candidate individual in the historical sample set are mapped to a target time window that matches their gestational cycle parameters. The multimodal monitoring data of all candidate individuals within the same target time window are aggregated to generate a reference population corresponding to each gestational cycle.

[0015] Secondly, another embodiment of the present invention provides a nocturnal hypoglycemia early warning system for pregnancy based on continuous glucose monitoring, the system comprising: The acquisition module is used to acquire multimodal monitoring data of the target object during nighttime sleep. The multimodal monitoring data includes glucose time-series data characterizing changes in interstitial fluid glucose concentration, sleep physiological signals characterizing sleep stage status, and heart rate variability parameters characterizing the degree of autonomic nerve excitation. The determination module is used to construct blood glucose baselines for each standard gestational week based on historical databases and determine the initial hypoglycemia threshold for each gestational week. The standard gestational week blood glucose baseline is used to characterize the average nocturnal blood glucose metabolism level of the reference group in the same gestational week, and the initial hypoglycemia threshold is used to characterize the safe threshold value of nocturnal blood glucose metabolism of the reference group as it dynamically changes with gestational week. The analysis module is used to extract the blood glucose fluctuation characteristics of the target object based on multimodal monitoring data, and calculate the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational blood glucose baseline. The blood glucose fluctuation characteristics include multidimensional feature vectors of the target object in different sleep stages at night, and the deviation is used to characterize the degree of abnormality of the target object's nighttime glucose metabolism state relative to the standard physiological pattern. The correction module is used to correct the initial hypoglycemia threshold based on the deviation, determine the warning threshold for the target object, and issue a nighttime hypoglycemia warning for the target object based on the warning threshold.

[0016] Thirdly, in another embodiment of the present invention, an electronic device is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect.

[0017] Fourthly, in another embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0018] The present invention has the following beneficial effects: This invention integrates three types of multi-source physiological monitoring data: glucose time-series data, sleep stage signals, and heart rate variability. Based on a sample database, it divides gestational weeks into intervals, constructs differentiated standard blood glucose baselines, and generates initial hypoglycemia thresholds suitable for each gestational week. Based on sleep physiological stages such as deep sleep and REM sleep, it extracts multi-dimensional blood glucose fluctuation characteristics (statistical, extreme values, and trends) hierarchically, quantifies the deviation of the target individual's glucose metabolism from the baseline for the same gestational week, and dynamically corrects the deviation to generate personalized warning thresholds, achieving intelligent detection and warning of nocturnal hypoglycemia. This invention aligns with the hormonal and metabolic changes during pregnancy, and the warning thresholds can be dynamically updated with gestational week progression and individual metabolic levels, reducing false alarms and missed alarms for nocturnal hypoglycemia warnings and improving the effectiveness of nocturnal hypoglycemia warnings during pregnancy. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart illustrating a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring, as provided in one embodiment of the present invention. Detailed Implementation

[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for a nocturnal hypoglycemia early warning method during pregnancy based on continuous blood glucose monitoring, provided by the present invention.

[0021] This invention proposes a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a schematic flowchart of a method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring, according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain multimodal monitoring data of the target subject during nighttime sleep. The multimodal monitoring data includes glucose time-series data characterizing changes in interstitial fluid glucose concentration, sleep physiological signals characterizing sleep stage status, and heart rate variability parameters characterizing the degree of autonomic nerve excitation.

[0022] The target group refers to pregnant women who are pregnant and need nighttime hypoglycemia risk monitoring, including pregnant women with gestational diabetes, high-risk pregnant women, and healthy pregnant women awaiting delivery.

[0023] Nighttime sleep refers to the continuous period from when a pregnant or postpartum woman falls asleep in bed at night until she wakes up in the morning (e.g., 10:00 PM to 6:00 AM the next day). During this period, pregnant women experience interrupted eating and significant hormonal fluctuations, making it a high-risk window for hypoglycemia during pregnancy.

[0024] Multimodal monitoring data includes combined monitoring data collected simultaneously from three different physiological dimensions, integrating three physiological indicators: blood glucose, sleep, and cardiovascular autonomic nervous system, which is different from single blood glucose data monitoring.

[0025] For example, multimodal monitoring data includes three types of data collected simultaneously: glucose time-series data (continuous changes in interstitial fluid glucose concentration, continuous glucose monitoring (CGM)); sleep physiological signals (physiological data that distinguishes sleep stages such as light sleep, deep sleep, and REM sleep); and heart rate variability parameters (indicators reflecting the degree of autonomic nervous system excitation). The combination of these three types of heterogeneous data constitutes multimodal monitoring data.

[0026] Interstitial fluid glucose concentration refers to the glucose content in the intercellular fluid of human tissues. It is a direct indicator detected by continuous glucose monitoring (CGM) devices. Its value changes lag behind blood glucose and can continuously and dynamically reflect changes in blood glucose metabolism in the body, serving as the basis for non-invasive blood glucose monitoring.

[0027] Glucose time series data refers to the sequence of glucose values ​​in interstitial fluid collected continuously at fixed sampling time intervals, with timestamps, which can characterize the continuous change of blood glucose over time.

[0028] Sleep physiological signals refer to the brain, body movement, respiration, and electrocardiogram-derived signals collected by sleep monitoring equipment, which are used to automatically classify sleep stages such as light sleep, deep sleep, and REM sleep.

[0029] Sleep stages refer to the sleep stages classified according to sleep physiological signals. They are divided into the waking stage, light sleep stage, deep sleep stage, and rapid eye movement (REM) sleep stage. The secretion patterns of insulin and glucagon differ in different sleep stages.

[0030] The level of autonomic nervous system excitation refers to the activity level of the sympathetic and parasympathetic nervous systems. Fluctuations in the autonomic nervous system during pregnancy can affect liver glucose output and insulin sensitivity, indirectly inducing a sudden drop in blood sugar at night.

[0031] Heart rate variability (HRV) is a quantitative indicator of the subtle fluctuations in successive heartbeat intervals. It is a general parameter for assessing the excitability of the autonomic nervous system and includes derived values ​​such as the standard deviation of all sinus RR intervals (SDNN) and the root mean square of successive differences (RMSSD) between adjacent RR intervals.

[0032] It should be noted that, in the embodiments of the present invention, by integrating sleep stage and heart rate variability parameters, and combining the physiological effects of sleep rhythm and autonomic nerve fluctuations on gestational blood glucose, false alarms and missed alarms are reduced; at the same time, relying on multimodal monitoring data of glucose, sleep, and heart rate variability, the individualized glucose metabolism characteristics of pregnant women at different gestational weeks are accurately characterized, providing reliable data support for dynamically correcting individualized early warning thresholds based on gestational week standard baselines, and adapting to the needs of physiological dynamic changes throughout the entire pregnancy cycle.

[0033] Step S2: Construct blood glucose baselines for each standard gestational week based on historical databases and determine the initial hypoglycemia threshold for each gestational week. The standard gestational week blood glucose baseline is used to characterize the average nocturnal blood glucose metabolism level of the reference group in the same gestational week, and the initial hypoglycemia threshold is used to characterize the safe threshold value of nocturnal blood glucose metabolism of the reference group as the gestational week changes dynamically.

[0034] The historical database refers to a database that collects multimodal historical monitoring data of women at different gestational weeks. It contains a large amount of measured sample data such as glucose time series, sleep physiology, heart rate variability, and gestational week parameters of the reference population. It is the data source for building baselines and thresholds and can filter and aggregate samples by gestational week, sleep period, and sampling time.

[0035] The standard gestational blood glucose baseline refers to a time-series blood glucose baseline curve generated based on actual nighttime blood glucose measurement data of the same gestational week reference group, through time alignment, point averaging, and curve fitting; it represents the average nighttime glucose metabolism level of a group of healthy pregnant women at the same gestational week, which changes dynamically with the number of days of pregnancy and nighttime time, and serves as a reference benchmark for individualized blood glucose deviation comparison.

[0036] Gestational week refers to the discrete division of the gestational period. Based on the time span rules, the gestational period is divided into multiple time windows. It is the core dividing dimension to distinguish the differences in physiological metabolism during pregnancy (pregnancy hormones and insulin sensitivity change with gestational week, and blood glucose benchmarks change synchronously).

[0037] The initial hypoglycemia threshold refers to the dynamic critical value curve formed by sampling points based on the blood glucose data of the reference population at the corresponding gestational week, taking preset percentile values, and fitting the curve. It is a general safety threshold for nocturnal hypoglycemia at the population level, without taking into account the individual physical differences of pregnant women, and does not provide a benchmark value for subsequent personalized threshold correction.

[0038] The reference group refers to a collection of pregnant women in the same gestational age group, aggregated according to the gestational week time window. It is generated by collecting candidate individual data from historical databases, removing extreme abnormal samples, and represents the general pattern of physiological blood glucose in pregnant women at that gestational week.

[0039] The safe threshold for nocturnal glucose metabolism is a critical blood glucose concentration determined based on big data statistics of a reference population at the same gestational age. This threshold changes dynamically with gestational age and the time of nighttime monitoring. When a pregnant woman's blood glucose is below this threshold, it indicates that the probability of nocturnal hypoglycemia is significantly increased. It is the population benchmark limit for assessing the risk of nocturnal hypoglycemia during pregnancy.

[0040] Furthermore, the step of constructing standard gestational blood glucose baselines based on historical databases includes: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point.

[0041] The preset gestational week range is a segmented gestational week range that is pre-divided based on the physiological and metabolic changes during pregnancy, such as 20-24 weeks of pregnancy and 28-32 weeks of pregnancy, which is used to screen reference samples that match the gestational week.

[0042] The sampling time points are the fixed times when the equipment collects blood glucose data during nighttime monitoring, such as 23:00, 01:00, and 03:00 at night. A unified time coordinate is a prerequisite for data alignment processing.

[0043] Alignment processing involves uniformly calibrating the glucose time-series data of all individuals within the same reference population to the same sampling time axis, eliminating individual sampling time deviations, and ensuring that the data of each sample at the same time node are available for comparative analysis.

[0044] For each sampling time point, a set of multiple glucose concentration values ​​corresponding to each sampling time point of the reference population is extracted.

[0045] The glucose concentration value set refers to the collection of blood glucose values ​​corresponding to all subjects in the reference group at the same sampling time point.

[0046] The mean value of each glucose concentration value set is calculated one by one, and the calculated mean value is determined as the baseline blood glucose value at each sampling time point.

[0047] The baseline blood glucose value refers to the arithmetic mean of the blood glucose set at a single sampling time point, representing the average blood glucose level of the population at that gestational week.

[0048] The baseline blood glucose values ​​corresponding to all sampling time points were fitted to generate a standard gestational week blood glucose baseline that changes over time.

[0049] Fitting refers to using mathematical curve fitting algorithms to generate a continuous curve from discrete time-point baseline blood glucose values.

[0050] It should be noted that, in the embodiments of the present invention, by using the mean of a large sample of the same gestational age to generate a baseline, the interference caused by individual special cases can be eliminated, forming a standardized reference that fits the physiological characteristics of pregnancy; this baseline, as a prerequisite for deviation calculation and individualized early warning threshold correction, can quantify the degree of deviation of the subject's blood glucose relative to the same gestational age group, overcoming the drawback that fixed thresholds cannot adapt to the metabolic changes in stages of pregnancy.

[0051] Furthermore, the step of determining the initial hypoglycemia threshold for each gestational week includes: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point.

[0052] For example, from a historical database storing a large amount of pregnancy monitoring information, a reference group sample corresponding to a preset gestational week interval is selected, and the glucose time-series monitoring data recorded throughout the entire process of this sample is extracted; then, using a unified sampling time as a reference coordinate, time-series alignment processing is performed on the glucose time-series data of all individuals to correct the problem of misaligned collection times of each sample, so that all data correspond one-to-one at the same time node.

[0053] For each sampling time point, a set of multiple glucose concentration values ​​corresponding to each sampling time point of the reference population is extracted.

[0054] For example, by traversing all sampling time points after time alignment, for a single fixed sampling time, the blood glucose test values ​​of all subjects at that time point are summarized from the regularized reference population data, thereby constructing a glucose concentration dataset corresponding to each sampling time point.

[0055] The preset percentile values ​​of each glucose concentration value set are calculated one by one, and the calculated preset percentile values ​​are fitted to generate an initial hypoglycemia warning threshold curve that changes over time. The values ​​corresponding to each sampling time point on the initial hypoglycemia warning threshold curve are determined as the initial hypoglycemia threshold.

[0056] The preset percentile value refers to the pre-selected statistical percentile (e.g., P5 and P10 percentiles are commonly used for clinical hypoglycemia), which represents the critical threshold value for low blood glucose at this time point in the same gestational age group, avoiding interference from extreme abnormal data.

[0057] The initial hypoglycemia warning threshold curve refers to a continuous dynamic curve generated by fitting a function with sampling time as the horizontal axis and percentile blood glucose value as the vertical axis. It reflects the changing pattern of the basic hypoglycemia threshold at different times of the night and with gestational week.

[0058] It should be noted that, in the embodiments of the present invention, a dynamic initial threshold is constructed by fitting percentiles by gestational week and time period, which can avoid interference from extreme blood glucose data and improve the scientific validity of the threshold; data standardization is achieved by aligning sampling time, ensuring horizontal comparison and aggregation modeling of sample data; this initial threshold serves as the basis for personalized threshold correction, which can reduce threshold correction error, and combined with the physiological patterns of nighttime, it can reduce false alarms and missed alarms of hypoglycemia; at the same time, the standardized processing procedure facilitates continuous expansion of samples and iterative optimization of thresholds, adapting to the clinical needs of diverse populations.

[0059] Step S3: Extract the blood glucose fluctuation characteristics of the target object based on multimodal monitoring data, and calculate the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational week blood glucose baseline. The blood glucose fluctuation characteristics include multidimensional feature vectors of the target object in different sleep stages at night. The deviation is used to characterize the degree of abnormality of the target object's nighttime glucose metabolism state relative to the standard physiological pattern.

[0060] Among them, the blood glucose fluctuation feature is a multi-dimensional feature vector obtained by splitting blood glucose data according to each sleep stage, integrating three types of indicators: blood glucose statistics, blood glucose extreme values, and blood glucose trends, and then normalizing and weighting them. The blood glucose statistics feature includes average blood glucose, standard deviation, and coefficient of variation; the blood glucose extreme value feature includes the nighttime blood glucose maximum value and its corresponding time; and the blood glucose trend feature includes the rate of blood glucose decline and the duration of hypoglycemia. It can characterize the nighttime blood glucose change characteristics of the subjects in stages.

[0061] The standard fluctuation feature is a reference feature selected within the time window corresponding to the monitoring period of the subjects, and is composed of the baseline mean blood glucose, the baseline blood glucose standard deviation, and the baseline blood glucose trend slope extracted from the baseline blood glucose at standard gestational weeks.

[0062] Deviation is a quantitative value obtained by calculating the deviation between individual characteristics and standard characteristics item by item, combined with weights that are positively correlated with sleep depth. The higher the value, the higher the deviation of glucose metabolism from the normal level and the higher the risk of hypoglycemia.

[0063] Nighttime sleep is divided into light sleep, deep sleep, and REM sleep based on sleep physiological signals. Each stage has different characteristics in hormone and autonomic nervous system regulation, and different physiological fluctuations in blood sugar. The feature weights are adjusted accordingly. Deep sleep has increased weights for extreme values ​​and trend features, while REM sleep has more emphasis on statistical feature weights.

[0064] Multidimensional feature vectors (MVAs) are structured data that aggregates three types of blood glucose indicators based on sleep stages, serving as a digital carrier for quantifying blood glucose fluctuations.

[0065] The standard physiological pattern is the normal nighttime blood glucose variation pattern formed by healthy individuals at the same gestational age based on the gestational age baseline, and is used as a reference for determining abnormal glucose metabolism.

[0066] It should be noted that, in the embodiments of the present invention, the warning threshold is dynamically adjusted based on the baseline of big data at the same gestational age and the individual deviation, which effectively reduces false alarms and missed alarms; blood glucose characteristics are broken down according to sleep stages and weights are allocated differently, which conforms to the blood glucose regulation rules of hormones and autonomic nervous system at different sleep stages, thus improving the scientific nature of abnormality identification; the deviation is used to quantify the metabolic differences between individuals and groups, so as to realize the personalized customization of the warning threshold and adapt to the individual physical differences of pregnant women; asymptomatic hypoglycemia in deep sleep can be detected early, reducing the risk of pregnancy complications such as fainting in pregnant women and fetal developmental abnormalities.

[0067] Furthermore, the step of extracting blood glucose fluctuation features of the target object based on multimodal monitoring data includes: Preprocessing of glucose time-series data from multimodal monitoring data yields standardized glucose time-series data, standardized sleep physiological signals, and standardized heart rate variability parameters.

[0068] Preprocessing is used to remove missing data, abnormal mutations, and errors caused by equipment drift from the original monitoring data. Then, normalization and dimensional unification are used to eliminate data interference caused by equipment model and individual basic physiological conditions, unify the data value range and measurement unit, form standardized data, and ensure that data of multiple types and different time periods can be calculated collaboratively.

[0069] Standardized sleep physiological signals are obtained by noise reduction and standardization of raw sleep data. The data comes from monitoring signals such as body movement, electrocardiogram, and electroencephalogram, and can be used to distinguish sleep stages such as light sleep, deep sleep, and REM sleep.

[0070] Standardized heart rate variability (SFR) parameters are standardized conversions of inter-heart rate fluctuations. They can reflect the level of autonomic nervous system excitation and help determine whether abnormal blood glucose levels are caused by autonomic nervous system fluctuations.

[0071] Sleep stage status during nighttime monitoring periods is identified based on standardized sleep physiological signals.

[0072] For example, based on the sleep physiological signals after noise reduction and standardization, signal feature parameters are extracted, and the monitoring intervals throughout the night are analyzed one by one according to the sleep physiological judgment criteria to distinguish light sleep, deep sleep and REM sleep, thus completing the sleep stage division of the entire night cycle.

[0073] Standardized glucose time-series data were segmented according to sleep stages, and statistical features, extreme blood glucose features, and blood glucose trend features were extracted for each sleep stage. The statistical features included the nighttime average blood glucose value, blood glucose standard deviation, and coefficient of variation. The extreme blood glucose features included the highest and lowest blood glucose values ​​at night and their corresponding time points. The blood glucose trend features included the rate of blood glucose decline and the duration of hypoglycemia.

[0074] Among them, blood glucose statistical characteristics are quantitative parameters that reflect the overall distribution of blood glucose in a single sleep stage, including the average blood glucose level, blood glucose standard deviation, and coefficient of variation for each stage, which are used to characterize the average level and stability of blood glucose fluctuations in the corresponding sleep stage.

[0075] Blood glucose extreme value characteristics cover the maximum and minimum blood glucose values ​​in the current period and their corresponding occurrence times, accurately pinpointing the timing of blood glucose peaks and troughs.

[0076] Blood glucose trend characteristics include the rate of blood glucose decline and the duration of hypoglycemia. These characteristics can quantify the speed of blood glucose decline and the duration of hypoglycemia, and are key criteria for early warning of nocturnal hypoglycemia during pregnancy.

[0077] The statistical characteristics, extreme values, and trends of blood glucose in different sleep stages are fused to obtain the blood glucose fluctuation characteristics of the target subject.

[0078] For example, firstly, blood glucose statistics, extreme values, and trends are extracted for each stage of light sleep, deep sleep, and REM sleep. Normalization is then performed on each of these three types of data to eliminate dimensional differences. Next, a preset feature weight matrix is ​​matched based on the sleep stage type, and the weights of the statistics, extreme values, and trends are dynamically configured according to circadian rhythm rules (increasing the weights of extreme values ​​and trends during deep sleep, and emphasizing the weights of statistics during REM sleep). Subsequently, multi-dimensional feature fusion is achieved through weighted operations, integrating all blood glucose parameter information from different sleep stages, and finally generating a multi-dimensional blood glucose fluctuation feature vector that comprehensively reflects the nighttime segmented blood glucose variation patterns of the target pregnant woman.

[0079] It should be noted that, in the embodiments of the present invention, blood glucose features are extracted based on sleep stages to reduce false alarms; multi-dimensional features are constructed by fusing multiple types of blood glucose parameters, and multi-modal data of blood glucose, sleep, and heart rate variability are combined to improve the accuracy of hypoglycemia determination.

[0080] Furthermore, the step of calculating the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational age blood glucose baseline includes: Based on the standard gestational blood glucose baseline, a time window matching the nighttime monitoring period of the target subjects was extracted, and the baseline mean blood glucose, baseline standard deviation blood glucose, and baseline blood glucose trend slope within the time window were calculated to form the standard fluctuation characteristics.

[0081] The time window is a continuous period of time extracted from the standard gestational blood glucose baseline and completely corresponding to the actual nighttime monitoring time of the pregnant woman to be tested, so as to achieve the time dimension matching between individual and reference data and ensure the time sequence comparability of blood glucose data.

[0082] The baseline mean blood glucose is the average blood glucose level of the reference population at the same gestational week within the corresponding time window, reflecting the overall baseline blood glucose level of the healthy population during that monitoring period.

[0083] The baseline blood glucose standard deviation is used to reflect the dispersion of blood glucose data in the reference group within this time window, and intuitively reflects the range of blood glucose fluctuations in healthy pregnant women at the same gestational age under normal physiological conditions.

[0084] The baseline blood glucose trend slope represents the rate of increase or decrease of blood glucose over time in the reference group within the window period, serving as a reference for the natural trend of nighttime blood glucose changes in healthy individuals.

[0085] The blood glucose statistical characteristics, blood glucose trend characteristics, and baseline blood glucose mean, baseline blood glucose standard deviation, and baseline blood glucose trend slope in the blood glucose fluctuation characteristics of the target object are calculated to obtain the blood glucose mean deviation, blood glucose fluctuation amplitude deviation, and blood glucose change trend deviation.

[0086] For example, the nighttime average blood glucose value, blood glucose standard deviation, and blood glucose decrease rate extracted from the target object are compared with the baseline blood glucose mean, baseline blood glucose standard deviation, and baseline blood glucose trend slope in the standard fluctuation characteristics to calculate the difference, blood glucose mean deviation, blood glucose fluctuation amplitude deviation, and blood glucose change trend deviation.

[0087] Among them, the blood glucose fluctuation characteristics of the target subjects were extracted according to different sleep stages, including three types of parameters: statistical, extreme values, and trend. The statistical features include the mean blood glucose, standard deviation, and coefficient of variation; the extreme value features include the highest and lowest blood glucose values ​​in each stage and their corresponding occurrence times; the trend features include the rate of blood glucose decline and the duration of hypoglycemia, which comprehensively characterizes the blood glucose change characteristics of the subjects in multiple dimensions.

[0088] The mean blood glucose deviation is calculated by subtracting the baseline mean blood glucose value from the corresponding standard fluctuation characteristics from the nighttime average blood glucose value in the blood glucose statistical characteristics. It is used to quantify the deviation of the overall average blood glucose level of the test subjects from the standard reference value.

[0089] Blood glucose fluctuation deviation is the difference between the standard deviation of blood glucose and the benchmark blood glucose standard deviation in blood glucose statistical characteristics. It is used to reflect the abnormality of the blood glucose fluctuation of the subject compared with the normal group.

[0090] The deviation in blood glucose trend is obtained by subtracting the baseline blood glucose trend slope from the rate of blood glucose decline in the blood glucose trend characteristics. It can intuitively reflect the degree to which the rate of blood glucose rise and fall in the tested pregnant woman deviates from the normal variation pattern of healthy people.

[0091] For example, for the nighttime monitoring data of the target pregnant woman after data cleaning and timestamp synchronization, three types of standardized blood glucose characteristic parameters are extracted: the nighttime average blood glucose value representing the overall blood glucose level, the blood glucose standard deviation reflecting the degree of blood glucose dispersion and fluctuation, and the blood glucose decline rate reflecting the nighttime blood glucose decline pattern. The standard blood glucose baseline corresponding to the pregnant woman's gestational week is retrieved, and the matching nighttime monitoring period from 22:00 to 6:00 the next day is selected to obtain three sets of reference parameters corresponding to the standard fluctuation characteristics (baseline mean blood glucose, baseline standard deviation blood glucose, and baseline blood glucose trend slope). Differences are calculated using a one-to-one correspondence method for parameters of the same type. Subtracting the baseline reference parameters from the individual blood glucose characteristic parameters yields three types of deviation indicators: the difference between the nighttime average blood glucose value and the baseline mean blood glucose is the blood glucose mean deviation; the difference between the blood glucose standard deviation and the baseline blood glucose standard deviation is the blood glucose fluctuation amplitude deviation; and the difference between the blood glucose decline rate and the baseline blood glucose trend slope is the blood glucose change trend deviation.

[0092] Based on the statistical characteristics, extreme values, and trends of blood glucose in different sleep stages at night, the weight coefficients for each feature are determined. The weight coefficients are positively correlated with the depth of the sleep stage.

[0093] Among them, the depth of sleep stage represents the physiological stage of sleep transitioning from light sleep to deep sleep. As sleep depth increases, the influence of the autonomic nervous system and various pregnancy regulatory hormones on blood sugar increases. Therefore, the feature weight coefficient is positively correlated with sleep depth. The deeper the sleep, the higher the weight of the corresponding blood sugar feature.

[0094] The weighting coefficient is the weighting ratio parameter assigned to the three types of deviation indicators: mean, extreme value, and trend of blood glucose. During deep sleep, the weighting ratio corresponding to extreme value and trend characteristics of blood glucose will increase.

[0095] For example, the weighting coefficient is a parameter that characterizes the importance of each blood glucose indicator during feature fusion. It is set differently according to different stages of deep sleep and REM sleep: during deep sleep, the weights of extreme values ​​and trend features are increased first, while during REM sleep, the weights of statistical features are increased.

[0096] The deviation of the mean blood glucose deviation, the blood glucose fluctuation amplitude deviation, and the blood glucose change trend deviation are weighted and summed based on the weighting coefficients to obtain the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational week blood glucose baseline.

[0097] It should be noted that, in the embodiments of the present invention, before performing weighted summation of the mean deviation of blood glucose, the amplitude deviation of blood glucose fluctuation, and the trend deviation of blood glucose change based on the weighting coefficients, the deviation distribution of the reference group in the same gestational period is used as a benchmark. The mean deviation of blood glucose, the amplitude deviation of blood glucose fluctuation, and the trend deviation of blood glucose change are processed by the maximum-minimum value normalization algorithm, and all deviations are uniformly mapped to the dimension range of 0 to 1, eliminating the differences in dimensions and scales between different indicators, so that each deviation has a unified comparison scale, and providing dimensionless input data for subsequent weighted fusion.

[0098] For example, according to the corresponding feature weight coefficients matched for different sleep stages, the deviation of mean blood glucose, deviation of blood glucose fluctuation amplitude, and deviation of blood glucose change trend are weighted and assigned values ​​one by one. Then, the weighted deviation values ​​are summed to finally quantify the overall deviation of the individual blood glucose fluctuation characteristics from the standard fluctuation characteristics of the same gestational week. This objectively reflects the degree of abnormality of the tested pregnant woman's nocturnal glucose metabolism relative to the normal physiological level.

[0099] It should be noted that, in the embodiments of the present invention, time sequence alignment is completed by relying on the matching time window to eliminate the comparison error caused by the difference in monitoring time periods; the deviation is calculated from three dimensions of average blood glucose, fluctuation amplitude, and rate of change, and the feature weights are dynamically configured in combination with sleep depth to conform to the physiological laws of hormones and glucose metabolism during pregnancy and improve the accuracy of deviation calculation.

[0100] Furthermore, the step of fusing the statistical characteristics, extreme values, and trends of blood glucose in each sleep stage to obtain the blood glucose fluctuation characteristics of the target object includes: The statistical characteristics, extreme values, and trend characteristics of blood glucose were normalized to obtain the normalized values ​​of blood glucose statistics, extreme values, and trend.

[0101] Normalization refers to converting characteristic data with different dimensions and numerical ranges into a unified numerical range, eliminating differences in units and numerical magnitudes, and facilitating subsequent weighted calculations.

[0102] Obtain the feature weight matrix, which includes the weight coefficients corresponding to blood glucose statistical features, the weight coefficients corresponding to blood glucose extreme value features, and the weight coefficients corresponding to blood glucose trend features.

[0103] The feature weight matrix is ​​a parameter matrix constructed based on the basic weights and after real-time sleep stage correction. It stores the corresponding weights of the three types of blood glucose features—statistics, extreme values, and trends—under each sleep stage, providing a basis for feature fusion.

[0104] Based on the feature weight matrix, the blood glucose statistical normalization value, blood glucose extreme value normalization value, and blood glucose trend normalization value are weighted and fused to obtain the blood glucose fluctuation characteristics of the target object.

[0105] Weighted fusion refers to the weighted operation and summarization of the three types of blood glucose features that have completed normalization, based on the allocation ratio of the weight matrix, integrating the scattered individual feature information, and finally forming a complete and unified blood glucose fluctuation feature of the target object.

[0106] It should be noted that in the embodiments of this invention, normalization is used to unify the dimensions of each feature to prevent differences in data magnitude from interfering with the fusion results; and the weights are dynamically configured based on sleep stages, which conforms to the physiological changes during pregnancy. The fusion of statistical, extreme value, and trend indicators improves blood glucose characteristics, comprehensively reflects glucose metabolism, provides reliable support for deviation calculation and threshold correction, and effectively improves the early warning effect for hypoglycemia.

[0107] Furthermore, the steps for constructing the feature weight matrix include: Construct a basic weight mapping table, which includes the baseline weight coefficients corresponding to each feature dimension. The feature dimensions include at least blood glucose statistical features, blood glucose extreme value features, and blood glucose trend features.

[0108] The basic weight mapping table is a pre-configured data table, in which the baseline weight parameters of three types of blood glucose characteristics—statistics, extreme values, and trends—are pre-entered as the original reference for subsequent dynamic adjustment of weights.

[0109] The feature dimensions are divided into three types of indicators: blood glucose statistics, blood glucose extreme values, and blood glucose trends, which constitute the basic classification for constructing fluctuation features.

[0110] The baseline weighting coefficient is a fixed weighting value without sleep state adjustment, representing the inherent contribution ratio of various characteristics to the evaluation of blood glucose fluctuations under normal physiological conditions.

[0111] The sleep stage status corresponding to each sampling point is obtained, and the baseline weight coefficients in the basic weight mapping table are corrected according to the preset physiological rhythm adjustment rules to obtain the feature weight matrix. The preset physiological rhythm adjustment rules include: when the sleep stage status indicates deep sleep, the weight coefficients associated with extreme blood glucose features and blood glucose trend features are adjusted to be greater than their corresponding baseline weight coefficients; when the sleep stage status indicates REM sleep, the weight coefficients associated with blood glucose statistical features are adjusted to be greater than their corresponding baseline weight coefficients.

[0112] The sampling points are where the CGM monitoring equipment collects multi-source data on glucose, sleep physiology, and heart rate variability at various time points according to a fixed sampling cycle, ensuring that all monitoring data are synchronized in time.

[0113] The physiological rhythm regulation rules are set in conjunction with the mechanism of the relationship between sleep and glucose metabolism during pregnancy, and the feature weight ratio is adjusted in a targeted manner according to the physiological differences in different sleep stages.

[0114] During deep sleep, the sleep level is deeper and the body's autonomic nervous system activity tends to be stable. During this stage, blood sugar is easily affected by hormones, resulting in a rapid drop in blood sugar, making it a high-risk period for hypoglycemia.

[0115] REM sleep is a shallow sleep stage, with higher levels of neurological excitability. It is also affected by endocrine fluctuations, making it easy for average blood sugar levels to fluctuate significantly.

[0116] It should be noted that, in the embodiments of the present invention, the feature weights are dynamically adjusted according to the physiological rhythm of sleep. Deep sleep is emphasized based on extreme values ​​and trend indicators, while REM sleep is emphasized based on statistical indicators, replacing the fixed weight model and aligning with the metabolic patterns during pregnancy. The blood glucose fluctuation features obtained after weight optimization are more accurate, effectively reducing the error in deviation calculation. The resulting personalized warning threshold is more reasonable, reducing false negatives and missed warnings for hypoglycemia.

[0117] Step S4: Correct the initial hypoglycemia threshold according to the deviation, determine the warning threshold for the target object, and issue a nighttime hypoglycemia warning for the target object based on the warning threshold.

[0118] Among them, threshold correction refers to the calculation process of determining the threshold adjustment amount based on the individual's deviation in glucose metabolism and the preset correction coefficient, and then adjusting the general initial hypoglycemia threshold of the group upward or downward accordingly, so as to realize the conversion from the group reference standard of gestational age to the individualized judgment standard of pregnant women.

[0119] The warning threshold is generated based on the baseline at the same gestational age and the individual deviation correction. It is a specific hypoglycemic threshold value, unlike a general fixed threshold, and can dynamically change with the progress of gestation and individual metabolic differences.

[0120] Nighttime hypoglycemia alert refers to the real-time collection of blood glucose data and comparison with personalized alert thresholds. When blood glucose is below the threshold and remains below the threshold for a specified duration, an alert is triggered to promptly detect hidden hypoglycemia without noticeable symptoms during sleep.

[0121] It should be noted that, in the embodiments of this invention, the deviation is calculated based on multimodal data of sleep and heart rate variability, and dynamic parameter adjustments are made in conjunction with different sleep physiological states to adapt to natural fluctuations in blood glucose at night. The threshold is updated in real time with gestational age and physical condition, effectively improving the problems of false alarms and missed alarms.

[0122] Furthermore, the step of correcting the initial hypoglycemia threshold based on the deviation and determining the warning threshold for the target object includes: The deviation is combined with a preset correction coefficient to obtain the threshold adjustment amount.

[0123] The preset correction coefficient is set by comprehensively considering the physiological characteristics of different gestational weeks, the metabolic patterns of each sleep stage, and the hormonal changes during pregnancy. It is used to quantify the degree of deviation on the warning threshold and constrain the adjustment range of the threshold.

[0124] The threshold adjustment amount is obtained by corresponding calculations of the deviation and the preset correction coefficient. This value intuitively reflects the specific amount that needs to be adjusted up or down for the initial hypoglycemia threshold of the group, and is the calculation parameter for completing individualized threshold correction.

[0125] For example, the deviation is a glucose metabolism evaluation index obtained by comprehensively analyzing multidimensional blood glucose characteristics, including an absolute difference index for blood glucose troughs and an average blood glucose difference index. Both the absolute difference index for blood glucose troughs and the average blood glucose difference index are calculated by comparing the individual's nighttime blood glucose curve (by fitting multimodal monitoring data to generate the individual's nighttime blood glucose curve corresponding to the target object) with a standard gestational week blood glucose baseline. The absolute difference index for blood glucose troughs and the average blood glucose difference are extracted from the deviation, multiplied by a preset correction weight coefficient, and then summed and fused to obtain the threshold adjustment amount.

[0126] The initial hypoglycemia threshold is corrected based on the threshold adjustment amount to obtain the warning threshold for the target object.

[0127] For example, the calculated threshold adjustment amount is used to adjust the initial hypoglycemia threshold of the population determined by data from the same gestational age group, and the population baseline value is corrected by combining the pregnant woman's own glucose metabolism deviation. The individual adaptation bias caused by the general threshold is eliminated, and finally, a unique personalized hypoglycemia judgment standard that is adapted to the gestational age, metabolic characteristics and nighttime sleep patterns of the tested pregnant woman is generated, which is the warning threshold of the target object.

[0128] Furthermore, the step of providing a nighttime hypoglycemia warning for the target object based on a warning threshold includes: Real-time acquisition of glucose concentration values ​​of the target object.

[0129] Among them, the glucose concentration value refers to the real-time value of interstitial fluid glucose collected by continuous blood glucose monitoring equipment, which is the raw data for determining hypoglycemia.

[0130] A hypoglycemia warning signal is generated when the detected glucose concentration is below the warning threshold and the duration reaches the preset duration threshold.

[0131] The preset duration threshold is set in advance based on the physiological characteristics of pregnancy to filter out short-term low blood glucose caused by instantaneous physiological fluctuations and sampling disturbances; only when blood glucose is continuously lower than the personalized warning threshold and the duration threshold is met is it determined to be clinical true hypoglycemia.

[0132] A hypoglycemia warning signal refers to a digital warning command generated by the system when both blood glucose level and duration conditions are met simultaneously. This triggers multiple levels of alerts, including device audio and visual prompts, pop-up windows, and push notifications from mobile apps.

[0133] The preset output device is controlled to perform the warning operation corresponding to the low blood sugar warning signal. The warning operation includes at least one of the following: audible and visual alarm, vibration reminder, and sending a warning message to the associated terminal.

[0134] Among them, the output devices are various hardware devices that carry the low blood sugar warning and reminder function. They are divided into three categories according to the usage scenario: wearable terminals, home monitoring terminals, and hospital medical management terminals, so as to realize alarm prompts.

[0135] For example, the audible and visual alarm relies on the intermittent flashing of the equipment indicator light combined with the sound of a buzzer to provide a close-range and intuitive early warning at the equipment deployment site.

[0136] For example, vibration alerts refer to those integrated into wearable monitoring devices, which rely on a built-in vibration motor to generate mechanical vibrations, reminding the pregnant woman wearing the device to check her blood sugar level in a timely manner.

[0137] For example, the associated terminal is an external mobile device that is bound to the blood glucose monitoring device via a communication link, including the pregnant woman's own device, the family member's bound mobile phone, the monitoring tablet used by medical staff, and the obstetrics back-end management system.

[0138] Furthermore, the method also includes: The gestational cycle parameters of each candidate individual in the historical sample set are obtained, and the continuous gestational cycles are divided into multiple discrete time windows according to the preset time span rules.

[0139] Among them, the historical sample set refers to the collection of massive historical records of pregnant women accumulated over a long period of time. It includes raw data on blood glucose, sleep, and heart rate variability monitoring and pregnancy information of different gestational weeks and healthy / gestational diabetes populations. It is the data source for constructing the gestational week baseline.

[0140] Candidate individuals refer to pregnant women who participated in the study during their pregnancy and were included in the historical sample set, serving as the basic sample for constructing the reference population.

[0141] Pregnancy cycle parameters refer to the actual gestational week data of a pregnant woman, with the number of days of pregnancy / gestational week as the core indicator, to distinguish different physiological stages of early, mid and late pregnancy.

[0142] The preset time span rule refers to the pre-set gestational week segmentation standard (such as 1 week / 2 weeks as a unit) used to divide the continuous pregnancy period and is the basis for dividing time windows.

[0143] Discrete time windows refer to non-overlapping gestational week intervals after being divided according to span rules, such as 20-21 weeks of gestation or 21-22 weeks of gestation. They are the smallest statistical units for data collection.

[0144] The multimodal monitoring data of each candidate individual in the historical sample set are mapped to a target time window that matches their gestational cycle parameters. The multimodal monitoring data of all candidate individuals within the same target time window are aggregated to generate a reference population corresponding to each gestational cycle.

[0145] Data mapping refers to the data matching operation that categorizes individual monitoring data into the corresponding gestational week time window based on the pregnant woman's actual gestational week.

[0146] Data aggregation refers to the summarization, collection, and integration of all pregnant women's data within the same time window, including the aggregation and calculation of blood glucose, sleep, and heart rate data for samples at the same gestational week.

[0147] It should be noted that, in the embodiments of the present invention, the gestational period is divided into discrete gestational week windows to adapt to the dynamic changes in hormones and metabolism during pregnancy. Samples are aggregated and matched according to gestational week to improve the homogeneity of the reference population; a standardized baseline is constructed based on large samples within the same window to mitigate individual and monitoring errors. The resulting population benchmark is used for individualized threshold correction, effectively improving the problem of false alarms and missed alarms.

[0148] It is important to emphasize that the nocturnal hypoglycemia warning results, individualized warning thresholds, and glucose metabolism deviation data output by this invention are only used to quantitatively assess the risk of nocturnal blood glucose fluctuations in pregnant women and to assist in screening for occult hypoglycemia. This provides intelligent reference data for medical personnel and effectively improves the efficiency of blood glucose risk identification during pregnancy. However, the final diagnosis, risk classification, and subsequent clinical interventions and blood glucose control for pregnant women must be determined by professional medical personnel based on clinical experience, the pregnant woman's overall health condition, and actual examination results. The algorithm results of this invention do not replace professional clinical diagnosis and treatment decisions.

[0149] In summary, this invention integrates three types of multi-source physiological monitoring data: glucose time-series data, sleep stage signals, and heart rate variability. Based on a sample database, it divides gestational weeks into intervals, constructs differentiated standard blood glucose baselines, and generates initial hypoglycemia thresholds suitable for each gestational week. Based on sleep physiological stages such as deep sleep and REM sleep, it extracts multi-dimensional blood glucose fluctuation characteristics (statistical, extreme values, and trends) hierarchically, quantifies the deviation of the target individual's glucose metabolism from the baseline for the same gestational week, and dynamically corrects the deviation to generate personalized warning thresholds, achieving intelligent judgment and warning of nocturnal hypoglycemia. This invention aligns with the hormonal and metabolic changes during pregnancy, and the warning thresholds can be dynamically updated with gestational week progression and individual metabolic levels, reducing false alarms and missed alarms for nocturnal hypoglycemia warnings and improving the effectiveness of nocturnal hypoglycemia warnings during pregnancy.

[0150] This invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described functions. Figure 1 Any step in the corresponding method embodiment can achieve the same technical effect, and will not be repeated here to avoid repetition.

[0151] The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: 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 or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0152] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0153] The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0154] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or terminal. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0155] This invention also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to achieve the method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring provided in the above embodiments.

Claims

1. A method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring, characterized in that, The method includes: Acquire multimodal monitoring data of the target object during nighttime sleep, wherein the multimodal monitoring data includes glucose time-series data characterizing changes in interstitial fluid glucose concentration, sleep physiological signals characterizing sleep stage status, and heart rate variability parameters characterizing the degree of autonomic nervous system excitation; Based on historical databases, standard gestational blood glucose baselines are constructed for each gestational week, and initial hypoglycemia thresholds for each gestational week are determined. The standard gestational blood glucose baselines are used to characterize the average nocturnal blood glucose metabolism level of the reference group in the same gestational week, and the initial hypoglycemia thresholds are used to characterize the safe threshold value of nocturnal blood glucose metabolism of the reference group as it dynamically changes with gestational week. Based on the multimodal monitoring data, the blood glucose fluctuation characteristics of the target object are extracted, and the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational blood glucose baseline is calculated. The blood glucose fluctuation characteristics include the multidimensional feature vectors of the target object in different sleep stages at night, and the deviation is used to characterize the degree of abnormality of the target object's nighttime glucose metabolism state relative to the standard physiological pattern. The initial hypoglycemia threshold is corrected based on the deviation to determine the warning threshold for the target object, and a nighttime hypoglycemia warning is issued for the target object based on the warning threshold.

2. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The steps for constructing standard gestational blood glucose baselines based on historical databases include: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point. For each sampling time point, extract a set of multiple glucose concentration values ​​corresponding to each sampling time point for the reference population; The mean value of each glucose concentration value set is calculated one by one, and the calculated mean value is determined as the baseline blood glucose value at each sampling time point. The baseline blood glucose values ​​corresponding to all sampling time points were fitted to generate a standard gestational week blood glucose baseline that changes over time.

3. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The step of extracting blood glucose fluctuation features of the target object based on the multimodal monitoring data includes: The glucose time-series data in the multimodal monitoring data are preprocessed to obtain standardized glucose time-series data, standardized sleep physiological signals, and standardized heart rate variability parameters. Based on the standardized sleep physiological signals, sleep stage status during the nighttime monitoring period is identified; The standardized glucose time-series data is segmented according to the sleep stage state, and the blood glucose statistical features, blood glucose extreme value features, and blood glucose trend features are extracted for each sleep stage state. The blood glucose statistical features include the nighttime average blood glucose value, blood glucose standard deviation, and coefficient of variation. The blood glucose extreme value features include the nighttime highest blood glucose value, lowest blood glucose value, and corresponding time points. The blood glucose trend features include the blood glucose decrease rate and the duration of hypoglycemia. The blood glucose statistical characteristics, blood glucose extreme value characteristics, and blood glucose trend characteristics of each sleep stage are fused to obtain the blood glucose fluctuation characteristics of the target object.

4. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 3, characterized in that, The step of calculating the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational age blood glucose baseline includes: Based on the standard gestational blood glucose baseline, a time window matching the nighttime monitoring period of the target object is extracted, and the baseline mean blood glucose, baseline standard deviation blood glucose, and baseline blood glucose trend slope within the time window are calculated to constitute the standard fluctuation characteristics. The blood glucose statistical features, blood glucose trend features, and baseline blood glucose mean, baseline blood glucose standard deviation, and baseline blood glucose trend slope in the blood glucose fluctuation features of the target object are calculated to obtain blood glucose mean deviation, blood glucose fluctuation amplitude deviation, and blood glucose change trend deviation. Based on the blood glucose statistical characteristics, blood glucose extreme value characteristics, and blood glucose trend characteristics of the target object under different sleep stages at night, the weight coefficients corresponding to each characteristic are determined respectively, wherein the weight coefficients are positively correlated with the depth of sleep stage. Based on the weighting coefficients, the deviation of the mean blood glucose value, the deviation of the blood glucose fluctuation amplitude, and the deviation of the blood glucose change trend are weighted and summed to obtain the deviation between the blood glucose fluctuation characteristics and the standard fluctuation characteristics corresponding to the standard gestational week blood glucose baseline.

5. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The step of determining the initial hypoglycemia threshold for each gestational week includes: Glucose time-series data of a reference population within a preset gestational age range are extracted from the historical database, and the glucose time-series data are aligned according to the sampling time point. For each sampling time point, extract a set of multiple glucose concentration values ​​corresponding to each sampling time point for the reference population; The preset percentile values ​​of each glucose concentration value set are calculated one by one, and the calculated preset percentile values ​​are fitted to generate an initial hypoglycemia warning threshold curve that changes over time. The values ​​corresponding to each sampling time point on the initial hypoglycemia warning threshold curve are determined as the initial hypoglycemia threshold.

6. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The step of correcting the initial hypoglycemia threshold based on the deviation and determining the warning threshold for the target object includes: The deviation is fused with a preset correction coefficient to obtain the threshold adjustment amount; The initial hypoglycemia threshold is corrected based on the threshold adjustment amount to obtain the warning threshold for the target object.

7. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The step of issuing a nighttime hypoglycemia warning to the target object based on the warning threshold includes: Real-time acquisition of the glucose concentration value of the target object; When the glucose concentration value is detected to be less than the warning threshold and the duration reaches a preset duration threshold, a hypoglycemia warning signal is generated. The preset output device is controlled to perform the warning operation corresponding to the low blood sugar warning signal, wherein the warning operation includes at least one of the following: audible and visual alarm, vibration reminder, and sending a warning message to the associated terminal.

8. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 3, characterized in that, The step of fusing the statistical features of blood glucose, the extreme value features of blood glucose, and the trend features of blood glucose in each sleep stage to obtain the blood glucose fluctuation features of the target object includes: The blood glucose statistical features, the blood glucose extreme value features, and the blood glucose trend features are normalized respectively to obtain the blood glucose statistical normalized value, the blood glucose extreme value normalized value, and the blood glucose trend normalized value. Obtain a feature weight matrix, wherein the feature weight matrix includes the weight coefficients corresponding to the blood glucose statistical features, the weight coefficients corresponding to the blood glucose extreme value features, and the weight coefficients corresponding to the blood glucose trend features; Based on the feature weight matrix, the blood glucose statistical normalized value, the blood glucose extreme value normalized value, and the blood glucose trend normalized value are weighted and fused to obtain the blood glucose fluctuation characteristics of the target object.

9. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 8, characterized in that, The steps for constructing the feature weight matrix include: Construct a basic weight mapping table, wherein the basic weight mapping table includes the benchmark weight coefficients corresponding to each feature dimension, and the feature dimension includes at least the blood glucose statistical feature, the blood glucose extreme value feature, and the blood glucose trend feature; The sleep stage status corresponding to each sampling point is obtained, and the baseline weight coefficients in the basic weight mapping table are corrected according to the preset physiological rhythm adjustment rules to obtain the feature weight matrix. The preset physiological rhythm adjustment rules include: when the sleep stage status indicates deep sleep, the weight coefficients associated with the extreme blood glucose feature and the blood glucose trend feature are adjusted to be greater than their corresponding baseline weight coefficients; when the sleep stage status indicates REM sleep, the weight coefficients associated with the blood glucose statistical feature are adjusted to be greater than their corresponding baseline weight coefficients.

10. The method for early warning of nocturnal hypoglycemia during pregnancy based on continuous blood glucose monitoring according to claim 1, characterized in that, The method further includes: Obtain the gestational cycle parameters of each candidate individual in the historical sample set, and divide the continuous gestational cycles into multiple discrete time windows according to the preset time span rules. The multimodal monitoring data of each candidate individual in the historical sample set are mapped to a target time window that matches their pregnancy cycle parameters, and the multimodal monitoring data of all candidate individuals within the same target time window are aggregated to generate a reference population corresponding to each pregnancy cycle.

Images