An artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system

Abstract

The present application relates to the technical field of signal processing and analysis, and discloses a fetal heart rate and uterine contraction double monitoring system based on artificial intelligence, which comprises a signal decoupling module for constructing a signal decomposition network of multi-channel physiological signals; a signal feature extraction module for extracting a time domain feature set and a frequency domain feature set of a fetal heart rate signal sequence and extracting a time-varying nonlinear feature set of a uterine contraction signal sequence; a coupling relationship analysis module for constructing a heart rate-contraction joint space-time feature map of the time-varying nonlinear feature set, the time domain feature set and the frequency domain feature set, and determining a heart rate-contraction coupling relationship of a fetal heart rate state and a contraction intensity; an abnormal joint index analysis module for calculating a fetal hypoxia probability and a contraction frequency probability of a pregnant woman to determine an abnormal joint index of the pregnant woman; and a monitoring report generation module for generating a monitoring report of the pregnant woman. The present application can improve the accuracy and comprehensiveness of fetal heart rate and uterine contraction double monitoring.

Description

Technical Field

[0001] This invention relates to the field of signal processing and analysis technology, and in particular to an artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions. Background Technology

[0002] Fetal heart rate and uterine contraction dual monitoring is an obstetric monitoring technique that simultaneously and continuously observes and records changes in fetal heart rate and maternal uterine contractions during prenatal and labor processes. By analyzing the dynamic relationship between the two, it assesses the fetus's oxygen supply and health status in the uterus, promptly identifies potential risks such as fetal distress, and thus guides clinical decisions to ensure maternal and infant safety.

[0003] Traditional methods for monitoring fetal heart rate and uterine contractions involve electronic fetal heart rate monitoring. This method uses a Doppler ultrasound probe and a uterine contraction pressure gauge to collect data on the movement of the fetal heart valves and changes in abdominal rigidity during uterine contractions, indirectly analyzing the fetal heart rate and uterine contraction pressure to generate fetal monitoring charts. However, this method relies on simple chart analysis and fails to capture deeper, hidden information within the signals. This results in limited information dimensionality, coarse feature extraction, and consequently, vague assessment results that cannot quantify risk. Summary of the Invention

[0004] This invention provides an artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions, the main purpose of which is to improve the accuracy and comprehensiveness of dual monitoring of fetal heart rate and uterine contractions.

[0005] To achieve the above objectives, the present invention provides an artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions, comprising: The signal decoupling module is used to collect multi-channel physiological signals from pregnant women and construct a signal decomposition network for the multi-channel physiological signals to extract the fetal heart rate signal sequence and uterine contraction signal sequence from the multi-channel physiological signals. The signal feature extraction module is used to extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence. The time-domain feature set includes baseline heart rate, baseline variation, and sine wave morphology. The frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. The module also extracts the time-varying nonlinear feature set of the uterine contraction signal sequence. The time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features. The coupling relationship analysis module is used to construct the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set, so as to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity. The abnormal joint index analysis module is used to calculate the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman based on the heart rate-uterine contraction coupling relationship, so as to determine the abnormal joint index of the pregnant woman. The monitoring report generation module is used to generate a monitoring report for the pregnant woman based on the abnormal joint index.

[0006] Optionally, constructing the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set includes: The time-varying nonlinear characteristics, the time-domain feature set, and the time axis and feature axis of the frequency-domain feature set are uniformly defined. Based on the time axis and the feature axis, construct a two-dimensional feature matrix of the time-varying nonlinear feature sequence, the time-domain feature sequence, and the frequency-domain feature sequence; The two-dimensional feature matrix is ​​standardized to obtain a standardized feature matrix; Analyze the feature relationships in the standardized feature matrix; The standardized feature matrix is ​​visualized to obtain a feature heatmap. The feature relationships are integrated into the feature heatmap to obtain the heart rate-uterine contraction combined spatiotemporal feature map.

[0007] Optionally, the analysis of the fetal heart rate status and uterine contraction intensity of the pregnant woman includes: Construct a feature map attention analysis network for the spatiotemporal feature map of the pregnant woman's heart rate-uterine contraction combination; Identify the feature nodes of the heart rate-uterine contraction combined spatiotemporal feature map; Based on the feature map attention analysis network, calculate the importance scores between the feature nodes; Based on the importance score, the fetal health index and uterine contraction coordination of the pregnant woman are calculated. Based on the fetal health index and the uterine contraction coordination, the fetal heart rate status and uterine contraction intensity of the pregnant woman are determined.

[0008] Optionally, constructing the signal decomposition network for the multi-channel physiological signal includes: Construct the signal decomposition network architecture for the multi-channel physiological signals; The signal decomposition network architecture is constructed with a multi-level processing layer, which includes: a signal preprocessing layer, a spatiotemporal feature extraction layer, a blind source separation simulation layer, and a signal reconstruction layer. Determine the communication protocols and signal processing logic between the multi-level processing layers; Based on the communication protocol and the signal processing logic, the multi-level processing layer and the signal decomposition network architecture are integrated to obtain the signal decomposition network.

[0009] Optionally, the extraction of the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence includes: Determine a fixed window length and a fixed sampling frequency for the fetal heart rate signal sequence; Based on the fixed window length and the fixed sampling frequency, the fetal heart rate signal sequence is segmented to obtain segmented fetal heart rate signals; Analyze the sine wave morphology of the segmented fetal heart rate signal; Calculate the average heart rate of each segment in the segmented fetal heart rate signal; Based on the average heart rate, the segmented fetal heart rate signal is divided into an abnormal heart rate signal segment and a normal heart rate signal segment; Based on the normal heart rate signal segment, determine the baseline heart rate of the fetal heart rate signal sequence; Based on the abnormal heart rate signal segments, determine the baseline variation of the fetal heart rate signal sequence; The time-domain feature set of the fetal heart rate signal sequence is determined based on the baseline heart rate, the variable baseline, and the sine wave morphology. The fetal heart rate signal sequence is frequency domain transformed to obtain the fetal heart rate signal frequency domain sequence; Calculate the power spectral density of the fetal heart rate signal frequency domain sequence to determine the frequency domain energy distribution of the fetal heart rate signal frequency domain sequence; Identify the dominant frequency band of the frequency domain energy distribution and calculate the dominant frequency band power of the dominant frequency band; The frequency domain feature set of the fetal heart rate signal sequence is determined based on the main frequency band power and the frequency domain energy distribution.

[0010] Optionally, the extraction of the time-varying nonlinear feature set of the uterine contraction signal sequence includes: Identify the contraction start point, contraction peak point, and contraction end point of the contraction signal sequence; Based on the contraction start point, the contraction peak point, and the contraction end point, the peak amplitude of the contraction signal sequence is determined; Based on the contraction start point and the contraction end point, calculate the contraction duration and contraction interval of the contraction signal sequence; Based on the peak point of the uterine contractions, determine the rise time and fall time of the uterine contraction signal sequence; Calculate the average amplitude of the contraction signal sequence based on the duration of the contractions; Calculate the amplitude change rate of the uterine contraction signal sequence based on the rise time and the fall time; Based on the duration of the contractions, the contraction frequency of the contraction signal sequence is analyzed; Calculate the standard deviation of the contraction intervals to analyze the periodic variability of the contraction signal sequence; By integrating the peak amplitude, the average amplitude, the amplitude change rate, the contraction duration, the contraction interval, the rise time, the fall time, the contraction frequency, and the periodic variability, a time-varying nonlinear feature set of the contraction signal sequence is obtained.

[0011] Optionally, determining the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity includes: The fetal heart rate status and the uterine contraction intensity were jointly constructed as heart rate-uterine contraction time-series curves. Identify the contraction time intervals of the heart rate-contraction timing curve; Calculate the ratio of the heart rate deceleration area to the contraction time interval of the heart rate-contraction time sequence curve; Identify the contraction rhythm frequency band of the heart rate-contraction timing curve; Calculate the correlation coefficient between the heart rate-contraction time series curve and the contraction-time series curve in the contraction rhythm frequency band; Based on the heart rate deceleration area ratio and the correlation coefficient, the coupling risk index of the heart rate-uterine contraction time series curve is calculated to determine the heart rate-uterine contraction coupling relationship of the heart rate-uterine contraction time series curve.

[0012] Optionally, calculating the probability of fetal hypoxia and the probability of excessive uterine contractions in the pregnant woman based on the heart rate-uterine contraction coupling relationship includes: Extract the coupling relationship features of the heart rate-uterine contraction coupling relationship; Based on the coupling relationship characteristics, fit the fetal hypoxia probability model and the uterine contraction frequency probability model of the pregnant woman; Based on the fetal hypoxia probability model, calculate the probability of fetal hypoxia in the pregnant woman. The probability of excessive uterine contractions in the pregnant woman is calculated based on the aforementioned probability model for excessive uterine contractions.

[0013] Optionally, determining the abnormal combined index of the pregnant woman includes: The probabilities of fetal hypoxia and uterine contraction frequency corresponding to the pregnant woman are nonlinearly transformed to obtain the transformed probabilities of fetal hypoxia and uterine contraction frequency. Determine the hypoxia probability regression coefficient for the change in fetal hypoxia probability and the contraction probability regression coefficient for the change in uterine contraction frequency probability; Calculate the interaction coefficient between the changed probability of fetal hypoxia and the changed probability of excessive uterine contractions; The abnormal combination index of the pregnant woman is calculated based on the regression coefficient of the uterine contraction probability, the regression coefficient of the hypoxia probability, and the interaction coefficient.

[0014] A method for dual monitoring of fetal heart rate and uterine contractions based on artificial intelligence, characterized in that the method includes: Multi-channel physiological signals of pregnant women are collected, and a signal decomposition network of the multi-channel physiological signals is constructed to extract the fetal heart rate signal sequence and uterine contraction signal sequence of the multi-channel physiological signals. Extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence, wherein the time-domain feature set includes baseline heart rate, variable baseline, and sine wave morphology, and the frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. Extract the time-varying nonlinear feature set of the uterine contraction signal sequence, wherein the time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features. Construct the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set of the heart rate-uterine contraction joint spatiotemporal feature map to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity. Based on the heart rate-uterine contraction coupling relationship, the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman are calculated to determine the abnormal combination index of the pregnant woman. A monitoring report for the pregnant woman is generated based on the aforementioned abnormal joint index.

[0015] This invention utilizes intelligent algorithms to collect, process, and jointly analyze the physiological signals of pregnant women in real time. This not only improves the accuracy and efficiency of monitoring but also provides strong support for clinical decision-making, demonstrating significant user-friendly effects. Firstly, by acquiring physiological signals through multiple channels and combining them with a signal decomposition network, this invention can accurately extract fetal heart rate and uterine contraction signal sequences. Multi-channel fusion significantly enhances the signal's anti-interference capability and stability, providing a high-quality data foundation for subsequent analysis. Building upon this, this invention further extracts the temporal characteristics (such as baseline heart rate, baseline variation, and sine wave morphology) and frequency domain characteristics (such as frequency domain energy distribution and dominant frequency band power) of fetal heart rate, as well as the time-varying nonlinear characteristics of uterine contraction signals (such as contraction amplitude, time characteristics, and periodic characteristics), achieving a multi-dimensional characterization of fetal status and uterine contraction intensity. Secondly, by constructing a heart rate-uterine contraction joint spatiotemporal feature map, this invention deeply integrates heart rate and uterine contraction signals in the spatiotemporal dimension, intuitively presenting the coupling relationship between the two. This joint analysis not only reveals the dynamic response of fetal heart rate to uterine contractions but also effectively identifies abnormal coupling patterns, providing a scientific basis for early warning of risks such as fetal distress and excessively frequent uterine contractions. Based on this, the present invention further calculates the probability of fetal hypoxia and the probability of excessively frequent uterine contractions, and generates an abnormal joint index, achieving quantitative assessment and hierarchical management of risks. Finally, the present invention automatically generates a structured monitoring report based on the abnormal joint index, including risk trend charts, probability comparisons, risk levels, and clinical recommendations. This improves the standardization and readability of the report, facilitating cross-departmental collaboration and remote consultation. Furthermore, the present invention supports integration with electronic medical records and mobile monitoring platforms, achieving intelligent management throughout the entire process and laying the foundation for personalized medicine and precise intervention. In summary, the present invention, through multimodal signal fusion, intelligent feature extraction, and joint risk modeling, significantly improves the accuracy, real-time performance, and user-friendliness of fetal monitoring, bringing revolutionary technological advancements to obstetric clinical practice and providing more reliable protection for maternal and infant safety. Therefore, it enhances the accuracy and comprehensiveness of dual monitoring of fetal heart rate and uterine contractions. Attached Figure Description

[0016] Figure 1 A functional block diagram of a fetal heart rate and uterine contraction dual monitoring system based on artificial intelligence provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating an artificial intelligence-based method for dual monitoring of fetal heart rate and uterine contractions, provided in an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0019] In practice, the server-side equipment deployed in an AI-based fetal heart rate and uterine contraction dual monitoring system may consist of one or more devices. This AI-based fetal heart rate and uterine contraction dual monitoring system can be implemented as a business instance, a virtual machine, or hardware devices. For example, this AI-based fetal heart rate and uterine contraction dual monitoring system can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this AI-based fetal heart rate and uterine contraction dual monitoring system can be understood as software deployed on a cloud node, used to provide AI-based fetal heart rate and uterine contraction dual monitoring services to various user terminals. Alternatively, this AI-based fetal heart rate and uterine contraction dual monitoring system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing various user terminals. Alternatively, this AI-based fetal heart rate and uterine contraction dual monitoring system can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide AI-based fetal heart rate and uterine contraction dual monitoring services to various user terminals.

[0020] In terms of implementation, the AI-based fetal heart rate and uterine contraction dual monitoring system and the user terminal are mutually compatible. That is, if the AI-based fetal heart rate and uterine contraction dual monitoring system is implemented as an application installed on a cloud service platform, then the user terminal is a client that establishes a communication connection with the application; or if the AI-based fetal heart rate and uterine contraction dual monitoring system is implemented as a website, then the user terminal is implemented as a webpage; or if the AI-based fetal heart rate and uterine contraction dual monitoring system is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0021] Reference Figure 1 The diagram shown is a functional block diagram of a fetal heart rate and uterine contraction dual monitoring system based on artificial intelligence provided in an embodiment of the present invention.

[0022] The artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system 100 described in this invention can be installed on a cloud server. In terms of implementation, it can be used as one or more service devices, or as a batch of applications installed on the cloud (e.g., a server cluster, a server for dual monitoring of fetal heart rate and uterine contractions based on artificial intelligence), or it can be developed into a website. Depending on the functions implemented, the artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system 100 includes a signal decoupling module 101, a signal feature extraction module 102, a coupling relationship analysis module 103, an abnormal joint index analysis module 104, and a monitoring report generation module 105.

[0023] In this embodiment of the invention, in the tracking of fetal heart rate and uterine contractions based on artificial intelligence, each of the above-mentioned modules can be implemented independently and called upon other modules. Here, "called upon" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system provided in this embodiment of the invention, the applicability of the artificial intelligence-based fetal heart rate and uterine contraction dual monitoring architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system. In practical applications, the above-mentioned modules can be set in the same device or different devices, or in virtual devices, such as service instances in a cloud server.

[0024] The following describes, with reference to specific embodiments, each component and specific workflow of the artificial intelligence-based fetal heart rate and uterine contraction dual monitoring system.

[0025] The signal decoupling module 101 is used to collect multi-channel physiological signals from pregnant women and construct a signal decomposition network for the multi-channel physiological signals to extract the fetal heart rate signal sequence and uterine contraction signal sequence from the multi-channel physiological signals.

[0026] This invention, through the acquisition of multi-channel physiological signals from pregnant women, allows for mutual verification and supplementation, effectively suppressing noise such as maternal movement, electromyographic interference, and baseline drift, thereby improving the signal-to-noise ratio and laying the foundation for subsequent multimodal feature extraction and fusion analysis. The multi-channel physiological signals refer to various physiological signals from the pregnant woman's abdomen and related areas, such as fetal electrocardiogram signals, maternal electrocardiogram signals, and uterine contraction pressure signals, acquired simultaneously through multiple sensors or acquisition channels.

[0027] Optionally, the pregnant woman's multi-channel physiological signals can be synchronously acquired using a wearable sensor array.

[0028] This invention, through the construction of a signal decomposition network for the multi-channel physiological signals, can effectively separate fetal heart rate signals from uterine contraction signals, significantly reducing the impact of noise from maternal movement, electromyography, and other sources, thus improving signal quality. The signal decomposition network refers to a neural network model that accurately separates and extracts the target signal from mixed multi-channel physiological signals.

[0029] As an embodiment of the present invention, constructing the signal decomposition network of the multi-channel physiological signal includes: Construct the signal decomposition network architecture for the multi-channel physiological signals; The signal decomposition network architecture is constructed with a multi-level processing layer, which includes: a signal preprocessing layer, a spatiotemporal feature extraction layer, a blind source separation simulation layer, and a signal reconstruction layer. Determine the communication protocols and signal processing logic between the multi-level processing layers; Based on the communication protocol and the signal processing logic, the multi-level processing layer and the signal decomposition network architecture are integrated to obtain the signal decomposition network.

[0030] The signal decomposition network architecture refers to the overall framework structure used to process multi-channel physiological signals. The multi-level processing layer refers to the signal processing architecture within the signal decomposition network, constructed according to the signal processing logic sequence and composed of multiple functional levels. The communication protocol refers to the rules and mechanisms for data transmission, format conversion, and synchronization control between multi-level processing layers. The signal processing logic refers to the data processing rules, algorithm flow, and decision-making mechanisms within each processing layer and between layers in the signal decomposition network. The signal preprocessing layer refers to the module that performs preliminary processing on the original multi-channel physiological signals to eliminate noise, enhance signal quality, and standardize data formats. The spatiotemporal feature extraction layer refers to the module that extracts local temporal and spatial features from the preprocessed physiological signals. The blind source separation simulation layer refers to the module that simulates the blind source separation process using a neural network. The signal reconstruction layer refers to the module that further processes and optimizes the decoupled signal. The reconstructed signal refers to the output signal optimized by the signal reconstruction layer.

[0031] Optionally, the signal preprocessing layer of the multi-channel physiological signal can be constructed by selecting a filtering algorithm, such as a bandpass filtering algorithm or a wavelet denoising algorithm.

[0032] Optionally, the spatiotemporal feature extraction layer of the preprocessed physiological signal can be determined by a convolutional neural network.

[0033] This invention, through the extraction of fetal heart rate and uterine contraction signal sequences from the multi-channel physiological signals, comprehensively reflects the dynamic changes in fetal heart rate, capturing complex changes that are difficult to identify using traditional linear features, thus laying a solid data foundation for subsequent analysis. The fetal heart rate signal sequence refers to a continuous signal sequence extracted and separated from the pregnant woman's multi-channel physiological signals, reflecting the changes in fetal cardiac activity over time. The uterine contraction signal sequence refers to a continuous signal sequence extracted and separated from the pregnant woman's multi-channel physiological signals, reflecting the changes in uterine muscle contraction activity over time.

[0034] Optionally, the extraction of the fetal heart rate signal sequence and uterine contraction signal sequence from the multi-channel physiological signals includes: The multi-channel physiological signals are preprocessed through the signal preprocessing layer in the multi-channel physiological signal decomposition network to obtain preprocessed physiological signals. The local spatiotemporal features of the preprocessed physiological signal are extracted through the spatiotemporal feature extraction layer in the signal decomposition network. Based on the local spatiotemporal characteristics, the linear transformation matrix of the preprocessed physiological signal is determined through the blind source separation simulation layer in the signal decomposition network; Based on the linear transformation matrix, the decoupled signal of the preprocessed physiological signal is output; The decoupled signal is reconstructed through the signal reconstruction layer in the signal decomposition network to obtain the reconstructed signal; Extract the fetal heart rate signal sequence and uterine contraction signal sequence from the reconstructed signal.

[0035] The preprocessed physiological signal refers to the multi-channel physiological signal processed by the signal preprocessing layer. The spatiotemporal feature extraction layer is a module that extracts local temporal and spatial features from the preprocessed physiological signal. The local spatiotemporal features are feature vectors extracted from the preprocessed physiological signal that reflect the signal's change pattern within a local time window and the spatial correlation of multiple channels. The linear transformation matrix is ​​the demixing matrix learned by the blind source separation simulation layer, used to linearly map the mixed signal into independent source signals. The decoupled signal refers to the independent components separated from the mixed signal by the blind source separation simulation layer, such as fetal heart rate signals and uterine contraction signals. The reconstructed signal refers to the output signal optimized by the signal reconstruction layer.

[0036] Optionally, the decoupled signal can be obtained by blind source separation methods, such as independent component analysis, principal component analysis, etc.

[0037] The signal feature extraction module 102 is used to extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence. The time-domain feature set includes baseline heart rate, variable baseline, and sine wave morphology. The frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. The module also extracts the time-varying nonlinear feature set of the uterine contraction signal sequence. The time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features.

[0038] This invention, through extracting the time-domain and frequency-domain feature sets of the fetal heart rate signal sequence, transforms the raw, complex, and high-dimensional fetal heart rate signal into a structured, low-dimensional, and interpretable feature vector. This process is essentially signal representation learning, aiming to reduce dimensionality, denoise, and extract key information, providing high-quality input for subsequent tasks such as decision support, pattern recognition, and anomaly detection. The time-domain feature set refers to a set of quantitative indicators extracted directly from the time series of the raw heart rate signal, reflecting the signal's changing patterns and morphological characteristics over time. The baseline heart rate refers to the average level of the fetal heart rate over a relatively stable period without significant acceleration or deceleration. The variable baseline refers to the fluctuation amplitude of the baseline heart rate over a short period, reflecting the fetal autonomic nervous system's ability to regulate heart rate. The sine wave morphology refers to the smooth, regular, sinusoidal fluctuations of the fetal heart rate signal. The frequency-domain feature set refers to quantitative indicators extracted through spectral analysis of the signal, reflecting the energy distribution and dominant frequency characteristics of the signal at different frequency components. The frequency-domain energy distribution refers to the energy distribution of the signal across different frequency bands. The dominant frequency band power refers to the frequency band with the most concentrated energy in the power spectrum, and its power value reflects the dominant position of the frequency band in the signal.

[0039] As an embodiment of the present invention, the extraction of the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence includes: Determine a fixed window length and a fixed sampling frequency for the fetal heart rate signal sequence; Based on the fixed window length and the fixed sampling frequency, the fetal heart rate signal sequence is segmented to obtain segmented fetal heart rate signals; Analyze the sine wave morphology of the segmented fetal heart rate signal; Calculate the average heart rate of each segment in the segmented fetal heart rate signal; Based on the average heart rate, the segmented fetal heart rate signal is divided into an abnormal heart rate signal segment and a normal heart rate signal segment; Based on the normal heart rate signal segment, determine the baseline heart rate of the fetal heart rate signal sequence; Based on the abnormal heart rate signal segments, determine the baseline variation of the fetal heart rate signal sequence; The time-domain feature set of the fetal heart rate signal sequence is determined based on the baseline heart rate, the variable baseline, and the sine wave morphology. The fetal heart rate signal sequence is frequency domain transformed to obtain the fetal heart rate signal frequency domain sequence; Calculate the power spectral density of the fetal heart rate signal frequency domain sequence to determine the frequency domain energy distribution of the fetal heart rate signal frequency domain sequence; Identify the dominant frequency band of the frequency domain energy distribution and calculate the dominant frequency band power of the dominant frequency band; The frequency domain feature set of the fetal heart rate signal sequence is determined based on the main frequency band power and the frequency domain energy distribution.

[0040] The fixed window length refers to the fixed time length used to segment the continuous fetal heart rate signal during signal segmentation. The fixed sampling frequency refers to the number of times the fetal heart rate signal is sampled per unit time. The segmented fetal heart rate signal refers to multiple sub-signal segments obtained by dividing the original fetal heart rate signal according to the fixed window length. The average heart rate refers to the arithmetic mean of all sampling points of the fetal heart rate signal within the fixed window length. The abnormal heart rate signal segment refers to the segmented signal with an average heart rate exceeding the normal range (typically 110-160 bpm). The normal heart rate signal segment refers to the segmented signal with an average heart rate within the normal range (110-160 bpm). The fetal heart rate signal frequency domain sequence refers to the spectral representation obtained by frequency domain transformation of the segmented fetal heart rate signal. The power spectral density refers to the power distribution of each frequency component of the signal in the frequency domain. The dominant frequency band refers to the frequency band range with the highest energy in the power spectral density.

[0041] Optionally, the sinusoidal morphology of the segmented fetal heart rate signal can be analyzed using time-domain statistical methods, such as the mean method and the variance method.

[0042] Optionally, the average heart rate of each segment of the segmented fetal heart rate signal can be calculated using the arithmetic mean method.

[0043] Optionally, the frequency domain sequence of the fetal heart rate signal can be obtained by frequency domain transformation methods, such as Fourier transform, wavelet transform, etc.

[0044] This invention, through extracting a time-varying nonlinear feature set from the uterine contraction signal sequence, can comprehensively capture the dynamic changes of uterine contractions, enhance noise resistance, and thus improve the accuracy of fetal distress detection. The time-varying nonlinear feature set refers to a set of quantitative indicators extracted from the uterine contraction signal sequence that reflects the dynamic changes of the signal over time and possesses nonlinear characteristics. The contraction amplitude feature describes the intensity changes of the uterine contraction signal over time. The contraction time feature describes the duration and temporal evolution of the contraction event over time. The contraction cycle feature describes the periodicity of the contraction signal, reflecting the rhythm and coordination of the contractions.

[0045] As an embodiment of the present invention, the extraction of the time-varying nonlinear feature set of the uterine contraction signal sequence includes: Identify the contraction start point, contraction peak point, and contraction end point of the contraction signal sequence; Based on the contraction start point, the contraction peak point, and the contraction end point, the peak amplitude of the contraction signal sequence is determined; Based on the contraction start point and the contraction end point, calculate the contraction duration and contraction interval of the contraction signal sequence; Based on the peak point of the uterine contractions, determine the rise time and fall time of the uterine contraction signal sequence; Calculate the average amplitude of the contraction signal sequence based on the duration of the contractions;

[0046] Calculate the amplitude change rate of the uterine contraction signal sequence based on the rise time and the fall time; Based on the duration of the contractions, the contraction frequency of the contraction signal sequence is analyzed; Calculate the standard deviation of the contraction intervals to analyze the periodic variability of the contraction signal sequence; By integrating the peak amplitude, the average amplitude, the amplitude change rate, the contraction duration, the contraction interval, the rise time, the fall time, the contraction frequency, and the periodic variability, a time-varying nonlinear feature set of the contraction signal sequence is obtained.

[0047] Wherein, the contraction initiation point refers to the point where the contraction signal begins to rise. The contraction peak point refers to the point where the contraction signal reaches its maximum amplitude in a single contraction. The contraction end point refers to the point where the contraction signal returns to the baseline. The peak amplitude refers to the maximum value of the signal in a single contraction. The average amplitude refers to the average signal value from the initiation point to the end point of a single contraction. The amplitude change rate refers to the average rate of change of the signal from the initiation point to the peak point. The contraction duration refers to the length of time from the initiation point to the end point. The contraction interval refers to the length of time from the end point of the previous contraction to the initiation point of the next contraction. The rise time refers to the length of time from the initiation point to the peak point. The fall time refers to the length of time from the peak point to the end point. The contraction frequency refers to the number of contractions per unit time. The periodicity variability refers to the standard deviation of the contraction interval, reflecting the regularity of contractions.

[0048] Optionally, the contraction start point, contraction peak point, and contraction end point of the contraction signal sequence can be identified by the signal amplitude threshold method. For example, if the signal exceeds the baseline by 10%, the maximum amplitude is the contraction start point; if it is higher than the baseline by 20%, the maximum amplitude is the contraction peak point; and if it is lower than the baseline by 10%, the maximum amplitude is the contraction end point.

[0049] As another implementation, the rate of change of amplitude can be calculated using the following formula:

[0050]

[0051] in, This represents the rate of change of the upward amplitude in the rate of change of amplitude. This represents the rate of change of the decreasing amplitude in the rate of change of amplitude. This indicates the amplitude of the peak contraction. This indicates the amplitude at the onset of uterine contractions. This indicates the amplitude at the end of the contraction. Indicates the rise time. Indicates the descent time.

[0052] The coupling relationship analysis module 103 is used to construct the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set, so as to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity.

[0053] This invention, through the construction of a heart rate-uterine contraction joint spatiotemporal feature map comprising the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set, can fuse and align the time-domain, frequency-domain, and time-varying nonlinear features in the spatiotemporal dimension. This visually presents the dynamic correlation between fetal heart rate status and uterine contraction intensity, providing data support for subsequent analysis. Specifically, the heart rate-uterine contraction joint spatiotemporal feature map refers to a visualization or feature representation map used for the joint analysis of fetal heart rate and uterine contractions through multimodal, spatiotemporal fusion.

[0054] As an embodiment of the present invention, the construction of the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set includes: The time-varying nonlinear characteristics, the time-domain feature set, and the time axis and feature axis of the frequency-domain feature set are uniformly defined. Based on the time axis and the feature axis, construct a two-dimensional feature matrix of the time-varying nonlinear feature sequence, the time-domain feature sequence, and the frequency-domain feature sequence; The two-dimensional feature matrix is ​​standardized to obtain a standardized feature matrix; Analyze the feature relationships in the standardized feature matrix; The standardized feature matrix is ​​visualized to obtain a feature heatmap. The feature relationships are integrated into the feature heatmap to obtain the heart rate-uterine contraction combined spatiotemporal feature map.

[0055] The time axis refers to the time reference axis used to align different feature sequences. The feature axis refers to the dimension axis used to construct a two-dimensional matrix, composed of all features. The two-dimensional feature matrix is ​​a matrix where each row represents a time point and each column represents an feature value. The standardized feature matrix is ​​the matrix obtained by standardizing the two-dimensional feature matrix. The feature relationships refer to the correlation, mutual information, clustering relationships, linear relationships, and non-linear dependencies between features. The feature heatmap is a graphical representation of the standardized feature matrix, where color intensity represents the magnitude of feature values, the horizontal axis represents time, and the vertical axis represents feature type.

[0056] Optionally, the standardized feature matrix can be obtained by a standardization algorithm, such as Z-score standardization, Min-Max normalization, etc.

[0057] Optionally, the feature relationships in the standardized feature matrix can be obtained through relationship analysis, such as correlation matrix, cluster analysis, principal component analysis, etc.

[0058] This invention, through analysis of the fetal heart rate status and uterine contraction intensity of the pregnant woman, can deeply analyze the change pattern of fetal heart rate under uterine contraction pressure, reveal the coupling relationship between the two, and significantly improve the early identification ability of abnormal conditions. The fetal heart rate status refers to the comprehensive assessment result of the fetal cardiovascular system function, autonomic nervous system regulation ability, and overall health status. The uterine contraction intensity refers to the mechanical strength generated by the uterine muscles during contraction.

[0059] As an embodiment of the present invention, the analysis of the fetal heart rate status and uterine contraction intensity of the pregnant woman includes: Construct a feature map attention analysis network for the spatiotemporal feature map of the pregnant woman's heart rate-uterine contraction combination; Identify the feature nodes of the heart rate-uterine contraction combined spatiotemporal feature map; Based on the feature map attention analysis network, the importance score between the feature nodes is calculated using the following formula:

[0060] in, Represents feature nodes To feature nodes Importance score Indicated by An exponential function with base 0. This represents the leaky ReLU activation function in the feature map attention analysis network. This represents the learnable attention vector in the feature map attention analysis network. This represents the learnable weight matrix in the feature map attention analysis network. This represents the transpose of the learnable weight matrix. Represents feature nodes The original feature vector in the corresponding heart rate-uterine contraction spatiotemporal feature map, Represents feature nodes The original feature vector in the corresponding heart rate-uterine contraction spatiotemporal feature map. Represents feature nodes The original feature vector in the corresponding heart rate-uterine contraction spatiotemporal feature map, Feature nodes in a feature map attention analysis network and feature nodes The cross-attention value of the coupling features between them. Feature nodes in a feature map attention analysis network and feature nodes The cross-attention value of the coupling features between them. Represents feature nodes The field; Based on the importance score, the fetal health index and uterine contraction coordination of the pregnant woman are calculated. Based on the fetal health index and the uterine contraction coordination, the fetal heart rate status and uterine contraction intensity of the pregnant woman are determined.

[0061] The feature map attention analysis network refers to a neural network that learns the relationships between nodes from the heart rate-contraction joint spatiotemporal feature map based on a graph attention mechanism and calculates the importance scores between nodes. The feature node refers to the feature vector corresponding to each time point in the heart rate-contraction joint spatiotemporal feature map. The importance score represents the degree of influence of one feature node on another. The leaky ReLU activation function is an activation function that ensures that when the input is negative, the output is not strictly zero, but a small, non-zero negative value. The learnable attention vector is a trainable vector used to weight the concatenated feature vectors to generate attention scores. The learnable weight matrix is ​​a learnable matrix used to perform linear transformations on the original feature vectors. The cross-attention value is the embedded representation of the coupling features between two feature nodes. The neighborhood refers to the set of all nodes connected to the feature node.

[0062] Optionally, the feature map attention analysis network corresponding to the pregnant woman's heart rate-uterine contraction joint spatiotemporal feature map can be constructed using a spatiotemporal graph convolutional network and a multi-head self-attention mechanism.

[0063] Optionally, the fetal health index and uterine contraction coordination of the pregnant woman can be calculated using deep learning models, such as LSTM and Transformer.

[0064] This invention, through determining the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity, can identify fetal heart rate abnormalities caused by excessively strong or weak uterine contractions, thereby improving the detection accuracy of abnormal states such as fetal distress and hypoxia. The heart rate-uterine contraction coupling relationship refers to the dynamic, time-varying, and nonlinear interaction and dependence between the fetal heart rate and maternal uterine contraction signals.

[0065] As an embodiment of the present invention, determining the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity includes: The fetal heart rate status and the uterine contraction intensity were jointly constructed as heart rate-uterine contraction time-series curves. Identify the contraction time intervals of the heart rate-contraction timing curve; Calculate the ratio of the heart rate deceleration area to the contraction time interval of the heart rate-contraction time sequence curve; Identify the contraction rhythm frequency band of the heart rate-contraction timing curve; Calculate the correlation coefficient between the heart rate-contraction time series curve and the contraction-time series curve in the contraction rhythm frequency band; Based on the heart rate deceleration area ratio and the correlation coefficient, the coupling risk index of the heart rate-uterine contraction time series curve is calculated to determine the heart rate-uterine contraction coupling relationship of the heart rate-uterine contraction time series curve.

[0066] The heart rate-contraction time series curve refers to a bivariate time series curve formed by aligning and plotting fetal heart rate signals and uterine contraction pressure signals along the same time axis. The contraction time interval refers to the time range from the start to the end of a single contraction. The heart rate deceleration area ratio refers to the ratio of the area of ​​the fetal heart rate deceleration curve to the area of ​​the contraction intensity curve within a single contraction time interval. The contraction rhythm frequency band refers to the frequency range in the contraction signal that represents its dominant rhythm. The correlation coefficient refers to the amplitude-squared coherence value between the fetal heart rate signal and the contraction signal within the contraction rhythm frequency band. The coupling risk index is an indicator used to quantify the degree of abnormality in the fetal heart rate-uterine contraction coupling relationship.

[0067] Optionally, the fetal heart rate status and the heart rate-uterine contraction time-series curve of the uterine contraction intensity can be constructed using deep learning techniques, such as convolutional neural networks, long short-term memory networks, etc.

[0068] Optionally, the ratio of the heart rate deceleration area in the contraction time interval of the heart rate-contraction time sequence curve can be calculated by integration.

[0069] Optionally, the correlation coefficient can be calculated using wavelet packet transform correlation analysis or Hilbert-Huang transform correlation.

[0070] The abnormal joint index analysis module 104 is used to calculate the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman based on the heart rate-uterine contraction coupling relationship, so as to determine the abnormal joint index of the pregnant woman.

[0071] This invention, through calculations of the probability of fetal hypoxia and the probability of excessive uterine contractions based on the heart rate-uterine contraction coupling relationship, can quantitatively assess the risk of fetal hypoxia, accurately evaluate uterine contraction function, and improve the early identification rate of fetal distress. The probability of fetal hypoxia refers to the quantitative probability value of fetal hypoxia risk during the current contraction cycle, calculated through multi-parameter modeling based on heart rate-uterine contraction coupling relationship analysis. The probability of excessive uterine contractions refers to the likelihood that the current frequency of uterine contractions exceeds the normal physiological range, calculated through statistical models based on uterine contraction signal analysis.

[0072] As an embodiment of the present invention, the step of calculating the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman based on the heart rate-uterine contraction coupling relationship includes: Extract the coupling relationship features of the heart rate-uterine contraction coupling relationship; Based on the coupling relationship characteristics, fit the fetal hypoxia probability model and the uterine contraction frequency probability model of the pregnant woman; Based on the fetal hypoxia probability model, calculate the probability of fetal hypoxia in the pregnant woman. The probability of excessive uterine contractions in the pregnant woman is calculated based on the aforementioned probability model for excessive uterine contractions.

[0073] The coupling relationship feature refers to a quantitative indicator used to quantify the interaction between fetal heart rate and uterine contractions. The fetal hypoxia probability model is a mathematical model constructed based on the coupling relationship feature, used to output the probability of fetal hypoxia. The excessive uterine contraction probability model is a mathematical model constructed based on the coupling relationship feature, used to assess whether the pregnant woman is currently experiencing excessive uterine contractions.

[0074] In another implementation, the fetal hypoxia probability model is expressed by the following formula:

[0075] in, This indicates the probability of fetal hypoxia. Indicated by An exponential function with base 0. This represents the bias term in the hypoxia model. The weights representing the ratio of the characteristic centrality deceleration area to the coupling relationship are... This represents the ratio of heart rate deceleration area in the coupling relationship characteristics. This represents the weight of the correlation coefficient in the coupling relationship features. This represents the correlation coefficient in the characteristics of the coupling relationship.

[0076] For example, if the heart rate deceleration area ratio is 0.5 with a weight of 2, the correlation coefficient is 0.8 with a weight of 1.5, and the hypoxia model bias term is 1, then the probability of fetal hypoxia is... .

[0077] As another implementation, the probability model for excessive uterine contractions is expressed by the following formula:

[0078] in, This indicates the probability of excessively frequent uterine contractions. Indicated by An exponential function with base 0. This represents the bias term in the uterine contraction model. This represents the weight of the uterine contraction frequency in the coupling relationship features. This represents the frequency of uterine contractions in the coupling relationship characteristics. This represents the weight of the average contraction interval in the coupling relationship characteristics. This represents the average interval between uterine contractions in the coupling relationship characteristics.

[0079] For example, if the contraction frequency is 0.7 with a weight of 1.8, the average contraction interval is 0.6 with a weight of 1.2, and the contraction model bias term is 0.5, then the probability of excessively frequent contractions is... .

[0080] This invention, by determining the abnormal combined index of the pregnant woman, can achieve higher sensitivity and specificity, enabling early warning of potential risks. The abnormal combined index is a quantitative indicator used to comprehensively assess the degree of abnormality in multimodal physiological signals in the pregnant woman's abdomen.

[0081] As an embodiment of the present invention, determining the abnormal combined index of the pregnant woman includes: The probabilities of fetal hypoxia and uterine contraction frequency corresponding to the pregnant woman are nonlinearly transformed to obtain the transformed probabilities of fetal hypoxia and uterine contraction frequency. Determine the hypoxia probability regression coefficient for the change in fetal hypoxia probability and the contraction probability regression coefficient for the change in uterine contraction frequency probability; Calculate the interaction coefficient between the changed probability of fetal hypoxia and the changed probability of excessive uterine contractions; The abnormal combination index of the pregnant woman is calculated based on the regression coefficient of the uterine contraction probability, the regression coefficient of the hypoxia probability, and the interaction coefficient.

[0082] Wherein, the transformed fetal hypoxia probability refers to the value after a nonlinear transformation of the fetal hypoxia probability. The transformed excessive uterine contraction probability refers to the value after a nonlinear transformation of the excessive uterine contraction probability. The hypoxia probability regression coefficient refers to the contribution weight of the transformed fetal hypoxia probability to the abnormal joint index. The uterine contraction probability regression coefficient refers to the contribution weight of the transformed excessive uterine contraction probability to the abnormal joint index. The interaction coefficient refers to the strength of the influence of the interaction between the transformed fetal hypoxia probability and the transformed excessive uterine contraction probability on the abnormal joint index.

[0083] Optionally, the transformed fetal hypoxia probability can be obtained through nonlinear transformation, such as logarithmic transformation, sigmoid transformation, exponential transformation, etc.

[0084] Optionally, the hypoxia probability regression coefficient can be obtained by regression analysis techniques, such as logistic regression, linear regression, or generalized linear models.

[0085] Optionally, the interaction coefficient between the altered fetal hypoxia probability and the altered uterine contraction frequency probability can be calculated using causal inference techniques, such as DoWhy, CausalML, etc.

[0086] The monitoring report generation module 105 is used to generate a monitoring report for the pregnant woman based on the abnormal joint index.

[0087] This invention, through the generation of a monitoring report for the pregnant woman based on the abnormal joint index, can integrate multimodal physiological signals to generate a comprehensive risk index, providing quantitative evidence, assisting doctors in quickly assessing the intrauterine status of the fetus, and realizing an automatic alarm and graded reminder mechanism.

[0088] This invention utilizes intelligent algorithms to collect, process, and jointly analyze the physiological signals of pregnant women in real time. This not only improves the accuracy and efficiency of monitoring but also provides strong support for clinical decision-making, demonstrating significant user-friendly effects. Firstly, by acquiring physiological signals through multiple channels and combining them with a signal decomposition network, this invention can accurately extract fetal heart rate and uterine contraction signal sequences. Multi-channel fusion significantly enhances the signal's anti-interference capability and stability, providing a high-quality data foundation for subsequent analysis. Building upon this, this invention further extracts the temporal characteristics (such as baseline heart rate, baseline variation, and sine wave morphology) and frequency domain characteristics (such as frequency domain energy distribution and dominant frequency band power) of fetal heart rate, as well as the time-varying nonlinear characteristics of uterine contraction signals (such as contraction amplitude, time characteristics, and periodic characteristics), achieving a multi-dimensional characterization of fetal status and uterine contraction intensity. Secondly, by constructing a heart rate-uterine contraction joint spatiotemporal feature map, this invention deeply integrates heart rate and uterine contraction signals in the spatiotemporal dimension, intuitively presenting the coupling relationship between the two. This joint analysis not only reveals the dynamic response of fetal heart rate to uterine contractions but also effectively identifies abnormal coupling patterns, providing a scientific basis for early warning of risks such as fetal distress and excessively frequent uterine contractions. Based on this, the present invention further calculates the probability of fetal hypoxia and the probability of excessively frequent uterine contractions, and generates an abnormal joint index, achieving quantitative assessment and hierarchical management of risks. Finally, the present invention automatically generates a structured monitoring report based on the abnormal joint index, including risk trend charts, probability comparisons, risk levels, and clinical recommendations. This improves the standardization and readability of the report, facilitating cross-departmental collaboration and remote consultation. Furthermore, the present invention supports integration with electronic medical records and mobile monitoring platforms, achieving intelligent management throughout the entire process and laying the foundation for personalized medicine and precise intervention. In summary, the present invention, through multimodal signal fusion, intelligent feature extraction, and joint risk modeling, significantly improves the accuracy, real-time performance, and user-friendliness of fetal monitoring, bringing revolutionary technological advancements to obstetric clinical practice and providing more reliable protection for maternal and infant safety. Therefore, the present invention can improve the accuracy and comprehensiveness of dual monitoring of fetal heart rate and uterine contractions.

[0089] like Figure 2 The diagram shown is a flowchart illustrating an artificial intelligence-based dual monitoring method for fetal heart rate and uterine contractions according to an embodiment of the present invention. In this embodiment, the artificial intelligence-based dual monitoring method for fetal heart rate and uterine contractions includes: Multi-channel physiological signals of pregnant women are collected, and a signal decomposition network of the multi-channel physiological signals is constructed to extract the fetal heart rate signal sequence and uterine contraction signal sequence of the multi-channel physiological signals. Extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence, wherein the time-domain feature set includes baseline heart rate, variable baseline, and sine wave morphology, and the frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. Extract the time-varying nonlinear feature set of the uterine contraction signal sequence, wherein the time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features. Construct the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set of the heart rate-uterine contraction joint spatiotemporal feature map to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity. Based on the heart rate-uterine contraction coupling relationship, the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman are calculated to determine the abnormal combination index of the pregnant woman. A monitoring report for the pregnant woman is generated based on the aforementioned abnormal joint index.

[0090] In the several embodiments provided by this invention, it should be understood that the provided systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0091] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0092] Finally, it should be noted that in the above embodiments, each embodiment can be combined with each other or independent. Deleting any one of them will not affect the technical implementation of other embodiments. The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fetal heart rate and uterine contraction dual monitoring system based on artificial intelligence, characterized in that, The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions includes: The signal decoupling module is used to collect multi-channel physiological signals from pregnant women and construct a signal decomposition network for the multi-channel physiological signals to extract the fetal heart rate signal sequence and uterine contraction signal sequence from the multi-channel physiological signals. The signal feature extraction module is used to extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence. The time-domain feature set includes baseline heart rate, baseline variation, and sine wave morphology. The frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. The module also extracts the time-varying nonlinear feature set of the uterine contraction signal sequence. The time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features. The coupling relationship analysis module is used to construct the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set, so as to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity. The abnormal joint index analysis module is used to calculate the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman based on the heart rate-uterine contraction coupling relationship, so as to determine the abnormal joint index of the pregnant woman. The monitoring report generation module is used to generate a monitoring report for the pregnant woman based on the abnormal joint index.

2. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The construction of the heart rate-uterine contraction joint spatiotemporal feature map of the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set includes: The time-varying nonlinear characteristics, the time-domain feature set, and the time axis and feature axis of the frequency-domain feature set are uniformly defined. Based on the time axis and the feature axis, construct a two-dimensional feature matrix of the time-varying nonlinear feature sequence, the time-domain feature sequence, and the frequency-domain feature sequence; The two-dimensional feature matrix is ​​standardized to obtain a standardized feature matrix; Analyze the feature relationships in the standardized feature matrix; The standardized feature matrix is ​​visualized to obtain a feature heatmap. The feature relationships are integrated into the feature heatmap to obtain the heart rate-uterine contraction combined spatiotemporal feature map.

3. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The analysis of the fetal heart rate and uterine contraction intensity of the pregnant woman includes: Construct a feature map attention analysis network for the spatiotemporal feature map of the pregnant woman's heart rate-uterine contraction combination; Identify the feature nodes of the heart rate-uterine contraction combined spatiotemporal feature map; Based on the feature map attention analysis network, calculate the importance scores between the feature nodes; Based on the importance score, the fetal health index and uterine contraction coordination of the pregnant woman are calculated. Based on the fetal health index and the uterine contraction coordination, the fetal heart rate status and uterine contraction intensity of the pregnant woman are determined.

4. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The construction of the signal decomposition network for the multi-channel physiological signal includes: Construct the signal decomposition network architecture for the multi-channel physiological signals; The signal decomposition network architecture is constructed with a multi-level processing layer, which includes: a signal preprocessing layer, a spatiotemporal feature extraction layer, a blind source separation simulation layer, and a signal reconstruction layer. Determine the communication protocols and signal processing logic between the multi-level processing layers; Based on the communication protocol and the signal processing logic, the multi-level processing layer and the signal decomposition network architecture are integrated to obtain the signal decomposition network.

5. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The extraction of the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence includes: Determine a fixed window length and a fixed sampling frequency for the fetal heart rate signal sequence; Based on the fixed window length and the fixed sampling frequency, the fetal heart rate signal sequence is segmented to obtain segmented fetal heart rate signals; Analyze the sine wave morphology of the segmented fetal heart rate signal; Calculate the average heart rate of each segment in the segmented fetal heart rate signal; Based on the average heart rate, the segmented fetal heart rate signal is divided into an abnormal heart rate signal segment and a normal heart rate signal segment; Based on the normal heart rate signal segment, determine the baseline heart rate of the fetal heart rate signal sequence; Based on the abnormal heart rate signal segments, determine the baseline variation of the fetal heart rate signal sequence; The time-domain feature set of the fetal heart rate signal sequence is determined based on the baseline heart rate, the variable baseline, and the sine wave morphology. The fetal heart rate signal sequence is frequency domain transformed to obtain the fetal heart rate signal frequency domain sequence; Calculate the power spectral density of the fetal heart rate signal frequency domain sequence to determine the frequency domain energy distribution of the fetal heart rate signal frequency domain sequence; Identify the dominant frequency band of the frequency domain energy distribution and calculate the dominant frequency band power of the dominant frequency band; The frequency domain feature set of the fetal heart rate signal sequence is determined based on the main frequency band power and the frequency domain energy distribution.

6. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The extraction of the time-varying nonlinear feature set of the uterine contraction signal sequence includes: Identify the contraction start point, contraction peak point, and contraction end point of the contraction signal sequence; Based on the contraction start point, the contraction peak point, and the contraction end point, the peak amplitude of the contraction signal sequence is determined; Based on the contraction start point and the contraction end point, calculate the contraction duration and contraction interval of the contraction signal sequence; Based on the peak point of the uterine contractions, determine the rise time and fall time of the uterine contraction signal sequence; Calculate the average amplitude of the contraction signal sequence based on the duration of the contractions; Calculate the amplitude change rate of the uterine contraction signal sequence based on the rise time and the fall time; Based on the duration of the contractions, the contraction frequency of the contraction signal sequence is analyzed; Calculate the standard deviation of the contraction intervals to analyze the periodic variability of the contraction signal sequence; By integrating the peak amplitude, the average amplitude, the amplitude change rate, the contraction duration, the contraction interval, the rise time, the fall time, the contraction frequency, and the periodic variability, a time-varying nonlinear feature set of the contraction signal sequence is obtained.

7. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The determination of the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity includes: The fetal heart rate status and the uterine contraction intensity were jointly constructed as heart rate-uterine contraction time-series curves. Identify the contraction time intervals of the heart rate-contraction timing curve; Calculate the ratio of the heart rate deceleration area to the contraction time interval of the heart rate-contraction time sequence curve; Identify the contraction rhythm frequency band of the heart rate-contraction timing curve; Calculate the correlation coefficient between the heart rate-contraction time series curve and the contraction-time series curve in the contraction rhythm frequency band; Based on the heart rate deceleration area ratio and the correlation coefficient, the coupling risk index of the heart rate-uterine contraction time series curve is calculated to determine the heart rate-uterine contraction coupling relationship of the heart rate-uterine contraction time series curve.

8. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, The step of calculating the probability of fetal hypoxia and the probability of excessive uterine contractions in the pregnant woman based on the heart rate-uterine contraction coupling relationship includes: Extract the coupling relationship features of the heart rate-uterine contraction coupling relationship; Based on the coupling relationship characteristics, fit the fetal hypoxia probability model and the uterine contraction frequency probability model of the pregnant woman; Based on the fetal hypoxia probability model, calculate the probability of fetal hypoxia in the pregnant woman. The probability of excessive uterine contractions in the pregnant woman is calculated based on the aforementioned probability model for excessive uterine contractions.

9. The artificial intelligence-based dual monitoring system for fetal heart rate and uterine contractions as described in claim 1, characterized in that, Determining the abnormal combined index of the pregnant woman includes: The probabilities of fetal hypoxia and uterine contraction frequency corresponding to the pregnant woman are nonlinearly transformed to obtain the transformed probabilities of fetal hypoxia and uterine contraction frequency. Determine the hypoxia probability regression coefficient for the change in fetal hypoxia probability and the contraction probability regression coefficient for the change in uterine contraction frequency probability; Calculate the interaction coefficient between the changed probability of fetal hypoxia and the changed probability of excessive uterine contractions; The abnormal combination index of the pregnant woman is calculated based on the regression coefficient of the uterine contraction probability, the regression coefficient of the hypoxia probability, and the interaction coefficient.

10. A method for dual monitoring of fetal heart rate and uterine contractions based on artificial intelligence, characterized in that, The method includes: Multi-channel physiological signals of pregnant women are collected, and a signal decomposition network of the multi-channel physiological signals is constructed to extract the fetal heart rate signal sequence and uterine contraction signal sequence of the multi-channel physiological signals. Extract the time-domain feature set and frequency-domain feature set of the fetal heart rate signal sequence, wherein the time-domain feature set includes baseline heart rate, variable baseline, and sine wave morphology, and the frequency-domain feature set includes frequency-domain energy distribution and main frequency band power. Extract the time-varying nonlinear feature set of the uterine contraction signal sequence, wherein the time-varying nonlinear feature set includes uterine contraction amplitude features, uterine contraction time features, and uterine contraction cycle features. Construct the time-varying nonlinear features, the time-domain feature set, and the frequency-domain feature set of the heart rate-uterine contraction joint spatiotemporal feature map to analyze the fetal heart rate status and uterine contraction intensity of the pregnant woman, and determine the heart rate-uterine contraction coupling relationship between the fetal heart rate status and the uterine contraction intensity. Based on the heart rate-uterine contraction coupling relationship, the probability of fetal hypoxia and the probability of excessive uterine contractions of the pregnant woman are calculated to determine the abnormal combination index of the pregnant woman. A monitoring report for the pregnant woman is generated based on the aforementioned abnormal joint index.

Images