Prenatal monitoring risk assessment system and method

Abstract

The present invention discloses a prenatal monitoring risk assessment system and method, which relate to the technical field of medical health monitoring, including collecting physiological data of mother and fetus in real time through wearable devices, fetal heart rate monitors and ultrasonic devices; preprocessing the collected physiological data; using a bidirectional long short-term memory network Bi-LSTM to build a mother-fetus interaction model, inputting the preprocessed physiological data into the mother-fetus interaction model, capturing the dynamic association between the physiological data of the mother and the fetus, and outputting a prenatal risk score; at the same time, by inputting historical data into the mother-fetus interaction model, using an attention mechanism to dynamically adjust parameter weights, and optimizing the mother-fetus interaction model; dividing the risk into three risk levels of low, medium and high according to the prenatal risk score; and giving specific intervention suggestions according to the risk level. The present invention captures the complex dynamic interaction relationship between mother and fetus through data preprocessing and deep modeling, and improves the accuracy of risk assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of medical health monitoring, and in particular to a prenatal monitoring risk assessment system and method. Background Art

[0002] With the development of medical technology and information technology, prenatal monitoring has become one of the important means to ensure the health of pregnant women and fetuses. Traditional prenatal monitoring relies on regular hospital examinations and manual evaluations. However, due to the long monitoring intervals and insufficient data timeliness, sudden risks are not discovered in time. In recent years, the rapid development of wearable devices and medical sensor technology has made it possible to collect maternal and fetal physiological data in real time, such as heart rate, blood pressure and fetal heart rate. However, how to efficiently integrate multidimensional data and dynamically monitor the health status of mothers and fetuses remains a technical difficulty. At the same time, artificial intelligence technology has gradually been applied to the medical field, especially neural networks have shown great potential in mining complex data relationships and predicting disease risks, but existing technologies have not fully considered the dynamic interaction between mothers and fetuses and the deep modeling of multidimensional data;

[0003] Although the existing prenatal monitoring technology has made certain progress, it still has shortcomings. Traditional risk assessment methods are mostly based on a single or a small number of indicators, which make it difficult to capture the complex physiological relationship between mother and fetus, resulting in incomplete and inaccurate assessment results. In addition, most systems are based on shallow analysis and cannot perform deep modeling of time series data, making it difficult to reflect dynamic characteristics. At the same time, problems such as inconsistent sampling frequencies of different devices and a large number of abnormal data also limit the actual utilization value of the data. The existing system does not make sufficient use of historical data and lacks a dynamic optimization mechanism, making it difficult to achieve personalized risk prediction and intervention recommendations. Our invention solves the shortcomings of existing technologies in dynamic interactive modeling and risk assessment. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a prenatal monitoring risk assessment method to solve the problem that traditional risk assessment methods are mostly based on a single or a small number of indicators, which are difficult to capture the complex physiological relationship between mother and fetus, resulting in incomplete and inaccurate assessment results.

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

[0007] In a first aspect, the present invention provides a method for prenatal monitoring risk assessment, comprising:

[0008] Through wearable devices, fetal heart rate monitors and ultrasound equipment, physiological data of mothers and fetuses are collected in real time;

[0009] Preprocess the collected physiological data;

[0010] A bidirectional long short-term memory network Bi-LSTM was used to build a maternal-fetal interaction model. The preprocessed physiological data were input into the maternal-fetal interaction model to capture the dynamic relationship between maternal and fetal physiological data and output the prenatal risk score.

[0011] At the same time, the maternal-fetal interaction model is optimized by inputting historical data into the maternal-fetal interaction model and using the attention mechanism to dynamically adjust the parameter weights;

[0012] The risk is divided into three risk levels: low, medium, and high according to the prenatal risk score;

[0013] Based on the risk level, specific intervention recommendations are given.

[0014] As a preferred solution of the prenatal monitoring risk assessment method of the present invention, wherein: the physiological data of the mother and the fetus are collected in real time through a wearable device, a fetal heart monitor and an ultrasonic device, and the specific steps are:

[0015] Use FDA-approved smart bracelets to collect core maternal physiological data;

[0016] Use a Doppler-based fetal heart rate monitor to collect fetal heart rate and fetal movement frequency, and use ultrasound equipment to regularly collect fetal amniotic fluid index and fetal position change data;

[0017] Data from all devices is transmitted to the central server in real time via the 5G network;

[0018] The data is stored in the efficient distributed database MongoDB and timestamped;

[0019] Integrate the collected data to generate a time series data matrix, the expression is:

[0020] ;

[0021] in, Indicates at time The time series data matrix of time, Indicates the number of acquisition devices. Indicates the start time of collection. Indicates the end time of collection. The index variable representing the acquisition device. Indicates The weight corresponding to each device, Indicates Devices at time Physiological data collected at all times , Indicates Devices at time The physiological data collected, Indicates A normalization function, Represents a small time increment in an integration operation.

[0022] As a preferred solution of the prenatal monitoring risk assessment method of the present invention, the preprocessing of the collected physiological data is specifically as follows:

[0023] Set abnormal thresholds for the collected physiological data Perform outlier detection and elimination, the expression is:

[0024] ;

[0025] ;

[0026] in, Indicates Devices at time The abnormal flag, the value of 1 indicates abnormality, and the value of 0 indicates normality. Represents the dynamic proportional coefficient for outlier determination, Indicates the first Devices at time The physiological data collected, Indicates The average value of the data collected by each device, Indicates The standard deviation of the data collected by each device;

[0027] According to the cleaning Devices at time Physiological data collected Update time series data matrix .

[0028] As a preferred solution of the prenatal monitoring risk assessment method of the present invention, the bidirectional long short-term memory network Bi-LSTM is used to construct the maternal-fetal interaction model, specifically:

[0029] We selected the bidirectional long short-term memory network Bi-LSTM as the basis of the maternal-fetal interaction model;

[0030] The time series data matrix is a two-dimensional matrix, representing the time The data feature set has the dimension ,in, Indicates the number of acquisition devices. Indicates the number of features;

[0031] The time series data matrix Convert to time series data tensor , It is a three-dimensional data structure, which means arrive Multidimensional maternal, fetal, and environmental data collected over the entire time range, with dimensions of ,in, Represents the length of the time series, and the expression is: ,in, Indicates belonging to, Represents the field of real numbers.

[0032] As a preferred embodiment of the prenatal monitoring risk assessment method of the present invention, the pre-processed physiological data is input into the maternal-fetal interaction model to capture the dynamic relationship between the maternal and fetal physiological data and output the prenatal risk score. The specific steps are as follows:

[0033] Convert the time series data tensor The input maternal-fetal interaction model is used to extract features from the three-dimensional data to obtain the hidden state, which is expressed as:

[0034] ;

[0035] in, Represents the maternal-fetal interaction model at time The hidden state of is a bidirectional long short-term memory network, Represents a time series tensor at time slices, and are the hidden state and memory state of the previous time step respectively;

[0036] The dynamic relationship between the mother and the fetus is captured according to the hidden state, expressed as:

[0037] ;

[0038] in, and The index variable representing the feature, Indicates the dynamic incidence matrix and The dynamic correlation between the features, represents the length of the time series, and Respectively expressed in time Time and The hidden state of a feature, Indicates at time Time and The interaction eigenvalues ​​between the features are is the activation function Sigmoid, represents the hyperbolic tangent function, Indicates based on and The exponential decay function of , and is the weight parameter;

[0039] The attention mechanism is introduced to dynamically adjust the weight of the association matrix. The attention optimization formula is:

[0040] ;

[0041] ;

[0042] in, Indicates at time The attention weight vector, is the activation function Softmax function, represents the weight matrix of the attention mechanism, represents the bias term of the attention mechanism, Indicates at time The dynamic incidence matrix of Indicates at time The optimized dynamic correlation matrix;

[0043] Based on the optimized dynamic association matrix, the prenatal risk score is calculated. The risk score calculation formula is:

[0044] ;

[0045] in, represents the prenatal risk score, is the activation function Sigmoid, and Represents the weights and biases for risk score calculation.

[0046] As a preferred embodiment of the prenatal monitoring risk assessment method of the present invention, the risk is divided into three risk levels: low, medium and high according to the prenatal risk score, specifically:

[0047] Collect historical physiological data and input them into the risk score calculation formula to obtain the historical prenatal risk score ;

[0048] ;

[0049] in, , and Represent the output values ​​of low risk, medium risk and high risk levels respectively, represents the mean of the historical prenatal risk scores, represents the standard deviation of historical prenatal risk scores, and is the dynamically adjusted proportional coefficient, represents the smoothing parameter, represents the exponential function, represents the proportionality coefficient;

[0050] When calculated When , it indicates low risk;

[0051] When calculated , indicating medium risk;

[0052] When calculated , indicating high risk.

[0053] As a preferred embodiment of the prenatal monitoring risk assessment method of the present invention, the specific intervention suggestions are given according to the risk level, specifically:

[0054] When the risk level is low, basic health advice is provided to maintain a balanced diet and exercise moderately;

[0055] When the risk level is medium, monitor the patient based on the doctor's advice and consider mild intervention measures, such as adjusting lifestyle or supplementing nutrition.

[0056] When the risk level is high, seek medical attention immediately, conduct a comprehensive examination based on the doctor's advice, and initiate medical intervention.

[0057] In a second aspect, the present invention provides a prenatal monitoring risk assessment system, including a collection module, a preprocessing module, a maternal-fetal interaction module, a risk assessment module, and an intervention suggestion module;

[0058] The acquisition module is responsible for collecting physiological and environmental data of the mother, fetus and environment in real time;

[0059] The preprocessing module is responsible for cleaning, standardizing and interpolating the collected raw data to generate a high-quality time series data matrix;

[0060] The maternal-fetal interaction module is responsible for constructing a maternal-fetal interaction model using a bidirectional long short-term memory network Bi-LSTM, capturing the dynamic relationship between the mother and the fetus, optimizing model parameters, and calculating the prenatal risk score;

[0061] The risk assessment module divides the risk level according to the prenatal risk score and dynamically adjusts the threshold based on historical data;

[0062] The intervention suggestion module provides personalized intervention suggestions based on the risk assessment results.

[0063] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the prenatal monitoring risk assessment method as described in the first aspect of the present invention is implemented.

[0064] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the prenatal monitoring risk assessment method as described in the first aspect of the present invention is implemented.

[0065] The beneficial effects of the present invention are as follows: the present invention constructs a comprehensive and efficient prenatal monitoring risk assessment method by combining multi-source data collection, preprocessing, maternal-fetal dynamic modeling, risk level classification and personalized intervention recommendations. By real-time collection and integration of maternal, fetal and environmental data, the problem of single data and poor timeliness in traditional prenatal monitoring is solved; through data preprocessing and deep modeling, the complex dynamic interaction relationship between mother and fetus is captured, and the accuracy of risk assessment is improved; through risk level classification and personalized intervention recommendations, scientific and dynamic health management is achieved. On the whole, the invention has significant technical advantages in improving the accuracy and guidance of prenatal risk monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0067] Figure 1 This is a flow chart of the antenatal monitoring risk assessment method in Example 1.

[0068] Figure 2 This is an interval diagram for the risk level classification in Example 1. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0070] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0071] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0072] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for prenatal monitoring risk assessment, comprising the following steps:

[0073] S1: Real-time collection of maternal and fetal physiological data through wearable devices, fetal heart rate monitors and ultrasound equipment;

[0074] Use an FDA-certified smart bracelet to collect the mother's core physiological data, including heart rate, blood pressure, blood sugar, blood oxygen saturation, and body temperature;

[0075] Use a Doppler-based fetal heart rate monitor to collect fetal heart rate and fetal movement frequency, and use ultrasound equipment to regularly collect fetal amniotic fluid index and fetal position change data;

[0076] Air quality monitors are installed to collect air pollution index and noise level around pregnant women to assess the potential impact of the environment on maternal and fetal health;

[0077] Data from all devices is transmitted to the central server in real time via the 5G network;

[0078] The data is stored in the highly efficient distributed database MongoDB and timestamped to ensure time series consistency during subsequent processing;

[0079] All of the above equipment are medical grade, with high precision and reliability, ensuring data quality. They collect maternal, fetal and environmental data at the same time, and can provide comprehensive physiological information. The data is transmitted in real time through a high-speed network, supporting dynamic monitoring and model optimization. The data storage adopts a distributed architecture, which is convenient for adding more monitoring indicators in the future.

[0080] Integrate the collected data to generate a time series data matrix, the expression is:

[0081] ;

[0082] in, Indicates at time The time series data matrix of time, Indicates the number of acquisition devices. Indicates the start time of collection. Indicates the end time of collection. The index variable representing the acquisition device. Indicates The weight corresponding to each device, Indicates Devices at time Physiological data collected at all times , Indicates Devices at time The physiological data collected, Indicates A normalization function, Represents a small time increment in an integration operation;

[0083] Through The integral definition ensures the continuity and consistency of data at different time points, providing a basis for subsequent input models (such as the Bi-LSTM model in the subsequent step S3). The time series characteristics are the basis for capturing changes in the physiological states of the mother and fetus;

[0084] No. Normalization function , the expression is:

[0085] ;

[0086] in, Indicates The historical average value of data collected by each device, Indicates The standard deviation of the data collected by each device.

[0087] S2: preprocessing the collected physiological data;

[0088] Set abnormal thresholds for the collected physiological data Perform outlier detection and elimination, the expression is:

[0089] ;

[0090] ;

[0091] in, Indicates Devices at time The abnormal flag, the value of 1 indicates abnormality, and the value of 0 indicates normality. Represents the dynamic proportional coefficient for outlier determination, Indicates the first Devices at time The physiological data collected, Indicates The average value of the data collected by each device, Indicates The standard deviation of the data collected by each device;

[0092] According to the cleaning Devices at time Physiological data collected Update time series data matrix ;

[0093] Compared with the prior art, the present invention proposes an outlier detection method based on proportional coefficient by combining the dynamic statistical characteristics of historical average and standard deviation, which can more flexibly adapt to the needs of abnormal data removal in different devices and scenarios. Especially in the processing scenario of multi-source heterogeneous data, the traditional fixed threshold method is prone to misjudgment, while the dynamic adjustment mechanism of the present invention significantly improves the accuracy and adaptability of anomaly detection;

[0094] This step solves the common outlier problem in multi-device data collection (such as equipment error, environmental interference, etc.), and avoids the impact of abnormal data on subsequent analysis. By eliminating outliers, the accuracy and reliability of the data are guaranteed, providing a clean data foundation for the generation of time series data matrix;

[0095] Since the sampling frequencies of maternal and fetal data are different (for example, the fetal heart rate is sampled once per second, while the blood pressure is sampled once every 10 minutes), a method based on spline interpolation is used to interpolate and synchronize the data.

[0096] S3: A bidirectional long short-term memory network Bi-LSTM is used to build a maternal-fetal interaction model, and the preprocessed physiological data is input into the maternal-fetal interaction model to capture the dynamic association between maternal and fetal physiological data and output the prenatal risk score;

[0097] At the same time, the maternal-fetal interaction model is optimized by inputting historical data into the maternal-fetal interaction model and using the attention mechanism to dynamically adjust the parameter weights;

[0098] We selected the bidirectional long short-term memory network Bi-LSTM as the basis of the maternal-fetal interaction model;

[0099] Time series data matrix is a two-dimensional matrix, representing the time The data feature set has the dimension ,in, Indicates the number of acquisition devices. Indicates the number of features;

[0100] The time series data matrix Convert to time series data tensor , Representation is a three-dimensional data structure that represents the entire time range (from arrive ) collected multi-dimensional maternal, fetal and environmental data, with dimensions of ,in, Represents the length of the time series, and the expression is: ,in, Indicates belonging to, represents the field of real numbers;

[0101] Convert the time series data tensor The input maternal-fetal interaction model is used to extract features from the three-dimensional data to obtain the hidden state, which is expressed as:

[0102] ;

[0103] in, Represents the maternal-fetal interaction model at time The hidden state of is a bidirectional long short-term memory network, Represents a time series tensor at time slices, and are the hidden state and memory state of the previous time step respectively;

[0104] The memory state of the previous time step It can be directly obtained through the existing LSTM model framework;

[0105] The dynamic relationship between the mother and the fetus is captured according to the hidden state, expressed as:

[0106] ;

[0107] in, and The index variable representing the feature, Indicates the dynamic incidence matrix and The dynamic correlation between the features, represents the length of the time series, and Respectively expressed in time Time and The hidden state of a feature, Indicates at time Time and The interaction eigenvalues ​​between the features are is the activation function Sigmoid, represents the hyperbolic tangent function, Indicates based on and The exponential decay function of , and is the weight parameter, For adjustment The influence of items on dynamic correlation, For adjustment The impact of items on dynamic correlation, For adjustment The impact of nonlinear changes in items on dynamic correlation;

[0108] The attention mechanism is introduced to dynamically adjust the weight of the association matrix. The attention optimization formula is:

[0109] ;

[0110] ;

[0111] in, Indicates at time The attention weight vector, is the activation function Softmax function, which is used to normalize the vector into a probability distribution. Represents the weight matrix of the attention mechanism, which is used to transform the hidden state Mapped to the attention space, represents the bias term of the attention mechanism, Indicates at time The dynamic incidence matrix of Indicates at time The optimized dynamic correlation matrix;

[0112] Based on the optimized dynamic association matrix, the prenatal risk score is calculated. The risk score calculation formula is:

[0113] ;

[0114] in, represents the prenatal risk score, ranging from [0,1], is the activation function Sigmoid, and Represents the weights and biases for risk score calculation.

[0115] S4: The risk is divided into three risk levels: low, medium and high according to the prenatal risk score;

[0116] Collect historical physiological data and input them into the risk score calculation formula to obtain the historical prenatal risk score ;

[0117] ;

[0118] in, , and Represent the output values ​​of low risk, medium risk and high risk levels respectively, represents the mean of the historical prenatal risk scores, represents the standard deviation of historical prenatal risk scores, and is the dynamically adjusted proportional coefficient, represents the smoothing parameter, which is used to control the nonlinearity of risk classification. represents the exponential function, Represents the proportionality coefficient, which is used to adjust the slope of the exponential function, thereby affecting the threshold position of the risk level division;

[0119] When calculated When , it indicates low risk;

[0120] When calculated When the risk is medium, the patient should be monitored in combination with the doctor's advice, and mild intervention measures such as adjusting lifestyle or supplementing nutrition should be considered;

[0121] When calculated , indicating high risk.

[0122] S5: Give specific intervention recommendations based on risk level;

[0123] When the risk level is low, basic health advice is provided to maintain a balanced diet and exercise moderately;

[0124] When the risk level is medium, monitor the patient based on the doctor's advice and consider mild intervention measures, such as adjusting lifestyle or supplementing nutrition.

[0125] When the risk level is high, seek medical attention immediately, conduct a comprehensive examination based on the doctor’s advice, and initiate medical intervention;

[0126] Different from the traditional "one-size-fits-all" monitoring method, this invention directly links low-risk status with basic health management through risk level classification, reducing unnecessary use of medical resources and improving the autonomy of pregnant women in health management;

[0127] This step solves the problem of over-intervention in existing prenatal health management. In many cases, low-risk pregnant women are also included in the scope of strict medical monitoring, resulting in a waste of medical resources and increased psychological pressure on pregnant women. The present invention optimizes the health management process by matching basic health recommendations with low-risk status.

[0128] The hierarchical intervention mechanism of the present invention has remarkable flexibility and accuracy, which not only optimizes the use of medical resources, but also reduces the psychological and life burden of pregnant women, and maximizes the protection of maternal and fetal health and safety.

[0129] This embodiment also provides a prenatal monitoring risk assessment system, including: a collection module, a preprocessing module, a maternal-fetal interaction module, a risk assessment module and an intervention suggestion module.

[0130] The acquisition module is responsible for collecting physiological and environmental data of the mother, fetus and environment in real time;

[0131] The preprocessing module is responsible for cleaning, standardizing, and interpolating the collected raw data to generate a high-quality time series data matrix to prepare for model input;

[0132] The maternal-fetal interaction module is responsible for building a maternal-fetal interaction model using a bidirectional long short-term memory network (Bi-LSTM), capturing the dynamic relationship between the mother and the fetus, optimizing model parameters, and calculating prenatal risk scores;

[0133] The risk assessment module divides the risk level (low, medium, high) according to the prenatal risk score and dynamically adjusts the threshold based on historical data;

[0134] The intervention recommendation module provides personalized intervention recommendations based on risk assessment results to help users take appropriate measures to reduce prenatal risks.

[0135] This embodiment also provides a computer device, which is suitable for the case of a prenatal monitoring risk assessment method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the prenatal monitoring risk assessment method proposed in the above embodiment.

[0136] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0137] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the prenatal monitoring risk assessment method proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0138] In summary, the present invention constructs a comprehensive and efficient prenatal monitoring risk assessment method by combining multi-source data collection, preprocessing, maternal-fetal dynamic modeling, risk level classification and personalized intervention recommendations. It solves the problems of single data and poor timeliness in traditional prenatal monitoring by real-time collection and integration of maternal, fetal and environmental data. It captures the complex dynamic interaction relationship between mother and fetus through data preprocessing and deep modeling, and improves the accuracy of risk assessment. It realizes scientific and dynamic health management through risk level classification and personalized intervention recommendations. Overall, the invention has significant technical advantages in improving the accuracy and guidance of prenatal risk monitoring.

[0139] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Antenatal monitoring risk assessment method, characterized in that: include: Through wearable devices, fetal heart rate monitors and ultrasound equipment, the physiological data of the mother and fetus are collected in real time to generate a time series data matrix; Preprocess the collected physiological data; The bidirectional long short-term memory network Bi-LSTM is used to build a maternal-fetal interaction model. Convert to time series data tensor , transform the time series data tensor The maternal-fetal interaction model is input to capture the dynamic correlation between maternal and fetal physiological data. The attention mechanism is introduced to dynamically adjust the weight of the correlation matrix. Based on the optimized dynamic correlation matrix, the prenatal risk score R is calculated. At the same time, the maternal-fetal interaction model is optimized by inputting historical data into the maternal-fetal interaction model and using the attention mechanism to dynamically adjust the parameter weights; According to the prenatal risk score, the risk is divided into three risk levels: low, medium and high, as follows: Collect historical physiological data and input them into the risk score calculation formula to obtain the historical prenatal risk score ; ； in, , and Represent the output values ​​of low risk, medium risk and high risk levels respectively, represents the mean of the historical prenatal risk scores, represents the standard deviation of historical prenatal risk scores, and is the dynamically adjusted proportional coefficient, represents the smoothing parameter, represents the exponential function, represents the proportionality coefficient; When calculated When , it indicates low risk; When calculated When , it indicates medium risk; When calculated , indicating high risk.

2. The method for prenatal monitoring risk assessment according to claim 1, wherein: The physiological data of the mother and the fetus are collected in real time by using wearable devices, fetal heart rate monitors and ultrasound equipment. The specific steps are as follows: Use FDA-approved smart bracelets to collect core maternal physiological data; Use a Doppler-based fetal heart rate monitor to collect fetal heart rate and fetal movement frequency, and use ultrasound equipment to regularly collect fetal amniotic fluid index and fetal position change data; Data from all devices is transmitted to the central server in real time via the 5G network; The data is stored in the efficient distributed database MongoDB and timestamped; Integrate the collected data to generate a time series data matrix, the expression is: ； in, Indicates at time The time series data matrix of time, Indicates the number of acquisition devices. Indicates the start time of collection. Indicates the end time of collection. The index variable representing the acquisition device. Indicates The weight corresponding to each device, Indicates Devices at time Physiological data collected at all times , Indicates Devices at time The physiological data collected, Indicates A normalization function, Represents a small time increment in an integration operation.

3. The prenatal monitoring risk assessment method according to claim 2, characterized in that: The preprocessing of the collected physiological data is specifically as follows: Set abnormal thresholds for the collected physiological data Perform outlier detection and elimination, the expression is: ； ； in, Indicates Devices at time The abnormal flag, the value of 1 indicates abnormality, and the value of 0 indicates normality. Represents the dynamic proportional coefficient for outlier determination, Indicates the first Devices at time The physiological data collected, Indicates The average value of the data collected by each device, Indicates The standard deviation of the data collected by each device; According to the cleaning Devices at time Physiological data collected Update time series data matrix .

4. The method for prenatal monitoring risk assessment according to claim 3, wherein: The bidirectional long short-term memory network Bi-LSTM is used to construct the maternal-fetal interaction model, specifically: We selected the bidirectional long short-term memory network Bi-LSTM as the basis of the maternal-fetal interaction model; The time series data matrix is a two-dimensional matrix, representing the time The data feature set has the dimension ,in, Indicates the number of acquisition devices. Indicates the number of features; The tensor representation of time series data is a three-dimensional data structure, represented in arrive Multidimensional maternal, fetal, and environmental data collected over the entire time range, with dimensions of ,in, Represents the length of the time series, and the expression is: ,in, Indicates belonging to, Represents the field of real numbers.

5. The method for prenatal monitoring risk assessment according to claim 4, characterized in that: The preprocessed physiological data are input into the maternal-fetal interaction model to capture the dynamic relationship between maternal and fetal physiological data and output the prenatal risk score. The specific steps are as follows: Convert the time series data tensor Input the maternal-fetal interaction model, extract features from the three-dimensional data, and obtain the hidden state. The expression is: ； in, Represents the maternal-fetal interaction model at time The hidden state of is a bidirectional long short-term memory network, Represents a time series tensor at time slices, and are the hidden state and memory state of the previous time step respectively; The dynamic relationship between the mother and the fetus is captured according to the hidden state, expressed as: ； in, and The index variable representing the feature, Indicates the first and The dynamic correlation between the features, represents the length of the time series, and Respectively expressed in time Time and The hidden state of a feature, Indicates at time Time and The interaction eigenvalues ​​between the features are is the activation function Sigmoid, represents the hyperbolic tangent function, Indicates based on and The exponential decay function of , and is the weight parameter; The attention optimization formula is: ； ； in, Indicates at time The attention weight vector, is the activation function Softmax function, represents the weight matrix of the attention mechanism, represents the bias term of the attention mechanism, Indicates at time The dynamic incidence matrix of Indicates at time The optimized dynamic correlation matrix; The risk score calculation formula is: ； in, represents the prenatal risk score, is the activation function Sigmoid, and Represents the weights and biases for risk score calculation.

6. The prenatal monitoring risk assessment method according to claim 5, characterized in that: According to the risk level, specific intervention recommendations are given, including: When the risk level is low, basic health advice is provided to maintain a balanced diet and exercise moderately; When the risk level is medium, monitor the patient based on the doctor's advice and consider mild intervention measures, such as adjusting lifestyle or supplementing nutrition. When the risk level is high, seek medical attention immediately, conduct a comprehensive examination based on the doctor's advice, and initiate medical intervention.

7. A prenatal monitoring risk assessment system, based on the prenatal monitoring risk assessment method according to any one of claims 1 to 6, characterized in that: It includes acquisition module, pre-processing module, maternal-fetal interaction module, risk assessment module and intervention suggestion module; The acquisition module is responsible for collecting physiological and environmental data of the mother, fetus and environment in real time; The preprocessing module is responsible for cleaning, standardizing and interpolating the collected raw data to generate a high-quality time series data matrix; The maternal-fetal interaction module is responsible for constructing a maternal-fetal interaction model using a bidirectional long short-term memory network Bi-LSTM, capturing the dynamic relationship between the mother and the fetus, optimizing model parameters, and calculating the prenatal risk score; The risk assessment module divides the risk level according to the prenatal risk score and dynamically adjusts the threshold based on historical data; The intervention suggestion module provides personalized intervention suggestions based on the risk assessment results.

Images