A method for risk assessment and management of multiple gestation

Abstract

The application discloses a risk assessment and management method for multiple pregnancy, and relates to the technical field of multiple pregnancy risk assessment, and comprises the following steps: constructing a time sequence behavior representation tensor of each fetus; based on the time sequence behavior representation tensor of each fetus, extracting a time-dependent growth evolution curve reflecting individual development trend, so as to construct a fetal inter-coupling coefficient matrix representing the interaction relationship of multiple fetus development; taking the growth evolution curve of each fetus and the extracted fetal inter-coupling coefficient matrix as model input, for quantifying the development cooperation degree and potential pathological interference risk between different fetuses; adjusting the dosing parameters of the individualized drug regulation module through a preset intervention strategy; generating a prediction result describing the risk grade of individual fetus; outputting a comprehensive health score and a corresponding intervention strategy suggestion; the application can provide individualized intervention strategies and treatment plans for each fetus.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multiple pregnancy risk assessment, in particular to a risk assessment and management method for multiple pregnancy. BACKGROUND

[0002] Multiple pregnancy, also known as multiple birth, refers to the presence of more than one fetus in the uterus during a single pregnancy. Pregnant women with multiple pregnancies have a higher risk of complications during pregnancy, such as gestational diabetes, hypertension, and other conditions such as premature rupture of membranes and premature birth. In severe cases, it can even endanger the life of the fetus. Therefore, the risk of multiple pregnancy is generally high. If a pregnant woman has multiple pregnancies, she needs to regularly visit the hospital for prenatal care and monitor the growth and development of the fetus.

[0003] The existing multiple pregnancy risk assessment system mainly relies on overall pregnancy risk assessment, while ignoring the development imbalance between fetuses (such as fetal growth restriction) and the mutual influence between fetuses (such as the coupling effect between fetuses). This leads to the inability to effectively identify the small differences and potential mutual influence between fetuses (such as a fetus with growth restriction may affect another fetus through placental blood flow). This lack of recognition of the coupling-feedback mechanism between multiple fetuses not only affects the accurate assessment of the growth status of each fetus, but also limits the implementation of early warning and personalized intervention strategies. SUMMARY

[0004] To solve the defects in the prior art, the present application provides a risk assessment and management method for multiple pregnancy.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] The present application provides a risk assessment and management method for multiple pregnancy, comprising the following steps:

[0007] Collecting the multi-modal physiological behavior time series data of each fetus in the pregnant woman in a multiple pregnancy state, and standardizing the multi-modal physiological behavior time series data to construct a time series behavior representation tensor for each fetus;

[0008] Based on the time series behavior representation tensor of each fetus, extracting a time-dependent growth evolution curve reflecting the individual development trend, analyzing the physiological interaction index between multiple fetuses in combination with the coupling relationship between fetal blood flow parameters, and constructing a fetal inter-coupling coefficient matrix representing the development interaction relationship between multiple fetuses;

[0009] The growth evolution curve of each fetus and the extracted inter-fetal coupling coefficient matrix are taken as model inputs, a multi-objective joint modeling structure based on time series modeling is embedded, and the development coordination degree and potential pathological interference risk between different fetuses are quantified;

[0010] According to the interaction prediction index output by the multi-objective joint modeling structure, abnormal trends in multiple pregnancy are identified, and based on the current time series behavior representation tensor state and its development trend of each fetus, the dosing parameters of the individualized drug regulation module are adjusted through a preset intervention strategy;

[0011] The time series behavior representation tensor of the fetus during the entire pregnancy period is transmitted to the cloud monitoring platform in real time through an edge collection device with data uploading function, and the data is modeled and processed by an artificial intelligence analysis model deployed in the cloud monitoring platform to generate a prediction result describing the risk level of the individual fetus;

[0012] The inter-fetal coupling coefficient matrix, the dosing parameters of the drug regulation module, and the prediction result of the individual fetus risk level are taken as joint inputs, introduced into a multi-parameter fusion analysis system, and a feature fusion and state mapping method is used to output a comprehensive health score and a corresponding intervention strategy suggestion.

[0013] As a preferred technical solution of the present application, the multi-modal physiological behavior time series data includes fetal structure development data obtained through an ultrasonic imaging system, umbilical artery and vein blood flow rate data obtained through a placental Doppler blood flow monitoring system, and fetal electrocardiogram or electroencephalogram signal data obtained through a fetal bioelectricity monitoring device.

[0014] As a preferred technical solution of the present application, the time-dependent growth evolution curve is defined as a continuous time function with respect to time and a fetal physiological behavior state vector, and a personalized growth evolution function is constructed based on fetal structure development data, umbilical artery and vein blood flow rate data, fetal electrocardiogram or electroencephalogram signal data, and time factors.

[0015] As a preferred technical solution of the present application, the coupling effect between multiple fetuses is modeled through a functional relationship between the individual growth evolution function of each fetus and the inter-fetal coupling coefficient matrix, and the specific steps include:

[0016] Interaction coupling effect modeling, for each pair of fetuses, the interaction coupling effect between the fetuses is calculated using their respective growth evolution functions and inter-fetal coupling coefficient matrices;

[0017] Responsiveness measurement, the inter-fetal interaction coupling responsiveness is calculated through the model to measure the transmission strength and mutual influence of physiological resources between fetuses.

[0018] As a preferred technical solution of the present application, the drug administration parameter of the drug regulation module is dynamically adjusted according to the development state of the fetus and real-time feedback, and the adjustment rule of the drug administration parameter is controlled by a feedback adjustment function, and the specific steps include:

[0019] The dynamic regulation mechanism sets a feedback mechanism for the release dose of the drug, and adjusts the dose of the drug in real time according to the physiological state of the fetus;

[0020] The intervention trigger time and the feedback rate are related to the release of the drug dose, and the drug administration scheme is adjusted according to the model output.

[0021] As a preferred technical solution of the present application, the prediction result of the fetal risk level adopts a weighted summation model, and the specific steps include:

[0022] Physiological behavior index integration, combined with the physiological behavior time series data of each fetus;

[0023] Weighted summation calculation, according to the weight obtained by training in the historical data, the physiological behavior time series data of each fetus is weighted and summed to obtain a comprehensive risk assessment value.

[0024] As a preferred technical solution of the present application, the generation of the comprehensive health score is realized by the following steps:

[0025] The time-dependent growth evolution curve of each fetus, the inter-fetal coupling coefficient matrix, the drug administration parameter and the prediction result of the fetal risk level are fused;

[0026] Feature normalization or tensor fusion method is used for integration, and finally the comprehensive health score of each fetus is generated.

[0027] As a preferred technical solution of the present application, the intervention strategy suggestion includes an independent drug administration path for a single fetus, risk prompt information provided to medical personnel, and dynamic sorting suggestion of fetal monitoring priority.

[0028] The present application has the following advantages:

[0029] In the present application, the physiological behavior time series data of the fetus is comprehensively analyzed, the risks in multiple pregnancy and the mutual influence between fetuses are accurately identified, and individualized intervention strategies and treatment schemes can be provided for each fetus. Real-time prediction and multi-parameter fusion analysis of artificial intelligence not only improve the accuracy of risk assessment, but also realize dynamic monitoring and precise intervention of the development state of the fetus, and bring an innovative solution to the management of multiple pregnancy. BRIEF DESCRIPTION OF DRAWINGS

[0030] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of the specification, illustrate embodiments of the application, and are used to explain the present application, but are not intended to limit the present application. In the drawings:

[0031] Figure 1 A flowchart of a method for risk assessment and management of multiple pregnancy according to the present application. DETAILED DESCRIPTION

[0032] The preferred embodiments of the present application will be described below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to explain and illustrate the present application, and do not limit the present application.

[0033] Embodiment one

[0034] As shown in the figure, a method for risk assessment and management of multiple pregnancy includes the following steps: Figure 1

[0035] Collecting multi-modal physiological behavior time series data of each fetus in a pregnant woman in a multiple pregnancy state, which covers information in multiple dimensions such as fetal weight estimation, heart rate variability, blood flow velocity, and standardizing the multi-modal physiological behavior time series data to construct a time series behavior representation tensor for each fetus. These time series data have important physiological significance and reflect the growth and development trend of the fetus at different stages of pregnancy.

[0036] By continuously monitoring the physiological data of the fetus, the doctor can understand the development status of the fetus and its physiological interaction with other fetuses in real time. For example, if a fetus has intrauterine growth restriction (IUGR), it may affect the growth and development of other fetuses through changes in placental blood flow.

[0037] Based on the time series behavior representation tensor of each fetus, a time-dependent growth evolution curve reflecting the individual development trend is extracted, and the physiological interaction index between multiple fetuses is analyzed based on the coupling relationship between fetal blood flow parameters, further mining the time sequence correlation that may exist between fetuses at the development stage, and constructing a fetal inter-coupling coefficient matrix representing the development interaction relationship between multiple fetuses.

[0038] For fetuses with abnormal development risk, the time-dependent growth evolution curve can help doctors predict the future growth trend of the fetus and identify potential developmental delay or abnormalities in a timely manner. For fetal development abnormalities, doctors can develop appropriate treatment plans based on the prediction results.

[0039] The growth evolution curve of each fetus and the extracted inter-coupling coefficient matrix are used as model inputs, and a multi-objective joint modeling structure based on time series modeling is embedded to quantify the development coordination degree and potential pathological interference risk between different fetuses. ​

[0040] In multiple pregnancies, not only are there individual developmental differences between fetuses, but if one fetus has abnormal blood flow or nutrient restriction, it can affect other fetuses through the physiological mechanisms of the placenta. Through the coupling coefficient matrix, this interaction can be quantified, providing a scientific basis for early intervention.

[0041] According to the interaction prediction indicators output by the multi-objective joint modeling structure, abnormal trends in multiple pregnancies are identified, such as early warning signals of developmental retardation or abnormal blood flow in a fetus, and based on the current temporal behavior representation tensor state and its development trend of each fetus, the dosing parameters of the individualized drug regulation module are adjusted through the preset intervention strategy, achieving individualized precision treatment in multiple pregnancies.

[0042] Through this model, abnormal trends in fetal development, such as developmental retardation and abnormal placental blood flow, can be effectively identified and early warning can be performed to help doctors quickly determine which fetuses need priority intervention and which fetuses are in a healthy development stage.

[0043] The drug regulation module can dynamically adjust the dosage and timing of drug administration based on the physiological state and prediction results of the fetus, and through a drug release function based on feedback control, the drug release dose is dynamically adjusted to ensure the optimal development state of the fetus.

[0044] The temporal behavior representation tensor of the fetus during the entire pregnancy period is transmitted in real time to the cloud monitoring platform through edge collection devices with data uploading function, and the data is modeled and processed by artificial intelligence analysis models deployed on the cloud monitoring platform to generate prediction results describing the risk level of individual fetuses.

[0045] The cloud monitoring platform can not only process the physiological data of the fetus in real time, but also provide dynamic health risk assessment and prediction based on artificial intelligence analysis results to help doctors make more accurate treatment decisions.

[0046] The coupling coefficient matrix between fetuses, the dosing parameters of the drug regulation module, and the prediction results of the risk level of individual fetuses are imported into a multi-parameter fusion analysis system as joint inputs, and a feature fusion and state mapping method is used to output a comprehensive health score and corresponding intervention strategy suggestions, including drug dosage adjustment, diagnosis and treatment process prompts, and follow-up plan suggestions.

[0047] Further, the multi-modal physiological behavior temporal data includes fetal structure development data obtained through an ultrasonic imaging system, umbilical artery and vein blood flow rate data obtained through a placental Doppler blood flow monitoring system, fetal electrocardio or electroencephalogram signal data obtained through a fetal bioelectricity monitoring device, and can also include genomic data obtained through prenatal genetic screening means.

[0048] Specifically, the step of standardizing the time series data is as follows:

[0049] Data collection: First, collect the multi-modal physiological behavior time series data of the fetuses in the multiple pregnancy pregnant woman, including the estimated fetal weight, heart rate variability, blood flow velocity, etc.

[0050] Calculate the mean and standard deviation: For each time series data feature of each fetus (such as estimated fetal weight, heart rate variability, etc.), calculate the mean and standard deviation of the feature at all time points;

[0051] Standardization operation: Apply the standardization formula to each data point to convert it to zero mean and unit variance;

[0052] Time series behavior representation tensor construction: Merge the standardized time series data of each fetus to form a multi-dimensional time series behavior representation tensor, in which the standardized data of each fetus at each time point is organized into a vector, and the time series data of all fetuses can be uniformly modeled and analyzed.

[0053] Further, the current multiple pregnancy risk assessment system generally uses single-point or stage static physiological indicators for analysis, ignoring the physiological behavior change trend of the fetus in the entire pregnancy period, especially in the multiple pregnancy scene, it is difficult to dynamically capture the evolution track of each fetus in different development stages, Therefore, the existing method is difficult to accurately model and quantitatively evaluate the potential time-dependent risk, slight development deviation and coupling feedback effect between fetuses.

[0054] In view of the above problems, the present application proposes to take the "time-dependent growth evolution curve" as the core modeling element to dynamically track and model the development process of each fetus, thereby supporting high-precision modeling of the physiological state evolution path of individual fetuses, and further serving as the basis input for subsequent multi-fetal interaction analysis and risk prediction.

[0055] The time-dependent growth evolution curve is defined as a continuous time function with respect to time and fetal physiological behavior state vector, and is constructed based on fetal structure development data, umbilical artery and vein blood flow rate data, fetal electrocardiogram or electroencephalogram signal data and time factors. Personalized growth evolution function.

[0056] Specifically, the time-dependent growth evolution curve is defined as a continuous time function with respect to time and fetal physiological behavior state vector, and is implemented based on a linear model and feature weighting, and the specific expression form is:

[0057] ;

[0058] Where, x if i (x i ,t) represents the individual growth evolution function fitted by historical case data and current monitoring data, reflecting the comprehensive physiological health level, often in the form of a standardized score (such as [0, 1]) or a risk score, also represented as the physiological behavior state vector of the i-th fetus at time t;

[0059] w f is a pre-trained weight vector, used to represent the importance of each physiological indicator to the growth and development, the weight can be obtained by regression fitting of historical medical record data, for example, the growth rate of fetal biometric values (such as head circumference, abdominal circumference, femur length) measured by ultrasound can be used as the target value, and linear regression, ridge regression, etc. can be used to train;

[0060] b f is a bias term, and σ(·) is an activation function (Sigmoid function) for mapping the output value to the interval [0, 1].

[0061] Specifically, when f i (x i ,t) < 0.3, it represents a serious development restriction (emergency intervention is required), when 0.3 < f i (x i ,t) < 0.7, it represents a medium-low risk, and when f i (x i ,t) > 0.7, it represents a good development state.

[0062] Further, when the fetus is restricted in growth (IUGR) or the blood flow of a certain fetus is abnormal, this coupling mechanism may trigger a "chain reaction" and affect other fetuses, thereby inducing systemic risk. Therefore, the present application proposes a modeling scheme based on the analysis mechanism of the coupling effect between fetuses, which is used to quantify and reveal the cooperative development state and potential pathological interference channels between multiple fetuses, thereby providing a basis for more accurate risk prediction and intervention strategy formulation.

[0063] The coupling effect between multiple fetuses is modeled by the functional relationship between the individual growth evolution function of each fetus and the inter-fetal coupling coefficient matrix, and the specific steps include:

[0064] Interaction coupling effect modeling, for each pair of fetuses, the individual growth evolution function and the inter-fetal coupling coefficient matrix are used to calculate the interaction coupling effect between fetuses;

[0065] Responsiveness measurement, the interaction coupling responsiveness between fetuses is calculated by the model to measure the transmission strength and mutual influence of physiological resources between fetuses.

[0066] Specifically, the coupling effect between multiple fetuses is modeled by the functional relationship between the individual growth evolution function of each fetus and the inter-fetal coupling coefficient matrix, which is expressed as follows:

[0067] ;

[0068] where E is the total coupling response index between the fetal population, which is used to measure the strength of system interaction between multiple fetuses. In normal healthy multiple pregnancy samples, the average stable value of E is in the interval [0.5, 1.2]. If E > 1.5, it indicates that the system coupling response strength deviates significantly, suggesting the risk of development imbalance. If the growth rate of abnormal E value (dE / dt > 0.2 / week) can be used as a trigger for intervention.

[0069] f i (x i ,t) represents the growth evolution function of the i-th fetus, g j (x j ,t) represents the auxiliary characteristic evolution function of the j-th fetus, which is usually related to the metabolic state, placental blood flow, and other physiological characteristics of the fetus, and changes with time, i.e. f i (x i ,t) is used to describe the growth process of the fetus, and g j (x j ,t) describes the characteristics related to the physiological state of the fetus (such as blood flow, metabolism, etc.);

[0070] Similarly, the auxiliary characteristic evolution function g j (x j ,t) can also use a linear model with the same structure:

[0071] ;

[0072] Here, the Sigmoid function can not be used to preserve the original meaning of the physiological characteristics.

[0073] C ij represents the interaction coupling coefficient between fetus i and fetus j, reflecting the correlation strength in physiological aspects such as nutrient transmission, blood flow regulation, pressure coupling, etc. The recommended setting range of C ij is [0, 1], where C ij = 0 indicates that there is no significant interaction between fetus i and j, and C ij = 1 indicates that there is strong coupling dependence between fetus i and j (such as sharing blood supply).

[0074] The initial value of C ij can be set as follows:

[0075] Clinical pathway dependency, e.g. twin fetus in dichorionic diamniotic structure, C ij Usually less than 0.3;

[0076] Data-driven method, using time-delayed mutual information of fetal physiological states to estimate;

[0077] n represents the total number of fetuses in multiple pregnancy (usually 2-4 fetuses).

[0078] The specific workflow is as follows:

[0079] Collect the behavior state vector x i of each fetus (such as weight estimation, heart rate variability, blood flow velocity, etc.) and auxiliary features x j (such as umbilical vein, cerebral artery, placental blood oxygen, etc.);

[0080] Fit individual functions f i (x i ,t) and g j (x j ,t) based on historical cases and current data, and construct individual growth trend expression;

[0081] Based on biological mechanisms such as blood flow coupling model and tissue perfusion delay model, construct C ij , the initial value can be set from clinical priori, or can be learned from data-driven;

[0082] Substitute the above three types of quantities into the formula, and calculate E in real time;

[0083] Set the risk threshold τ, when E > τ, trigger the intervention system, prompt the medical staff to pay attention to the fact that a fetus may be affecting other fetuses;

[0084] Take E as input into the subsequent multi-objective joint model, which is used to predict systemic risk and develop individualized intervention plan.

[0085] For example, in a case of twin pregnancy, the growth rate f A (x A ,t) of fetus A is significantly lower than that of fetus B, but the fluctuation of B's blood flow parameters increases, at this time C AB = 0.42 (indicating that there is a shared area in the placental distribution of the two), when the system monitors: f A (x A ,t) is continuously decreasing, the cerebral blood flow index RI in g B (x B ,t) is rising, the overall E value rises from 1.1 to 1.6 and dE / dt increases rapidly;

[0086] It can be prompted that the development retardation of fetus A has affected B, and the drug regimen of B needs to be adjusted or the monitoring frequency needs to be increased to prevent systemic deterioration.

[0087] Further, the drug delivery parameters of the drug regulation module are dynamically adjusted by a feedback regulation function. In this way, the release strategy of the drug can be accurately adjusted according to the real-time monitoring of the fetal growth evolution, coupling effect and health status, so as to maximize the effect of individualized treatment.

[0088] The drug delivery parameters of the drug regulation module are dynamically adjusted according to the real-time feedback of the fetal development state, and the adjustment rule of the drug delivery parameters is controlled by a feedback regulation function. The specific steps include:

[0089] A dynamic regulation mechanism is provided to adjust the feedback mechanism of the drug release dose and the dose of the drug in real time according to the physiological state of the fetus;

[0090] The intervention trigger time and the feedback rate are related to the release of the drug dose, and the drug delivery scheme is adjusted according to the model output.

[0091] Specifically, the drug delivery parameters of the drug regulation module are dynamically adjusted by the following feedback regulation function:

[0092] ;

[0093] Where D(t) represents the drug release dose at time t, which is a key parameter dynamically adjusted according to the physiological changes of the fetus;

[0094] A represents the regulation peak release amount, which provides rapid intervention capability in high risk, i.e. the maximum drug dose required to be released for rapid intervention in high risk, in the case of acute high risk (such as fetal growth restriction, insufficient placental blood flow, etc.), the value of A needs to be set higher to provide sufficient drug intervention. Generally, this value can be set according to clinical experience and treatment requirements, such as 1.5-2 times the target drug dose;

[0095] k1 is the change rate coefficient of the regulation function, which determines the time response speed of the drug release, which is set according to the response time window. In general, k1 needs to be adjusted according to the change speed of the fetal health status. If the fetal health status changes rapidly (such as sudden placental dysfunction), a larger k1 needs to be set to accelerate the release of the drug, and when the fetal health status is relatively stable, the value of k1 should be smaller to avoid excessive intervention. Generally, the value of k1 can be set between 0.1 and 1;

[0096] t0 represents the time point of intervention response trigger, which is usually the time node of abnormal fetal health status, and is determined according to the clinical risk assessment system. When the physiological behavior of the fetus is monitored to be abnormal, the value will be automatically adjusted to the time point of risk occurrence;

[0097] B is the minimum basic maintenance dose, which is the minimum dose of the drug that must be maintained in any case to ensure the stability of the basic physiological function, which is generally small, set as the minimum value required to maintain the basic physiological function of the fetus, and is usually set to 10%-20% of the normal drug dose in clinical practice.

[0098] The purpose of designing this control function is to dynamically adjust the release amount of the drug according to the real-time health status of the fetus, and to provide the ability to quickly intervene at high-risk moments:

[0099] Dynamic response, the core feature of this function is to reflect the time dependence of drug release through exponential decay This design allows the release amount of the drug to be adjusted in real time according to the health status of the fetus, especially when the health status of the fetus changes suddenly, the function can quickly respond and provide a peak value of drug intervention, and then gradually stabilize to the basic maintenance dose;

[0100] Risk perception and adjustment ability, when the fetus has a risk signal (such as abnormal blood flow, fetal growth retardation, etc.), the setting of A value ensures that the drug can be quickly released in the shortest time to deal with the crisis, and by adjusting k1, the system can adjust the response speed of drug release, so as to provide a more appropriate drug intervention scheme under different health conditions;

[0101] Ensure safety, the minimum basic dose B of the drug ensures that the drug release always remains at a safe level in the absence of obvious risks, which can avoid the side effects of excessive drug use and ensure the safety of the mother and fetus.

[0102] Further, the existing multiple pregnancy risk assessment method usually ignores the individual differences of each fetus and does not consider the interaction between fetuses, which leads to the inability to accurately predict the health problems such as growth restriction and abnormal heart rate of each fetus, therefore, the present application proposes a risk prediction method based on multi-modal physiological behavior data and weighted summation model, which can comprehensively evaluate the health status of each fetus, thereby providing a basis for medical intervention.

[0103] The prediction result of the fetal risk level adopts a weighted summation model, and the specific steps include:

[0104] Physiological behavior index integration, combined with the physiological behavior time series data of each fetus;

[0105] Weighted summation calculation, according to the weight trained in the historical data, the physiological behavior time series data of each fetus is weighted and summed to obtain a comprehensive risk assessment value.

[0106] Specifically, the prediction result of the fetal risk level adopts a weighted summation model, which is expressed as follows:

[0107] ;

[0108] wherein P risk is the comprehensive risk assessment value at the current time t, since the physiological state of the fetus changes at different time points, the risk assessment value P risk is a dynamic quantity, when weighted summation is performed, the real-time monitored data is constantly updated according to the latest situation, and the finally derived risk assessment value also reflects the health status of the fetus, a i is the index importance weight obtained by training historical data, which represents the influence degree of each index on the final risk assessment, the value of a i is obtained by training historical data, which can reflect the importance of the index under a specific pregnancy state, for example, the importance of fetal movement frequency under normal pregnancy condition may be low, but when fetal growth restriction occurs, the change of fetal movement frequency may better reflect the health status of the fetus, these weight coefficients can be dynamically adjusted according to different gestational age, fetal health status, etc. in actual application.

[0109] In actual use, each index (X i (t)) is usually standardized or normalized to make all input dimensions consistent, for example: unitizing the fetal heart rate (such as converting per minute to standard value per unit time), normalizing the fetal movement frequency to make its dimension consistent with other indexes, and standardizing the fetal weight estimation value.

[0110] Further, the generation of the comprehensive health score is realized by the following steps:

[0111] fusing the time-dependent growth evolution curve of each fetus, the inter-fetal coupling coefficient matrix, the drug use parameter, and the prediction result of the fetal risk level;

[0112] integrating by using feature normalization or tensor fusion method, and finally generating the comprehensive health score of each fetus, to provide a quantifiable personalized health status reference for the clinic, and to support the intelligent generation of drug adjustment and medical decision path.

[0113] Specifically, the comprehensive health score is expressed by the following integration function:

[0114] ;

[0115] wherein S is the final health assessment score of fusing all key factors, f(·) is a function with feature normalization, tensor fusion and nonlinear mapping, supporting multi-factor integrated prediction.

[0116] Generally, the value range of the health score S will be adjusted according to different clinical applications. For the comprehensive score function of the present application, the value range of the health score S is usually between 0 and 1, which depends on the design of the model and the weight and input of different variables. For example, if the physiological state of the fetus is good and the risk is low, the score S will be close to 1, indicating a good health state. If the score is close to 0, it indicates that the fetus has a serious health risk and needs emergency intervention.

[0117] Suppose we have a deep neural network model to represent f(·), whose basic structure can be represented as:

[0118] ;

[0119] where σ is the activation function, usually Sigmoid or ReLU, to ensure that the output score is between 0 and 1, and α and β are weight coefficients to adjust the influence of drug intervention and risk score on the final result.

[0120] Further, the present application accurately depicts the individual development trend of the fetus and the dynamic interaction between the fetuses by constructing the fetal individual growth evolution function and the fetal coupling coefficient matrix in the form of time series modeling, and further generates a comprehensive health score S in real time by fusing and analyzing the physiological risk factors output by the artificial intelligence engine. Based on the health score result, an intervention strategy generation module is constructed to intelligently recommend possible intervention methods during multiple pregnancy.

[0121] The intervention strategy suggestions include an independent drug administration path for single fetus, risk prompt information provided to medical personnel, and dynamic sorting suggestions for fetal monitoring priority.

[0122] The above score S is used as an input parameter for the intervention strategy decision module to realize the following three intelligent strategy outputs:

[0123] Independent drug administration path suggestion, when the system determines that the comprehensive health score S and the individual risk assessment value P risk are in the high risk interval, the drug intervention feedback control model D(t) is enabled;

[0124] Risk prompt information (provided to medical personnel), the system determines whether to issue a risk warning according to the coupling influence coefficient C ij between fetuses and the individual risk assessment value P risk , for example, when the coupling influence coefficient C ij between two fetuses exceeds the threshold value 0.5 and the risk value P risk is greater than the threshold value, the system will automatically trigger a risk warning to prompt medical personnel to pay attention to the health of the fetus;

[0125] The fetal monitoring priority dynamic ranking suggestion ranks the P risk , relative coupling influence and relative coupling influence comprehensive calculation monitoring priority R i :

[0126] ;

[0127] Wherein, σ i is the physiological variation rate of the fetus, which describes the fluctuation of the physiological state of the fetus. For a certain physiological index x i (t) (such as fetal heart rate), it is sampled N times in the time window [t-Δt, t], and its variation rate is defined as:

[0128] ;

[0129] Wherein, x i (t k ) represents the physiological data at time point t k , represents the mean value of multiple sets of physiological data in the time window Δt;

[0130] C i is the coupling influence of the fetus, that is, the influence strength with other fetuses, which is represented as:

[0131] ;

[0132] λ1, λ2, λ3 are the corresponding weight coefficients, preferably λ1∈[0.4, 0.6], the clinical risk dominant weight; λ2∈[0.2, 0.3], the coupling influence dominant weight; λ3∈[0.1, 0.2], the behavior volatility contribution, the system dynamically adjusts the monitoring frequency suggestion of the fetus (for example: it is suggested to perform monitoring at least 3 times within 24 hours). i

[0133] Take a twin pregnant woman who receives real-time fetal monitoring by edge devices as an example:

[0134] The system detects that fetus A has an increased umbilical artery pulsatility index (PI) and reduced fetal movement;

[0135] It is found that fetus B has obvious positive coupling (C AB =0.72) with A;

[0136] The system triggers a coupling warning and suggests that the medical staff increase the monitoring frequency of B;

[0137] For A, the drug path feedback function is started, and A=5mg / h and k1=0.4 are initially set;

[0138] ​Simultaneous output of fetal priority monitoring ranking: B>A (because it is susceptible to coupled conduction).

[0139] Finally, it should be noted that the above only for the preferred embodiments of the present application, and is not intended to limit the present application, although the foregoing embodiments of the present application has been described in detail, for those skilled in the art, it still can be modified, or the equivalent replacement of part of the technical features described in the foregoing embodiments. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, shall be included within the scope of the present application.

Claims

1. A method for risk assessment and management of multiple pregnancies, characterized in that, Includes the following steps: Multimodal physiological behavior temporal data of each fetus in multiple pregnancies were collected and standardized to construct a temporal behavior representation tensor for each fetus. Based on this temporal behavior representation tensor, time-dependent growth evolution curves are extracted, and a coupling coefficient matrix reflecting multiple fetal interactions is constructed by combining the coupling relationship of physiological parameters between fetuses. The growth evolution curves and coupling coefficient matrix are input into a multi-objective joint modeling structure to quantify the developmental synergy among fetuses and the risk of pathological interference. Based on the interactive indicators output by the multi-objective joint modeling structure, abnormal pregnancy trends are identified, and dosing parameters are adjusted according to the temporal status and development trend of each fetus. The coupling effect among multiple fetuses is modeled by the functional relationship between the individual growth and evolution function of each fetus and the coupling coefficient matrix between fetuses, and its expression is as follows: ； Where E is the total coupling response index among fetal populations, used to measure the strength of system interactions among multiple pregnancies, f i (x i (t) represents the growth and evolution function of the i-th fetus, g j (x j (t) represents the auxiliary feature evolution function of the j-th fetus, C ij denoted by , where i represents the interaction coupling coefficient between fetus i and fetus j, and n represents the total number of fetuses in a multiple pregnancy; Growth evolution function f i (x i The specific form of ,t) is: ； Where, x i w represents the behavioral state vector of the i-th fetus at time t. f It is a pre-trained weight vector used to characterize the importance of each physiological indicator in contributing to growth and development, b f The bias term is σ(·), and the activation function is σ(·). Auxiliary feature evolution function g j (x j Linear models with the same structure are used for t): ； The temporal behavior representation tensor throughout the entire pregnancy cycle is uploaded to a cloud-based artificial intelligence analysis platform for processing to obtain an individual fetal risk level prediction. By integrating the coupling coefficient matrix, drug administration parameters, and risk level predictions, a comprehensive health score and intervention strategy recommendations are output.

2. The method for risk assessment and management of multiple pregnancies according to claim 1, characterized in that, The multimodal physiological behavior time-series data includes fetal structural development data acquired through an ultrasound imaging system, umbilical artery and vein blood flow velocity data acquired through a placental Doppler blood flow monitoring system, and fetal electrocardiogram or electroencephalogram signal data acquired through a fetal bioelectric monitoring device.

3. The method for risk assessment and management of multiple pregnancies according to claim 2, characterized in that, The time-dependent growth evolution curve is defined as a continuous time function of time and fetal physiological behavior state vector. A personalized growth evolution function is constructed based on fetal structural development data, umbilical artery and vein blood flow velocity data, fetal electrocardiogram or electroencephalogram signal data and time factors.

4. The method for risk assessment and management of multiple pregnancies according to claim 3, characterized in that, The coupling effect among multiple fetuses is modeled through the functional relationship between the individual growth and evolution functions of each fetus and the coupling coefficient matrix between fetuses. Specific steps include: Inter-coupling effect modeling: For each pair of fetuses, the inter-fetal interaction coupling effect is calculated using their respective growth evolution functions and the inter-fetal coupling coefficient matrix. The responsiveness measure calculates the interfetal interaction coupling responsiveness through a model, measuring the intensity of physiological resource transfer and mutual influence between fetuses.

5. The method for risk assessment and management of multiple pregnancies according to claim 4, characterized in that, The drug administration parameters of the drug regulation module are dynamically adjusted based on the fetal development status and real-time feedback. The adjustment rules for the drug administration parameters are controlled by a feedback regulation function, and the specific steps include: A dynamic regulation mechanism is used to set a feedback mechanism for drug release dosage and adjust the drug dosage in real time according to the physiological state of the fetus. The intervention trigger time and feedback rate, as well as the release of drug dosage, are related to the intervention trigger time and feedback rate. The drug administration regimen is adjusted based on the model output.

6. The method for risk assessment and management of multiple pregnancies according to claim 5, characterized in that, The fetal risk level prediction uses a weighted summation model, and the specific steps include: Physiological behavior indicators are integrated, combining the time-series data of each fetus's physiological behavior; The weighted summation calculation, based on the weights trained from historical data, sums the time-series data of each fetus's physiological behavior to obtain a comprehensive risk assessment value.

7. The method for risk assessment and management of multiple pregnancies according to claim 6, characterized in that, The comprehensive health score is generated through the following steps: The time-dependent growth evolution curves of each fetus, the coupling coefficient matrix between fetuses, medication parameters, and the prediction results of fetal risk levels are fused together. By integrating features using feature normalization or tensor fusion methods, a comprehensive health score for each fetus is ultimately generated.

8. A method for risk assessment and management of multiple pregnancies according to claim 7, characterized in that, The intervention strategy recommendations include a separate dosing pathway for singleton pregnancies, risk warning information provided to healthcare professionals, and recommendations for dynamically prioritizing fetal monitoring.

Images