Obstetrical fetal heart intelligent monitoring method and system

Abstract

The invention relates to the technical field of obstetrical fetal heart monitoring, in particular to an obstetrical fetal heart intelligent monitoring method and system, and the method comprises the following steps: S1, data collection: collecting fetal heart signals through wearable equipment; s2, data preprocessing: acquiring fetal heart monitoring parameters; s3, self-adaptive adjustment of the personalized physiological state: dynamically adjusting fetal heart monitoring parameters by applying a self-adaptive algorithm; s4, data analysis and prediction, wherein the adjusted fetal heart monitoring parameters are analyzed, and the health problem of the fetus is predicted; s5, dynamic multi-mode signal interaction analysis: identifying relevance among multi-mode signals, performing abnormal mode detection, and performing comprehensive evaluation; and S6, visualization and user interaction: developing a user interface based on Web and mobile applications. According to the invention, the relevance among various signals can be comprehensively captured and analyzed, the abnormal mode can be accurately identified, and the accuracy and reliability of fetal health monitoring can be improved.

Description

Technical Field

[0001] This invention relates to the field of obstetric fetal heart rate monitoring technology, and in particular to an intelligent method and system for obstetric fetal heart rate monitoring. Background Technology

[0002] With the advancement of technology, obstetric monitoring techniques have continued to develop, and fetal heart rate monitoring has become a key means of assessing fetal health. Fetal heart rate monitoring can monitor the fetal heart rate and cardiac frequency in real time, helping doctors to promptly detect and address various abnormal conditions of the fetus in utero. Modern fetal heart rate monitoring devices are mostly wearable devices that transmit the collected fetal heart signals to a data processing terminal via wireless communication technology, making it convenient for doctors and pregnant women to view the fetus's health status in real time. However, a single fetal heart rate monitoring signal is often insufficient to fully reflect the fetus's health status and needs to be combined with the pregnant woman's physiological state and external environmental signals for comprehensive analysis.

[0003] Existing fetal heart rate monitoring systems have several shortcomings in practical applications. First, noise and interference significantly affect fetal heart rate signals, and current technologies have limitations in noise removal and signal enhancement, resulting in insufficient clarity and accuracy of monitoring data. Second, existing systems typically lack dynamic adjustments to the individual physiological state of pregnant women, failing to compensate for signal deviations caused by physiological fluctuations and affecting the reliability of monitoring results. Furthermore, their ability to fuse and analyze multimodal signals is limited, failing to fully utilize maternal physiological signals and environmental signals to improve the accuracy and comprehensiveness of abnormal pattern detection, leading to poor prediction and early warning effects on fetal health status and hindering comprehensive and accurate health monitoring.

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent fetal heart rate monitoring method and system for obstetrics, which can help doctors intervene in a timely manner and improve the overall performance and user experience of fetal health monitoring. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides a method and system for intelligent fetal heart rate monitoring in obstetrics.

[0006] A method for intelligent fetal heart rate monitoring in obstetrics includes the following steps: S1, Data Acquisition: Fetal heart rate signals are acquired through wearable devices and transmitted via wireless communication technology; S2, Data Preprocessing: The collected fetal heart rate signal is removed and enhanced by a filtering algorithm. The fetal heart rate signal is extracted by a time-domain and frequency-domain analysis method to obtain fetal heart rate monitoring parameters, including fetal heart rate and fetal movement frequency. S3, Personalized Physiological Adaptive Adjustment: Combining the pregnant woman's individual physiological data and past health records, an adaptive algorithm is applied to dynamically adjust fetal heart rate monitoring parameters, specifically including: S31, Individual physiological data collection: Real-time collection of pregnant women's individual physiological data, including blood pressure, heart rate and body temperature, and acquisition of pregnant women's past health records, including past medical history and allergy history; S32, Data Fusion and Matching: The collected individual physiological data, previous health records and fetal heart signals are fused into comprehensive data, and data features with correlation above a predetermined threshold are found through data matching algorithms; S33, Adaptive parameter adjustment: Based on the fused comprehensive data, the Bayesian estimation algorithm is applied to dynamically adjust the fetal heart rate monitoring parameters to compensate for signal deviations caused by physiological fluctuations in pregnant women; S4, Data Analysis and Prediction: By analyzing the adjusted fetal heart rate monitoring parameters through the fetal heart rate health prediction model, we can predict the health problems that may occur in the fetus. S5, Dynamic Multimodal Signal Interaction Analysis: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, detect abnormal patterns, and perform comprehensive evaluation. S6, Visualization and User Interaction: Develop a web- and mobile application-based user interface that displays real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends, and supports users in viewing health reports and obtaining health guidance and support.

[0007] Optionally, the data preprocessing in S2 includes: S21, Noise Removal and Signal Enhancement: The acquired fetal heart rate signal is processed using a bandpass filter to remove low-frequency and high-frequency noise; S22, Signal Enhancement: The Kalman filter algorithm is used to enhance the denoised signal. The Kalman filter estimates the signal through recursive minimum mean square error. S23, Feature Extraction: A time-domain and frequency-domain-based analysis method is used to extract features from the filtered and enhanced fetal heart rate signal. Time-domain analysis includes calculating the instantaneous heart rate (HR) of the signal, and frequency-domain analysis includes performing a Fast Fourier Transform (FFT) on the signal to extract frequency components, as shown below: ; in, The time interval between adjacent R waves; ; in, It is a frequency domain signal. For time-domain signals, For signal length, For frequency.

[0008] Optionally, the individual physiological data collection in S31 includes: S311, Real-time collection of pregnant women's individual physiological data: Real-time monitoring of pregnant women's blood pressure, heart rate, and body temperature through wearable devices (such as smart bracelets or smart patches), specifically including: Blood pressure monitoring: Blood pressure is measured in real time using a combination of photoplethysmography (PPG) and piezoelectric sensors and a pulse wave conduction time (PWTT) algorithm. Heart rate monitoring: Uses a built-in PPG sensor to measure heart rate in real time via optical methods; Body temperature monitoring: Uses a built-in temperature sensor to measure body temperature in real time through direct skin contact; S312, Obtaining the pregnant woman's past health records: By connecting to the electronic health record (EHR) platform, obtain the pregnant woman's past medical history and allergy history records, specifically including: Data interface: Extract relevant health records from the electronic health record platform through a data interface (such as HL7 or FHIR); Data integration: Integrate the acquired past health records with the real-time collected individual physiological data to form an individual health data profile.

[0009] Optionally, the data fusion and matching in S32 includes: S321, Data Fusion: Integrating real-time collected individual physiological data, acquired past health records, and fetal heart rate signal data to form a comprehensive dataset; S322, Data Matching: Using the Kalman filter algorithm, find data features with correlations above a predetermined threshold, represented as: S3221, State estimation: Initial state estimation Covariance Matrix ; S3222, Prediction Phase: ; ; in, It is a priori state estimation. It is the posterior state estimate of the previous moment. It is the state transition matrix. It is a control matrix. It is a control input. It is the covariance estimated a priori. It is the process noise covariance; S3223, Update Phase: ; ; ; in, It is Kalman gain. These are measured values. It is a measurement matrix. It measures the noise covariance. It is a posterior state estimate. It is the covariance of the posterior estimate; S3224, Correlation Determination: Based on the updated state estimate and covariance matrix, calculate the correlation between each data feature and set a threshold. Data features with correlation values ​​above a predetermined threshold are selected and represented as follows: .

[0010] Optionally, the adaptive parameter adjustment in S33 includes: S331, Prior Distribution: Define the prior distribution of the fetal heart rate monitoring parameters. Assume the prior distribution is a normal distribution, expressed as: ; in, For fetal heart rate monitoring parameters, and These are the prior mean and variance, respectively; S332, Likelihood Function: Based on real-time collected individual physiological data, past health records, and fetal heart rate signals, a likelihood function is established, expressed as: ; in, For observation data, The variance of the observed noise; S333, Posterior Distribution: According to Bayes' theorem, the posterior distribution is calculated and expressed as: ; ; ; in, and These are the mean and variance of the posterior distribution, respectively; S334, Parameter Adjustment: Based on the mean of the posterior distribution, the fetal heart rate monitoring parameters are dynamically adjusted, as follows: ; ; in, This is the posterior mean obtained using the Bayesian estimation algorithm.

[0011] Optionally, the fetal heart rate health prediction model in S4 adopts a recurrent neural network (RNN) model, which includes: S41, Input layer: The input consists of adjusted fetal heart rate monitoring parameters, related individual physiological data, and past health records; S42, Dynamic Time Warping: Dynamically warps the input data to adjust the time scale of different time series, expressed as: ; in, and There are two time series. For time points and The distance between them; S43, Bidirectional RNN: Using a bidirectional RNN to process time series data, it captures information from both forward and backward directions, represented as: ; ; in, It is in a forward-hidden state. It is in a backward hidden state. and For the input weight matrix, and The hidden state weight matrix, and For bias; S44, Attention Mechanism: An attention mechanism is introduced based on the output of the bidirectional RNN to calculate the attention weight at each time step, expressed as: ; ; in, To score attention, For attention weight vectors, and This is the weight matrix. For bias, For time steps Attention weights; S45, Weighted Summation: The output of the bidirectional RNN is weighted and summed according to the attention weights to obtain the context vector, represented as: ; S46, Output Layer: Uses a linear transformation to convert the context vector into predicted values, represented as: ; in, For time The predicted value, To output the weight matrix, For the bias of the output layer; S47, Adaptive learning rate: Model training using the Adam optimizer, expressed as: ; ; ; ; ; in, and For estimation of first and second moments, and The exponential decay rate, For gradient, For learning rate, It is a tiny constant.

[0012] Optionally, the dynamic multimodal signal interaction analysis in S5 includes: S51, Multimodal signal feature extraction: Fourier transform and wavelet analysis are performed on fetal heart signals, maternal physiological signals and environmental signals to extract frequency domain and time-frequency domain features of fetal heart signals, maternal physiological signals and environmental signals, forming multimodal signal features; S52, Signal Fusion and Correlation Analysis: The extracted multimodal signal features are standardized and time-synchronized, fused into a comprehensive signal dataset, and the correlation between different signals is calculated to construct a signal correlation matrix; S53, Abnormal Pattern Detection and Comprehensive Evaluation: Based on the fused comprehensive signal dataset and correlation analysis results, anomaly patterns in the signal are detected using pattern recognition algorithms, a comprehensive evaluation report is generated, and early warning information is provided.

[0013] Optionally, the signal fusion and correlation analysis in S52 includes: S521, Signal Standardization Processing: Standardize the extracted multimodal signal features; S522, Time Synchronization Processing: Time alignment of signals from different sources is performed through interpolation, represented as: ; in, The signal value at the interpolation point. and For the time of a known point, and The signal value at the known point; S523, Signal Fusion: This involves fusing the standardized and synchronized features of multimodal signals into a single comprehensive signal dataset, represented as: ; in, For the comprehensive signal dataset, , , These are the characteristics of standardized and synchronized fetal heart rate signals, maternal physiological signals, and environmental signals, respectively. S524, Correlation Calculation: The correlation between different signals is calculated using the Pearson correlation coefficient; S525, Constructing the Signal Correlation Matrix: Based on the calculated correlation coefficients, a signal correlation matrix is ​​constructed to represent the correlation strength between different signal features. elements Indicate signal characteristics and signal characteristics The correlation coefficient between them is expressed as: .

[0014] Optionally, the abnormal pattern detection and comprehensive evaluation in S53 includes: S531, Data Input: The integrated signal dataset and signal correlation matrix obtained from signal fusion and correlation analysis are used as input data; S532: Anomaly Pattern Detection: This section describes the detection of anomalous patterns in a composite signal dataset using a Gaussian Mixture Model (GMM)-based clustering algorithm. Specifically, it includes: S5321: Parameter estimation of Gaussian mixture model: ; in, Input signal In the model The probability density under, The number of Gaussian components, For the first The weights of each component and The first The mean and covariance matrix of each component; S5322: Parameter estimation using the Expectation-Maximization (EM) algorithm: : ; : ; ; ; in, It is the first The responsibility of each Gaussian component It is the first One sample point, The total number of samples; S533: Generate a comprehensive evaluation report: Based on the abnormal pattern detection results, generate a comprehensive evaluation report, including a description of the abnormality of each signal feature, the performance and health risks of the abnormal pattern, and a summary of the signal correlation analysis results; S534: Provide early warning information: Based on the comprehensive assessment report, generate early warning information and promptly notify doctors or relevant personnel to intervene, as shown in: ; in, Let be the probability density of the signal under the Gaussian mixture model. The set warning threshold.

[0015] An intelligent fetal heart rate monitoring system for obstetrics, used to implement the above-mentioned intelligent fetal heart rate monitoring method for obstetrics, includes the following modules: Data acquisition module: Collects fetal heart signals through wearable devices and transmits the collected fetal heart signals through wireless communication technology; Data preprocessing module: The module uses filtering algorithms to remove noise and enhance the acquired fetal heart rate signals, and uses time-domain and frequency-domain analysis methods to extract features from the fetal heart rate signals to obtain fetal heart monitoring parameters, including fetal heart rate and fetal movement frequency. Personalized physiological state adaptive adjustment module: Combining the pregnant woman's individual physiological data and past health records, the module uses an adaptive algorithm to dynamically adjust the fetal heart rate monitoring parameters; Data Analysis and Prediction Module: Through the fetal heart health prediction model, the adjusted fetal heart monitoring parameters are analyzed to predict fetal health problems; Dynamic multimodal signal interaction analysis module: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, detect abnormal patterns, and perform comprehensive evaluation; Visualization and User Interaction Module: Develop a web- and mobile application-based user interface to display real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends. It also supports users in viewing health reports and obtaining health guidance and support.

[0016] The beneficial effects of this invention are: This invention acquires fetal heart rate signals in real time through wearable devices and uses filtering algorithms for noise removal and signal enhancement to ensure data clarity and accuracy. Simultaneously, it employs time-domain and frequency-domain analysis methods to extract features from the fetal heart rate signals, obtaining key fetal heart rate monitoring parameters. Through personalized physiological adaptive adjustments, combined with the pregnant woman's individual physiological data and past health records, a Bayesian estimation algorithm is applied to dynamically adjust the fetal heart rate monitoring parameters, compensating for signal deviations caused by physiological fluctuations in the pregnant woman. This effectively improves the accuracy and reliability of fetal heart rate monitoring, providing a solid data foundation for subsequent health status assessment and early warning.

[0017] This invention introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis to extract frequency and time-frequency domain features of multimodal signals, performs standardization and time synchronization processing, and fuses them into a comprehensive signal dataset. The correlation between different signals is calculated using the Pearson correlation coefficient to construct a signal correlation matrix. Based on the comprehensive signal dataset and correlation analysis results, a clustering algorithm based on a Gaussian mixture model is used to detect abnormal patterns, generate a comprehensive assessment report, and provide early warning information. This invention can comprehensively capture and analyze the correlation between multiple signals, accurately identify abnormal patterns, and improve the accuracy and reliability of fetal health monitoring.

[0018] This invention allows users to access monitoring data anytime, anywhere via mobile applications and web pages, understand their own and their fetus's health status, and generate comprehensive assessment reports and early warning information to help doctors intervene in a timely manner, promptly detect and address potential fetal health problems, improve the level of maternal and infant health protection, and enhance the user's overall experience and early intervention capabilities. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the monitoring method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the system functional modules according to an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0022] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0023] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0024] like Figure 1 As shown, a method for intelligent fetal heart rate monitoring in obstetrics includes the following steps: S1, Data Acquisition: Fetal heart rate signals are acquired through wearable devices and transmitted via wireless communication technology; S2, Data Preprocessing: The collected fetal heart signals are subjected to noise removal and signal enhancement using filtering algorithms to ensure the clarity and accuracy of the data. The fetal heart signals are then subjected to feature extraction using time-domain and frequency-domain analysis methods to obtain fetal heart monitoring parameters, including fetal heart rate and fetal movement frequency. S3, Personalized Physiological Adaptive Adjustment: Combining the pregnant woman's individual physiological data and past health records, an adaptive algorithm is applied to dynamically adjust fetal heart rate monitoring parameters to compensate for signal deviations caused by physiological fluctuations in the pregnant woman. Specifically, this includes: S31, Individual physiological data collection: Real-time collection of pregnant women's individual physiological data, including blood pressure, heart rate and body temperature, and acquisition of pregnant women's past health records, including past medical history and allergy history; S32, Data Fusion and Matching: The collected individual physiological data, previous health records and fetal heart signals are fused into comprehensive data, and data features with correlation above a predetermined threshold are found through data matching algorithms; S33, Adaptive parameter adjustment: Based on the fused comprehensive data, the Bayesian estimation algorithm is applied to dynamically adjust the fetal heart rate monitoring parameters to compensate for signal deviations caused by physiological fluctuations in pregnant women; S4, Data Analysis and Prediction: By analyzing the adjusted fetal heart rate monitoring parameters through the fetal heart rate health prediction model, we can predict fetal health problems and help doctors intervene in a timely manner. S5, Dynamic Multimodal Signal Interaction Analysis: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, conducts comprehensive abnormal pattern detection, and performs comprehensive evaluation to provide more comprehensive health status reports and accurate early warning information; S6, Visualization and User Interaction: Develop a web- and mobile application-based user interface to display real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends. It also supports users in viewing health reports and obtaining health guidance and support. Through mobile applications and web pages, users can access monitoring data anytime, anywhere to understand their own and their fetus's health status. Through the above steps, more accurate, personalized, and comprehensive fetal heart rate health monitoring is achieved, which helps to detect and intervene in potential fetal health problems in a timely manner, and improves the reliability of monitoring and user experience.

[0025] Data preprocessing in S2 includes: S21, Noise Removal and Signal Enhancement: The acquired fetal heart rate signal is processed using a bandpass filter to remove low-frequency and high-frequency noise, retaining the effective frequency band of the fetal heart rate signal, as shown below: ; in, and These are the low-frequency and high-frequency cutoff frequencies of the bandpass filter, respectively. S22, Signal Enhancement: The Kalman filter algorithm is used to enhance the denoised signal. The Kalman filter estimates the signal through recursive minimum mean square error, expressed as: ; ; in, Estimate the current state. Estimate the previous state. For Kalman gain, For the observed values, For the observation matrix, The covariance estimated from the previous state. To observe the noise covariance; S23, Feature Extraction: A time-domain and frequency-domain-based analysis method is used to extract features from the filtered and enhanced fetal heart rate signal. Time-domain analysis includes calculating the instantaneous heart rate (HR) of the signal, and frequency-domain analysis includes performing a Fast Fourier Transform (FFT) on the signal to extract frequency components, as shown below: ; in, The time interval between adjacent R waves; ; in, It is a frequency domain signal. For time-domain signals, For signal length, For frequency; The above measures effectively improve the clarity and accuracy of the signals, ensuring the reliable extraction of key fetal heart rate monitoring parameters (such as fetal heart rate and fetal movement frequency), thereby providing a solid data foundation for subsequent health status assessment and early warning, and enhancing the overall performance and reliability of the monitoring system.

[0026] Individual physiological data collection in S31 includes: S311, Real-time collection of pregnant women's individual physiological data: Real-time monitoring of pregnant women's blood pressure, heart rate, and body temperature through wearable devices (such as smart bracelets or smart patches), specifically including: Blood pressure monitoring: Blood pressure is measured in real time using a combination of photoplethysmography (PPG) and piezoelectric sensors, employing a pulse wave transit time (PWTT) algorithm, and is expressed as follows: ; in, This is the blood pressure value. The pulse wave conduction time. and These are constants determined through calibration; Heart rate monitoring: Using a built-in PPG sensor, heart rate is measured in real time via optical methods and expressed as: ; in, Heart rate, The time interval (in seconds) between adjacent R waves; Body temperature monitoring: Using a built-in temperature sensor, body temperature is measured in real time through direct skin contact, and is displayed as follows: ; in, This is the actual body temperature. The temperature measured by the sensor, This is a correction value; S312, Obtaining the pregnant woman's past health records: By connecting to the electronic health record (EHR) platform, obtain the pregnant woman's past medical history and allergy history records, specifically including: Data interface: Extract relevant health records from the electronic health record platform through a data interface (such as HL7 or FHIR); Data integration: Integrate the acquired past health records with the real-time collected individual physiological data to form an individual health data profile; The above ensures the comprehensiveness and accuracy of the data, providing a solid data foundation for personalized physiological adaptive adjustment, and greatly improving the overall performance and reliability of the fetal heart rate monitoring system. This helps to detect and intervene in potential health problems early and improve the level of maternal and infant health protection.

[0027] Data fusion and matching in S32 include: S321, Data Fusion: Integrating real-time collected individual physiological data, acquired past health records, and fetal heart rate signal data to form a comprehensive dataset, specifically including: Data standardization: Standardize data from different sources to ensure consistency in data format and units; Data synchronization: Time-align real-time data and historical data to ensure synchronization between different data sources; S322, Data Matching: Using the Kalman filter algorithm, find data features with correlations above a predetermined threshold, represented as: S3221, State estimation: Initial state estimation Covariance Matrix ; S3222, Prediction Phase: ; ; in, It is a priori state estimation. It is the posterior state estimate of the previous moment. It is the state transition matrix. It is a control matrix. It is a control input. It is the covariance estimated a priori. It is the process noise covariance; S3223, Update Phase: ; ; ; in, It is Kalman gain. These are measured values. It is a measurement matrix. It measures the noise covariance. It is a posterior state estimate. It is the covariance of the posterior estimate; S3224, Correlation Determination: Based on the updated state estimate and covariance matrix, calculate the correlation between each data feature and set a threshold. Data features with correlation values ​​above a predetermined threshold are selected and represented as follows: ; The above methods can effectively integrate and analyze pregnant women's individual physiological data, past health records, and fetal heart signals, extracting data features with correlations above a predetermined threshold determined by the Kalman filter algorithm. This provides accurate data support for subsequent personalized physiological state adaptive adjustment and health monitoring.

[0028] The adaptive parameter adjustment in S33 includes: S331, Prior Distribution: Define the prior distribution of the fetal heart rate monitoring parameters. Assume the prior distribution is a normal distribution, expressed as: ; in, For fetal heart rate monitoring parameters, and These are the prior mean and variance, respectively; S332, Likelihood Function: Based on real-time collected individual physiological data, past health records, and fetal heart rate signals, a likelihood function is established, expressed as: ; in, For observation data, The variance of the observed noise; S333, Posterior Distribution: According to Bayes' theorem, the posterior distribution is calculated and expressed as: ; ; ; in, and These are the mean and variance of the posterior distribution, respectively; S334, Parameter Adjustment: Based on the mean of the posterior distribution, the fetal heart rate monitoring parameters are dynamically adjusted, as follows: ; ; in, The posterior mean is obtained using the Bayesian estimation algorithm; Through the above steps, based on the fused comprehensive data, the fetal heart rate monitoring parameters can be dynamically adjusted using a Bayesian estimation algorithm to compensate for signal deviations caused by physiological fluctuations in pregnant women, thereby improving the accuracy and reliability of fetal heart rate monitoring and providing high-quality data support for subsequent health monitoring and analysis.

[0029] The fetal heart rate health prediction model in S4 uses a recurrent neural network (RNN) model, which includes: S41, Input layer: The input consists of adjusted fetal heart rate monitoring parameters, related individual physiological data, and past health records; S42, Dynamic Time Warping: Dynamically warps the input data to adjust the time scale of different time series, expressed as: ; in, and There are two time series. For time points and The distance between them; S43, Bidirectional RNN: Using a bidirectional RNN to process time series data, it captures information from both forward and backward directions, represented as: ; ; in, It is in a forward-hidden state. It is in a backward hidden state. and For the input weight matrix, and The hidden state weight matrix, and For bias; S44, Attention Mechanism: An attention mechanism is introduced based on the output of the bidirectional RNN to calculate the attention weight at each time step, expressed as: ; ; in, To score attention, For attention weight vectors, and This is the weight matrix. For bias, For time steps Attention weights; S45, Weighted Summation: The output of the bidirectional RNN is weighted and summed according to the attention weights to obtain the context vector, represented as: ; S46, Output Layer: Uses a linear transformation to convert the context vector into predicted values, represented as: ; in, For time The predicted value, To output the weight matrix, For the bias of the output layer; S47, Adaptive learning rate: Model training using the Adam optimizer, expressed as: ; ; ; ; ; in, and For estimation of first and second moments, and The exponential decay rate, For gradient, For learning rate, It is a tiny constant; The above methods enable better processing of time series data, capture of important time features, and improved prediction accuracy and generalization ability, making the model more suitable for complex fetal heart rate monitoring data analysis and prediction tasks.

[0030] Dynamic multimodal signal interaction analysis in S5 includes: S51, Multimodal signal feature extraction: Fourier transform and wavelet analysis are performed on fetal heart signals, maternal physiological signals and environmental signals to extract frequency domain and time-frequency domain features of fetal heart signals, maternal physiological signals and environmental signals, forming multimodal signal features; S52, Signal Fusion and Correlation Analysis: The extracted multimodal signal features are standardized and time-synchronized, fused into a comprehensive signal dataset, and the correlation between different signals is calculated to construct a signal correlation matrix; S53, Abnormal Pattern Detection and Comprehensive Evaluation: Based on the fused comprehensive signal dataset and correlation analysis results, anomaly patterns in the signal are detected using pattern recognition algorithms, a comprehensive evaluation report is generated, and early warning information is provided; The above methods enable comprehensive capture and analysis of the correlation between various signals, accurate identification of abnormal patterns, improved accuracy and reliability of fetal health monitoring, timely detection of potential problems, and assistance to doctors in early intervention.

[0031] Multimodal signal feature extraction in S51 includes: S511, Fetal Heart Rate Signal Extraction: Perform Fourier transform on the fetal heart rate signal to extract its frequency domain features, and perform wavelet transform on the fetal heart rate signal to extract its time-frequency domain features, as shown below: ; in, This represents the frequency domain of the fetal heart rate signal. This is the time-domain representation of the fetal heart rate signal. For signal length, For frequency; ; in, The wavelet transform coefficients of the fetal heart rate signal are... This is a fetal heartbeat signal. For the mother wavelet, For scale parameters, These are translation parameters; S512, Maternal Physiological Signal Extraction: Fourier transform is performed on maternal physiological signals (such as blood pressure, heart rate, and body temperature) to extract their frequency domain features. Wavelet transform is also performed on the maternal physiological signals to extract their time-frequency domain features, as shown below: ; in, This represents the frequency domain of maternal physiological signals. This represents the time domain of maternal physiological signals. ; in, These are the wavelet transform coefficients of the maternal physiological signal. These are physiological signals from the mother. S513, Environmental Signal Extraction: Perform Fourier transform on environmental signals (such as noise level and temperature) to extract their frequency domain features, and perform wavelet transform on the environmental signals to extract their time-frequency domain features, as shown below: ; in, This represents the frequency domain of the environmental signal. This represents the time domain of the environmental signal. ; in, These are the wavelet transform coefficients of the environmental signal. For environmental signals; Through the above steps, multimodal signal feature extraction can effectively perform Fourier transform and wavelet analysis on fetal heart signals, maternal physiological signals, and environmental signals, extracting frequency domain and time-frequency domain features of each signal to form multimodal signal features, providing a comprehensive data foundation for subsequent signal fusion and correlation analysis.

[0032] Signal fusion and correlation analysis in S52 include: S521, Signal Standardization Processing: The extracted multimodal signal features are standardized to ensure consistency in data format and units between different signals, represented as: ; in, For the standardized signal, The original signal, The mean of the signal. The standard deviation of the signal; S522, Time Synchronization Processing: Time alignment of signals from different sources is performed through interpolation to ensure signal synchronization, enabling effective fusion and analysis. This is represented as: ; in, The signal value at the interpolation point. and For the time of a known point, and The signal value at the known point; S523, Signal Fusion: This involves fusing the standardized and synchronized features of multimodal signals into a single comprehensive signal dataset, represented as: ; in, For the comprehensive signal dataset, , , These are the characteristics of standardized and synchronized fetal heart rate signals, maternal physiological signals, and environmental signals, respectively. S524, Correlation Calculation: The correlation between different signals is calculated using the Pearson correlation coefficient, expressed as follows: ; in, For signal and The correlation coefficient between them and Let the characteristic values ​​of the two signals be... and The mean of the signal; S525, Constructing the Signal Correlation Matrix: Based on the calculated correlation coefficients, a signal correlation matrix is ​​constructed to represent the correlation strength between different signal features. elements Indicate signal characteristics and signal characteristics The correlation coefficient between them is expressed as: ; The above content enables a comprehensive integration of fetal heart rate signals, maternal physiological signals, and environmental signals, improving the accuracy and reliability of multimodal signal analysis. This helps to accurately identify the relationships between signals, detect abnormal patterns in a timely manner, and provide a solid data foundation for fetal health monitoring and early warning.

[0033] Anomaly pattern detection and comprehensive evaluation in S53 include: S531, Data Input: The integrated signal dataset and signal correlation matrix obtained from signal fusion and correlation analysis are used as input data; S532: Anomaly Pattern Detection: This section describes the detection of anomalous patterns in a composite signal dataset using a Gaussian Mixture Model (GMM)-based clustering algorithm. Specifically, it includes: S5321: Parameter estimation of Gaussian mixture model: ; in, Input signal In the model The probability density under, The number of Gaussian components, For the first The weights of each component and The first The mean and covariance matrix of each component; S5322: Parameter estimation using the Expectation-Maximization (EM) algorithm: : ; : ; ; ; in, It is the first The responsibility of each Gaussian component It is the first One sample point, The total number of samples; S533: Generate a comprehensive evaluation report: Based on the abnormal pattern detection results, generate a comprehensive evaluation report, including a description of the abnormality of each signal feature, the performance and health risks of the abnormal pattern, and a summary of the signal correlation analysis results; S534: Provide early warning information: Based on the comprehensive assessment report, generate early warning information and promptly notify doctors or relevant personnel to intervene, as shown in: ; in, Let be the probability density of the signal under the Gaussian mixture model. The set warning threshold; Warning threshold The specific settings include: Data analysis: Collect and analyze a large amount of historical monitoring data to calculate the probability distribution of normal signals; Statistical threshold: Based on the probability distribution of normal signals, a 95% confidence interval is set, expressed as: ; in, This represents the mean of a normal signal. The standard deviation of a normal signal; Experimental verification: The set threshold was tested in a real dataset or simulation environment. Its impact on the early warning system was evaluated by calculating sensitivity and specificity. By adjusting the threshold, the optimal balance between sensitivity and specificity was found, expressed as: ; ; Dynamic adjustment: Based on real-time monitoring data and feedback, the thresholds are continuously adjusted and optimized to ensure the accuracy and reliability of the early warning system.

[0034] Through the above steps, based on the fused integrated signal dataset and correlation analysis results, anomaly pattern detection can be performed using a Gaussian mixture model to generate a comprehensive assessment report and provide accurate early warning information, which helps to promptly detect and respond to potential fetal health problems.

[0035] like Figure 2 As shown, an intelligent fetal heart rate monitoring system for obstetrics, used to implement the above-mentioned intelligent fetal heart rate monitoring method for obstetrics, includes the following modules: Data acquisition module: Collects fetal heart signals through wearable devices and transmits the collected fetal heart signals through wireless communication technology; Data preprocessing module: The module uses filtering algorithms to remove noise and enhance the acquired fetal heart rate signals, and uses time-domain and frequency-domain analysis methods to extract features from the fetal heart rate signals to obtain fetal heart monitoring parameters, including fetal heart rate and fetal movement frequency. Personalized physiological state adaptive adjustment module: Combining the pregnant woman's individual physiological data and past health records, the module uses an adaptive algorithm to dynamically adjust the fetal heart rate monitoring parameters; Data Analysis and Prediction Module: Through the fetal heart health prediction model, the adjusted fetal heart monitoring parameters are analyzed to predict fetal health problems; Dynamic multimodal signal interaction analysis module: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, detect abnormal patterns, and perform comprehensive evaluation; Visualization and User Interaction Module: Develop a web- and mobile application-based user interface to display real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends. It also supports users in viewing health reports and obtaining health guidance and support.

[0036] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0037] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring of fetal heart rate in obstetrics, characterized in that, Includes the following steps: S1, Data Acquisition: Fetal heart rate signals are acquired through wearable devices and transmitted via wireless communication technology; S2, Data Preprocessing: The collected fetal heart rate signal is removed and enhanced by a filtering algorithm. The fetal heart rate signal is extracted by a time-domain and frequency-domain analysis method to obtain fetal heart rate monitoring parameters, including fetal heart rate and fetal movement frequency. S3, Personalized Physiological Adaptive Adjustment: Combining the pregnant woman's individual physiological data and past health records, an adaptive algorithm is applied to dynamically adjust fetal heart rate monitoring parameters, specifically including: S31, Individual physiological data collection: Real-time collection of pregnant women's individual physiological data, including blood pressure, heart rate and body temperature, and acquisition of pregnant women's past health records, including past medical history and allergy history; S32, Data Fusion and Matching: The collected individual physiological data, previous health records and fetal heart signals are fused into comprehensive data, and data features with correlation above a predetermined threshold are found through data matching algorithms; S33, Adaptive parameter adjustment: Based on the fused comprehensive data, the Bayesian estimation algorithm is applied to dynamically adjust the fetal heart rate monitoring parameters to compensate for signal deviations caused by physiological fluctuations in pregnant women; S4, Data Analysis and Prediction: By analyzing the adjusted fetal heart rate monitoring parameters through the fetal heart rate health prediction model, we can predict the health problems that may occur in the fetus. S5, Dynamic Multimodal Signal Interaction Analysis: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, detect abnormal patterns, and perform comprehensive evaluation. S6, Visualization and User Interaction: Develop a web- and mobile application-based user interface that displays real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends, and supports users in viewing health reports and obtaining health guidance and support.

2. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 1, characterized in that, The data preprocessing in S2 includes: S21, Noise Removal and Signal Enhancement: The acquired fetal heart rate signal is processed using a bandpass filter to remove low-frequency and high-frequency noise; S22, Signal Enhancement: The Kalman filter algorithm is used to enhance the denoised signal. The Kalman filter estimates the signal through recursive minimum mean square error. S23, Feature Extraction: A time-domain and frequency-domain-based analysis method is used to extract features from the filtered and enhanced fetal heart rate signal. Time-domain analysis includes calculating the instantaneous heart rate of the signal, and frequency-domain analysis includes performing a Fast Fourier Transform on the signal to extract frequency components, as shown below: ； in, The time interval between adjacent R waves; ； in, It is a frequency domain signal. For time-domain signals, For signal length, For frequency.

3. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 1, characterized in that, The individual physiological data collection in S31 includes: S311, real-time collection of pregnant women's individual physiological data: Real-time monitoring of the pregnant woman's blood pressure, heart rate, and body temperature via a wearable device, specifically including: Blood pressure monitoring: Blood pressure is measured in real time using a combination of photoplethysmography (PPG) and piezoelectric sensors, and a pulse wave transit time algorithm is used. Heart rate monitoring: Uses a built-in PPG sensor to measure heart rate in real time via optical methods; Body temperature monitoring: Uses a built-in temperature sensor to measure body temperature in real time through direct skin contact; S312, Obtaining the pregnant woman's past health records: By connecting to the electronic health record platform, obtain the pregnant woman's past medical history and allergy history records, specifically including: Data interface: Extract relevant health records from the electronic health record platform through the data interface; Data integration: Integrate the acquired past health records with the real-time collected individual physiological data to form an individual health data profile.

4. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 3, characterized in that, The data fusion and matching in S32 includes: S321, Data Fusion: Integrating real-time collected individual physiological data, acquired past health records, and fetal heart rate signal data to form a comprehensive dataset; S322, Data Matching: Using the Kalman filter algorithm, find data features with correlations above a predetermined threshold, represented as: S3221, State estimation: Initial state estimation Covariance Matrix ; S3222, Prediction Phase: ； ； in, It is a priori state estimation. It is the posterior state estimate of the previous moment. It is the state transition matrix. It is a control matrix. It is a control input. It is the covariance estimated a priori. It is the process noise covariance; S3223, Update Phase: ； ； ； in, It is Kalman gain. These are measured values. It is a measurement matrix. It measures the noise covariance. It is a posterior state estimate. It is the covariance of the posterior estimate; S3224, Correlation Determination: Based on the updated state estimate and covariance matrix, calculate the correlation between each data feature and set a threshold. Data features with correlation values ​​above a predetermined threshold are selected and represented as follows: 。 5. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 4, characterized in that, The adaptive parameter adjustment in S33 includes: S331, Prior Distribution: Define the prior distribution of the fetal heart rate monitoring parameters. Assume the prior distribution is a normal distribution, expressed as: ； in, For fetal heart rate monitoring parameters, and These are the prior mean and variance, respectively; S332, Likelihood Function: Based on real-time collected individual physiological data, past health records, and fetal heart rate signals, a likelihood function is established, expressed as: ； in, For observation data, The variance of the observed noise; S333, Posterior Distribution: According to Bayes' theorem, the posterior distribution is calculated and expressed as: ； ； ； in, and These are the mean and variance of the posterior distribution, respectively; S334, Parameter Adjustment: Based on the mean of the posterior distribution, the fetal heart rate monitoring parameters are dynamically adjusted, as follows: ； ； in, This is the posterior mean obtained using the Bayesian estimation algorithm.

6. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 1, characterized in that, The fetal heart rate health prediction model in S4 adopts a recurrent neural network model, which includes: S41, Input layer: The input consists of adjusted fetal heart rate monitoring parameters, related individual physiological data, and past health records; S42, Dynamic Time Warping: Dynamically warps the input data to adjust the time scale of different time series, expressed as: ； in, and There are two time series. For time points and The distance between them; S43, Bidirectional RNN: Using a bidirectional RNN to process time series data, it captures information from both forward and backward directions, represented as: ； ； in, It is in a forward-hidden state. It is in a backward hidden state. and For the input weight matrix, and The hidden state weight matrix, and For bias; S44, Attention Mechanism: An attention mechanism is introduced based on the output of the bidirectional RNN to calculate the attention weight at each time step, expressed as: ； ； in, To score attention, For attention weight vectors, and This is the weight matrix. For bias, For time steps Attention weights; S45, Weighted Summation: The output of the bidirectional RNN is weighted and summed according to the attention weights to obtain the context vector, represented as: ； S46, Output Layer: Uses a linear transformation to convert the context vector into predicted values, represented as: ； in, For time The predicted value, To output the weight matrix, For the bias of the output layer; S47, Adaptive learning rate: Model training using the Adam optimizer, expressed as: ； ； ； ； ； in, and For estimation of first and second moments, and The exponential decay rate, For gradient, For learning rate, It is a tiny constant.

7. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 6, characterized in that, The dynamic multimodal signal interaction analysis in S5 includes: S51, Multimodal signal feature extraction: Fourier transform and wavelet analysis are performed on fetal heart signals, maternal physiological signals and environmental signals to extract frequency domain and time-frequency domain features of fetal heart signals, maternal physiological signals and environmental signals, forming multimodal signal features; S52, Signal Fusion and Correlation Analysis: The extracted multimodal signal features are standardized and time-synchronized, fused into a comprehensive signal dataset, and the correlation between different signals is calculated to construct a signal correlation matrix; S53, Abnormal Pattern Detection and Comprehensive Evaluation: Based on the fused comprehensive signal dataset and correlation analysis results, anomaly patterns in the signal are detected using pattern recognition algorithms, a comprehensive evaluation report is generated, and early warning information is provided.

8. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 7, characterized in that, The signal fusion and correlation analysis in S52 includes: S521, Signal Standardization Processing: Standardize the extracted multimodal signal features; S522, Time Synchronization Processing: Time alignment of signals from different sources is performed through interpolation, represented as: ； in, The signal value at the interpolation point. and For the time of a known point, and The signal value at the known point; S523, Signal Fusion: This involves fusing the standardized and synchronized features of multimodal signals into a single comprehensive signal dataset, represented as: ； in, For the comprehensive signal dataset, , , These are the characteristics of standardized and synchronized fetal heart rate signals, maternal physiological signals, and environmental signals, respectively. S524, Correlation Calculation: The correlation between different signals is calculated using the Pearson correlation coefficient; S525, Constructing the Signal Correlation Matrix: Based on the calculated correlation coefficients, a signal correlation matrix is ​​constructed to represent the correlation strength between different signal features. elements Indicate signal characteristics and signal characteristics The correlation coefficient between them is expressed as: 。 9. The method for intelligent fetal heart rate monitoring in obstetrics according to claim 8, characterized in that, The abnormal pattern detection and comprehensive evaluation in S53 includes: S531, Data Input: The integrated signal dataset and signal correlation matrix obtained from signal fusion and correlation analysis are used as input data; S532: Anomaly Pattern Detection: This section describes the detection of anomalous patterns in a composite signal dataset using a clustering algorithm based on a Gaussian mixture model. Specifically, it includes: S5321: Parameter estimation of Gaussian mixture model: ； in, Input signal In the model The probability density under, The number of Gaussian components, For the first The weights of each component and The first The mean and covariance matrix of each component; S5322: Parameter estimation using the expectation-maximization algorithm: ： ； ： ； ； ； in, It is the first The responsibility of each Gaussian component It is the first One sample point, The total number of samples; S533: Generate a comprehensive evaluation report: Based on the abnormal pattern detection results, generate a comprehensive evaluation report, including a description of the abnormality of each signal feature, the performance and health risks of the abnormal pattern, and a summary of the signal correlation analysis results; S534: Provide early warning information: Based on the comprehensive assessment report, generate early warning information and promptly notify doctors or relevant personnel to intervene, as shown in: ； in, Let be the probability density of the signal under the Gaussian mixture model. The set warning threshold.

10. An intelligent fetal heart rate monitoring system for obstetrics, used to implement the intelligent fetal heart rate monitoring method for obstetrics as described in any one of claims 1-9, characterized in that, Includes the following modules: Data acquisition module: Collects fetal heart signals through wearable devices and transmits the collected fetal heart signals through wireless communication technology; Data preprocessing module: The module uses filtering algorithms to remove noise and enhance the acquired fetal heart rate signals, and uses time-domain and frequency-domain analysis methods to extract features from the fetal heart rate signals to obtain fetal heart monitoring parameters, including fetal heart rate and fetal movement frequency. Personalized physiological state adaptive adjustment module: Combining the pregnant woman's individual physiological data and past health records, the module uses an adaptive algorithm to dynamically adjust the fetal heart rate monitoring parameters; Data Analysis and Prediction Module: Through the fetal heart health prediction model, the adjusted fetal heart monitoring parameters are analyzed to predict fetal health problems; Dynamic multimodal signal interaction analysis module: Based on predicted health problems, it introduces maternal physiological signals and environmental signals, uses Fourier transform and wavelet analysis methods to identify the correlation between multimodal signals, detect abnormal patterns, and perform comprehensive evaluation; Visualization and User Interaction Module: Develop a web- and mobile application-based user interface to display real-time fetal heart rate monitoring charts, health status prediction results, and historical data trends. It also supports users in viewing health reports and obtaining health guidance and support.

Images