Method and system for constructing fetal heart monitoring auxiliary interpretation model based on one-dimensional time sequence signal and double neural network architecture

Abstract

The invention relates to the technical field of medical care informatics, and discloses a method and a system for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time sequence signal and a double-neural network architecture. According to the method, a time-aligned double-signal sampling strategy is adopted, double-lead one-dimensional time sequence data of the fetal heart rate and uterine contraction are input in parallel, a double-branch neural network architecture is adopted, each branch independently processes a lead signal, and the two-branch neural network architecture is adopted to process the two-lead one-dimensional time sequence data of the fetal heart rate and uterine contraction, so that the two-lead one-dimensional time sequence data of the fetal heart rate and uterine contraction are obtained. Complementary information between different lead signals can be fully utilized, accurate recognition of fetus physiological status and pregnant woman fetus status abnormalities is achieved, and the recognition accuracy of a fetal heart monitoring auxiliary interpretation model on complex abnormal conditions such as fetal hypoxia and uterine contraction compression and time-dependent pathological states such as late deceleration and variation deceleration is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical care informatics, and in particular to a method and system for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture. Background Art

[0002] Fetal heart rate monitoring is a common tool for continuously monitoring fetal health during labor. It assesses the fetus's health in utero by monitoring real-time changes in fetal heart rate (FHR) and uterine contractions (UC). Fetal heart rate monitoring can help detect abnormalities such as fetal distress early, providing a basis for clinical intervention and reducing the incidence of adverse perinatal outcomes.

[0003] Traditional fetal heart rate monitoring relies primarily on subjective visual assessment by the obstetric medical team, but this approach has certain limitations. Due to the diverse variations in fetal heart rate monitoring graphs and the different standards for fetal heart rate monitoring diagnostic rules, there are often differences between and within observers, leading to inconsistent interpretations. The influence of this subjective factor may delay clinical decision-making and affect maternal and fetal outcomes. Artificial intelligence methods have also been introduced to analyze fetal heart rate monitoring data, but many existing methods rely solely on a single neural network or simple feature extraction methods, making it difficult to fully capture the complex features and temporal relationships in fetal heart rate monitoring data. In particular, they ignore the complex correlation between fetal heart rate (FHR) and uterine contractions (UC), which may lead to inaccurate assessments of the fetal physiological state and the inability to accurately identify subtle abnormal changes.

[0004] Therefore, there is an urgent need to build a new fetal heart monitoring auxiliary interpretation model to assist in achieving more accurate and efficient fetal heart monitoring. Summary of the Invention

[0005] The present invention provides a method and system for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture, which are used to solve the defect that the current artificial intelligence method used to analyze fetal heart monitoring data is not accurate enough.

[0006] The present invention provides a method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture, comprising: Obtain fetal heart monitoring data and corresponding monitoring result label data of a group of pregnant women, wherein the fetal heart monitoring data includes one-dimensional time series data of fetal heart rate and one-dimensional time series data of uterine contraction, and the monitoring result label data includes fetal physiological state labels and abnormal fetal state interval labels of pregnant women; Based on the fetal heart monitoring data of pregnant women and the corresponding monitoring result label data, a dual neural network architecture is used to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data, and a fetal heart monitoring auxiliary interpretation model is trained.

[0007] It should be noted that the one-dimensional time series data of fetal heart rate represents the sequence of the frequency of fetal heart beats changing over time, and the one-dimensional time series data of uterine contraction represents the sequence of the intensity and frequency of the contraction of the pregnant woman's uterine muscles changing over time.

[0008] In one embodiment, the fetal heart monitoring data is organized in the form of a two-dimensional array, wherein the first row of the array is specifically used to store the one-dimensional time series signal data of FHR, and the second row of the array is used to store the one-dimensional time series signal data of UC. Each column of the array represents a specific sampling data point, that is, the FHR value and UC value recorded at the same time point.

[0009] According to a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture provided by the present invention, the fetal physiological state includes any one of the following or any combination thereof: fetal hypoxia state, blood pH state, blood sample state, and blood carbon dioxide state, wherein the fetal hypoxia state is marked in a classified form as "yes" or "no", and the blood pH state, blood oxygen state, and blood carbon dioxide state are marked in segmented intervals and are divided into three categories: "good", "abnormal", and "severely abnormal".

[0010] According to the method for constructing a fetal heart rate monitoring assisted interpretation model based on one-dimensional time series signals and a dual neural network architecture, the present invention provides a method for constructing a fetal heart rate monitoring assisted interpretation model. The label for abnormal maternal fetal status is used to indicate the specific time series interval within a time series array where maternal and fetal abnormalities occur. Within the time series array, the start and end time points of the abnormality are precisely marked and represented as an abnormality index interval, such as "[start time, end time]," clearly indicating the specific time period during which the abnormality occurred. Furthermore, a textual description or code of the abnormality type can be added, such as fetal distress, bradycardia, tachycardia, uterine contractions, and other abnormalities.

[0011] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture can include the following steps after obtaining fetal heart monitoring data of a group of pregnant women and corresponding monitoring result label data: Preprocessing is performed on fetal heart monitoring data of a group of pregnant women and corresponding monitoring result label data. The preprocessing includes any one of the following or any combination thereof: linear interpolation, sample cropping, and normalization.

[0012] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The dual neural network architecture includes a classification and recognition neural network architecture and an abnormality monitoring and evaluation neural network architecture, wherein the classification and recognition neural network architecture includes a feature extraction module, a classification and recognition module, and a first embedding layer connecting the feature extraction module and the classification and recognition module. The feature extraction module includes a time series representation submodule (a one-dimensional convolutional neural network submodule), a global average pooling submodule, an encoder, and a projector. The classification and recognition module includes a fully connected layer, a batch normalization submodule, and a Softmax (normalized exponential function) activation function. The abnormality monitoring and evaluation neural network architecture includes a second embedding layer, an abnormality monitoring module based on a Transformer architecture, and a projection layer. The second embedding layer includes a mask embedding submodule and a position encoding submodule. The abnormality monitoring module based on the Transformer architecture includes a first encoder layer, a second encoder layer, and a third encoder layer. The first encoder layer, the second encoder layer, and the third encoder layer respectively include a multi-head attention layer and a multi-layer perceptron.

[0013] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The method uses the dual neural network architecture to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data of a group of pregnant women, and trains the fetal heart monitoring auxiliary interpretation model, including: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, feature extraction is performed through the feature extraction module of the classification and recognition neural network architecture in the dual neural network architecture to obtain comprehensive features; According to the comprehensive features, a comprehensive feature vector is obtained through the first embedding layer of the classification recognition neural network architecture in the dual neural network architecture; According to the comprehensive feature vector, the comprehensive feature vector is mapped to the output category through the classification recognition module of the classification recognition neural network architecture in the dual neural network architecture to realize the classification of the fetal physiological state.

[0014] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The method extracts features based on the fetal heart monitoring data of a group of pregnant women and the corresponding monitoring result label data through a feature extraction module of a classification and recognition neural network architecture in the dual neural network architecture to obtain comprehensive features, including: Based on the fetal heart monitoring data of pregnant women and the corresponding monitoring result label data, a time series representation submodule with feature extraction function is constructed. The different sizes and step sizes of the convolution kernel are used to capture the different time series relationships between the fetal heart monitoring data and the corresponding monitoring result label data, and multiple convolution features are obtained. According to the multiple convolved features, the encoder of the feature extraction module is used to encode them respectively to obtain multiple encoded features; According to the multiple encoded features, the encoded features are mapped to the classification space through the projector of the feature extraction module, and the multiple mapped features are spliced ​​in the channel dimension to obtain the comprehensive features.

[0015] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The method uses the dual neural network architecture to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data of a group of pregnant women, and trains the fetal heart monitoring auxiliary interpretation model, including: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to obtain the time series features and location information; According to the timing characteristics and location information, the abnormality monitoring module based on the Transformer architecture of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture calculates the attention weight and association difference to identify abnormal points in the timing characteristics and location information, thereby realizing the abnormal identification of the fetal status of pregnant women.

[0016] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The method obtains time series features and position information based on fetal heart monitoring data of a group of pregnant women and corresponding monitoring result label data through the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture, including: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the mask embedding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to obtain the time series features of the corresponding monitoring result labels; According to the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the position information of the corresponding monitoring result label is obtained through the position encoding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture.

[0017] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture is provided. According to the time series features and position information, an abnormality monitoring module based on a Transformer architecture of an abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to calculate attention weights and association differences to identify abnormal points in the time series features and position information, thereby realizing abnormal status recognition of pregnant women and fetuses, including: Based on time series features and location information, the multi-head attention layer in the anomaly monitoring module based on the Transformer architecture is combined with a sliding window method and a minimization optimization function to identify error sequences. The error sequences are then expanded through anomaly filling, and continuous or overlapping error sequences are merged to form an abnormality recognition signal for the maternal and fetal status. Through the multi-layer perceptron in the anomaly monitoring module based on the Transformer architecture, feature extraction is performed based on the output of the corresponding multi-head attention layer to obtain the anomaly recognition signal features; Through the projection layer in the abnormality monitoring module based on the Transformer architecture, the abnormality recognition signal features are mapped to the required output dimension, which is the abnormality recognition result of the fetal status of the pregnant woman.

[0018] According to the present invention, a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is provided. The method uses the dual neural network architecture to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data of a group of pregnant women, and trains the fetal heart monitoring auxiliary interpretation model, further comprising: The fetal physiological status classification results obtained by the classification and recognition neural network architecture in the dual neural network architecture and the abnormality recognition results of the fetal status of pregnant women obtained by the abnormality monitoring and evaluation neural network architecture are integrated to obtain the fetal heart monitoring results.

[0019] The present invention also provides a fetal heart monitoring auxiliary system, comprising: The data receiving module is used to receive fetal heart monitoring data of the pregnant woman to be tested from at least one terminal; A fetal heart monitoring auxiliary module is configured to obtain a fetal heart monitoring result of the pregnant woman to be tested based on the fetal heart monitoring data of the pregnant woman to be tested and using a fetal heart monitoring auxiliary interpretation model obtained by any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture; The data output module is used to output the fetal heart monitoring results of the pregnant woman to be tested to at least one terminal.

[0020] It should be noted that a terminal refers to an input and output device connected to a computer system. Depending on their functions, terminals can be divided into various types: smart terminals or intelligent terminals, dumb terminals, interactive terminals or online terminals. Terminals can specifically be various mobile communication devices, such as mobile phones, tablet computers, etc. The purpose of this article is to provide users with the function of inputting and outputting data.

[0021] The present invention also provides an electronic device comprising a processor and a memory storing a computer program. When the processor executes the computer program, it implements any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture.

[0022] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture.

[0023] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture.

[0024] The present invention provides a method and system for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture, which can bring at least the following beneficial effects: (1) By inputting the dual-lead one-dimensional time series data of fetal heart rate and uterine contraction in parallel, the signal coupling interference problem in traditional single-channel monitoring is solved, and the recognition accuracy of the fetal heart monitoring auxiliary interpretation model for complex working conditions such as fetal hypoxia and uterine contraction compression is greatly improved.

[0025] (2) The time-aligned dual-signal sampling strategy is adopted to retain the time correlation characteristics of the original monitoring waveform, which can improve the prediction accuracy of the model in time-dependent pathological states such as late deceleration and variable deceleration.

[0026] (3) A dual-branch neural network architecture is adopted, where each branch processes the signal of a lead independently, making full use of the complementary information between signals of different leads.

[0027] (4) Using the classification recognition neural network architecture, feature extraction and dimensionality reduction are performed through the feature extraction module, the first embedding layer, and the classification recognition module. The features extracted by each branch are spliced ​​in the channel dimension to form a comprehensive feature vector, which improves the richness and accuracy of feature expression. The comprehensive feature vector is mapped to the output category through the fully connected layer to achieve accurate classification and recognition of the fetal physiological state.

[0028] (5) Using the anomaly monitoring and evaluation neural network architecture, the second embedding layer, the anomaly monitoring module based on the Transformer architecture, and the projection layer are used to calculate the signal error, statistically analyze the error distribution, and use optimization technology to find the anomaly detection threshold. The degree of abnormality of the signal is comprehensively evaluated, and the error sequence is found to achieve accurate identification of abnormal fetal status in fetal heart monitoring signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 This is one of the flow charts of a method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture provided by the present invention.

[0031] Figure 2 The second flow chart of the method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture provided by the present invention shows the dual neural network architecture of the fetal heart monitoring auxiliary interpretation model.

[0032] Figure 3 This is a structural diagram of a system for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture provided by the present invention.

[0033] Figure 4 This is a schematic structural diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0035] Figure 1 and 2 A flowchart of a method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture provided by the present invention.

[0036] The executor of the method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional timing signals and dual neural network architecture provided by the present invention can be any applicable terminal side device or network side device, such as a construction device for a fetal heart monitoring auxiliary interpretation model based on one-dimensional timing signals and dual neural network architecture.

[0037] See also Figure 1 The present invention provides a method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture, which may include: S110. Obtain fetal heart monitoring data and corresponding monitoring result label data of a group of pregnant women, wherein the fetal heart monitoring data includes one-dimensional time series data of fetal heart rate and one-dimensional time series data of uterine contraction, and the monitoring result label data includes a fetal physiological state label and a pregnant woman's fetal state abnormality interval label.

[0038] In one embodiment, fetal heart monitoring data can be collected by an external fetal heart monitoring device, and monitoring result label data can be manually labeled. Specifically, the one-dimensional time series data of fetal heart rate represents the sequence of changes in the frequency of fetal heartbeats over time, and the one-dimensional time series data of uterine contractions represents the sequence of changes in the intensity and frequency of uterine muscle contractions over time.

[0039] In one embodiment, fetal heart monitoring data is organized in the form of a two-dimensional array, where the first row of the array is specifically used to store one-dimensional time-series signal data for FHR, and the second row of the array is used to store one-dimensional time-series signal data for UC. Each column of the array represents a specific sampling data point, namely, the FHR value and UC value recorded at the same time point. All signal data can be directly acquired by a professional external fetal heart monitoring instrument. This instrument has high precision and real-time performance, ensuring that the acquired FHR and UC signals accurately reflect the actual physiological status of the fetus and mother. During the acquisition process, the instrument continuously records FHR and UC signals at a preset sampling frequency, forming continuous time series data. Each sampling data point is an element in the array, and its value represents the measured value of FHR or UC at a specific time point. The sequential order of the data points strictly corresponds to the chronological order, ensuring the effectiveness and accuracy of the time series analysis.

[0040] In one embodiment, the fetal physiological state includes any one of the following or any combination thereof: fetal hypoxia state, blood pH state, blood sample state, and blood carbon dioxide state, wherein the fetal hypoxia state is marked in a classified form, marked as "yes" or "no", and the blood pH state, blood oxygen state, and blood carbon dioxide state are marked in segmented intervals, divided into three categories: "good", "abnormal", and "severely abnormal".

[0041] In one embodiment, a maternal fetal status abnormality interval label is used to indicate the specific time series interval within a time series array where a maternal fetal abnormality occurs. Within the time series array, the start and end time points of the abnormality are precisely marked and represented as an abnormality index interval, such as "[start time, end time]," to clearly indicate the specific time period during which the abnormality occurred. A textual description or code of the abnormality type can also be added, such as fetal distress, bradycardia, tachycardia, uterine contractions, and other abnormalities.

[0042] In one embodiment, after acquiring fetal heart rate monitoring data and corresponding monitoring result label data, they can be preprocessed. The preprocessing can include any one or any combination of the following: linear interpolation, sample clipping, and normalization. In this embodiment, linear interpolation is first performed on the signal to fill in any missing data and ensure signal continuity. Then, sample clipping and normalization are performed on the signal to scale the signal values ​​to a uniform range, eliminating dimensional differences and improving the convergence speed and stability of the algorithm.

[0043] The simultaneous use of two physiological time-series signals—fetal heart rate (FHR) and uterine contraction (UC)—can more comprehensively reflect the state of fetal-maternal interaction. FHR changes are physiologically correlated with uterine contraction intensity (e.g., changes in placental oxygen supply during uterine contractions affect FHR). Joint dual-signal training helps the model capture this dynamic coupling, providing greater clinical significance than single-signal analysis. Simultaneously using fetal physiological status labels (e.g., hypoxia, pH) and abnormal interval labels (e.g., distress, bradycardia), it enables joint supervision of microphysiological indicators and macroclinical diagnosis. This hierarchical labeling system guides the neural network to separately learn basic physiological features (e.g., baseline variability) and advanced pathological patterns (e.g., late decelerations), improving model generalization. Traditional fetal monitoring interpretation relies on expert experience (e.g., baseline, acceleration, and deceleration rules in the FIGO standard). However, this approach, trained directly on raw one-dimensional time-series data, automatically discovers potential discriminative features (e.g., UC-FHR phase differences in specific frequency bands), potentially uncovering hidden abnormal patterns that are difficult to interpret manually.

[0044] S120. Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, a dual neural network architecture is used to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data, and a fetal heart monitoring auxiliary interpretation model is trained.

[0045] In one embodiment, see Figure 2The dual neural network architecture includes a classification and recognition neural network architecture and an anomaly monitoring and evaluation neural network architecture, wherein the classification and recognition neural network architecture includes a feature extraction module, a classification and recognition module, and a first embedding layer connecting the feature extraction module and the classification and recognition module, the feature extraction module includes a time series representation submodule (a one-dimensional convolutional neural network submodule), a global average pooling submodule, an encoder, and a projector, the classification and recognition module includes a fully connected layer, a batch normalization submodule and a Softmax activation function, and wherein the anomaly monitoring and evaluation neural network architecture includes a second embedding layer, an anomaly monitoring module based on the Transformer architecture, a projection layer, the second embedding layer includes a mask embedding submodule and a position encoding submodule, the anomaly monitoring module based on the Transformer architecture includes a first encoder layer, a second encoder layer and a third encoder layer, and the first encoder layer, the second encoder layer and the third encoder layer respectively include a multi-head attention layer and a multi-layer perceptron.

[0046] The classification and recognition neural network architecture is designed to output the specific physiological state of the fetus at that time, including fetal hypoxia, blood pH, blood sample status, and blood carbon dioxide status. Its input data is the preprocessed FHR-UC time series array. The preprocessed time series array is encoded with a sine-cosine hybrid encoding of the extracted features through the Gram angle field change. The specific calculation process is as follows:

[0047] in x i is the time series data that has been normalized and preprocessed. θ i is the polar coordinate angle, G GASF ( i , j ) is the Gram angle and the first i and j The matrix elements at the moment, G GADF ( i , j ) is the first Gram angle difference matrix i and j Matrix element at time.

[0048] The encoder extracts local features from the signal, implements downsampling, and outputs a high-dimensional feature representation, which helps the network better understand the complexity and diversity of the signal. The features output by the two branches are concatenated along the channel dimension to form a comprehensive feature vector. This is then input into the projection layer, which maps the encoded features to a low-dimensional space and maps the fused features to the final classification result.

[0049] In one embodiment, S120 may include: S1201, based on the fetal heart monitoring data of the pregnant women group and the corresponding monitoring result label data, extract features through the feature extraction module of the classification and recognition neural network architecture in the dual neural network architecture to obtain comprehensive features; S1202. Based on the comprehensive features, a comprehensive feature vector is obtained through the first embedding layer of the classification recognition neural network architecture in the dual neural network architecture. The vector is a learnable 64-dimensional embedding vector. S1203. Based on the comprehensive feature vector, the comprehensive feature vector is mapped to the output category through the classification recognition module of the classification recognition neural network architecture in the dual neural network architecture to realize the classification of the fetal physiological state. Specifically, the comprehensive feature vector is mapped to the output category through the fully connected layer to realize the classification and recognition of the fetal state (such as normal, abnormal, etc.). In the fully connected layer, through the learning and adjustment of the weights, the network can gradually learn how to judge the fetal state based on the input features and give the corresponding classification results.

[0050] Specifically, S1201 may include: Based on fetal heart rate monitoring data from pregnant women and the corresponding monitoring result labels, the feature extraction module's temporal representation submodule uses a fully connected layer and positional encoding to jointly model the preprocessed dual-lead temporal signals. A fully connected network maps the input one-dimensional temporal signal to a high-dimensional latent space using a linear transformation matrix of dimension [2, 128]. The input dimension is 2 for the dual-channel FHR-UC signal, and the output dimension is 128 for the expanded feature representation space. This linear projection operation, through parameterized learning of the weight matrix, can adaptively capture the multi-scale frequency components in the fetal heart rate monitoring signal, demonstrating stronger feature representation capabilities than fixed convolution kernel designs. Based on multiple convolutional features, the encoder of the feature extraction module is used to encode them separately to obtain multiple encoded features to reduce the feature dimension while retaining key information; According to the multiple encoded features, the projector of the feature extraction module maps the encoded features to a classification space that is easier to classify, and the multiple mapped features are spliced ​​in the channel dimension to obtain comprehensive features.

[0051] In another embodiment, LSTM (Long Short-Term Memory Network) can also be used to replace the time series representation submodule to implement data processing, directly process the preprocessed time series array, and LSTM can be combined with the attention mechanism output to calculate the prediction error.

[0052] In another embodiment, the self-attention mechanism in the Transformer architecture can also replace the temporal representation submodule to implement feature extraction. Through the multi-layer Transformer encoder, high-level features are gradually extracted, and the self-attention mechanism can simultaneously consider the relationship between all points in the sequence and capture global dependencies.

[0053] The anomaly detection and assessment neural network architecture aims to identify anomalous signals, including their location, severity, and possible anomaly type. The data embedding layer comprises two submodules: token encoding and position encoding, respectively, for extracting the signal's temporal features and positional information. This layer converts the input one-dimensional time series signal into a high-dimensional embedding vector for processing by subsequent network layers. The encoder layer consists of multiple encoder blocks, each of which includes an attention layer and a multi-layer perceptron (MLP). The attention layer identifies anomalies in the signal by calculating attention weights and correlation differences. The anomaly detection process first uses a sliding window to calculate the mean and standard deviation change of the prediction error for each signal within each window. The minimize optimization function is used to determine the anomaly detection threshold based on factors such as the mean and standard deviation change, the number of errors exceeding the threshold, and the number of consecutive sequences. Finally, error sequences exceeding the threshold are identified and expanded using an "anomaly filling" strategy to more comprehensively capture anomalous events. Consecutive or overlapping anomaly sequences are merged to form the final anomaly detection result. The MLP further processes the output of the attention layer to extract higher-level features. The final projection layer maps the output of the encoder layer to the desired output dimension, which is the recognition result of the abnormal signal.

[0054] In one embodiment, S120 may include: S1201′, based on the fetal heart monitoring data of the pregnant women group and the corresponding monitoring result label data, obtain time series features and position information through the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture; S1202', based on the time series features and position information, the abnormality monitoring module based on the Transformer architecture of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture calculates the attention weight and association difference to identify abnormal points in the time series features and position information, thereby realizing the abnormal identification of the fetal state of the pregnant woman.

[0055] Specifically, S1201′ may include: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the mask embedding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to obtain the time series features of the corresponding monitoring result labels; According to the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the position information of the corresponding monitoring result label is obtained through the position encoding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture.

[0056] Specifically, S1202' may include: Based on time series features and location information, the multi-head attention layer in the anomaly monitoring module based on the Transformer architecture is combined with a sliding window method and a minimization optimization function to identify error sequences. The error sequences are then expanded through anomaly filling, and continuous or overlapping error sequences are merged to form an abnormality recognition signal for the maternal and fetal status. Through the multi-layer perceptron in the anomaly monitoring module based on the Transformer architecture, feature extraction is performed based on the output of the corresponding multi-head attention layer to obtain the anomaly recognition signal features; Through the projection layer in the abnormality monitoring module based on the Transformer architecture, the abnormality recognition signal features are projected and mapped to the required output dimension, which is the abnormality recognition result of the pregnant woman's fetal status.

[0057] Specifically, the error sequence identification process may be as follows.

[0058] First, the original fetal heart monitoring time series signal in the fetal heart monitoring data is preprocessed, wherein linear interpolation fills the missing values ​​in the signal to ensure the continuity of the signal. The calculation formula is as follows:

[0059] in,( x 0, y 0) and ( x 1, y 1) are two known points, x is the value to be interpolated, y is the output value after interpolation; Sample cropping removes invalid or redundant parts of the signal to improve data quality; normalization processing adjusts the amplitude of the signal to a uniform range, eliminates dimensional differences, and facilitates subsequent processing.

[0060] A 1D CNN effectively extracts local features from preprocessed time series signals, capturing patterns and trends. The features then undergo global average pooling, an encoder, and a projector to further compress and refine the feature information, ultimately producing the input for the embedding layer. The embedding layer input is fed into a fully connected layer, which integrates and transforms the features. Finally, the network outputs monitoring results (e.g., fetal hypoxia) through a softmax function. The softmax function converts the output of the fully connected layer into a probability distribution, representing the probability of the fetus being in different states (e.g., normal or hypoxic).

[0061] In addition to classification and recognition, the process also includes an anomaly detection and assessment neural network, which utilizes the Transformer architecture. The Transformer architecture comprises components such as an embedding layer, a multi-head attention layer, a multi-layer perceptron, and an encoder. The embedding layer converts the input signal into a high-dimensional vector representation. The multi-head attention layer captures dependencies between different locations in the signal, focusing on important features and improving computational efficiency. The multi-layer perceptron performs a nonlinear transformation on the output of the attention layer. The encoder further integrates and refines features. Ultimately, the network outputs an evaluation metric through the projection layer, which is used to quantitatively assess the degree of fetal abnormality. To enhance the Transformer architecture's ability to identify abnormal patterns in fetal heart and contraction signals, anomaly detection algorithms based on statistical methods and optimization techniques can be deeply integrated with the Transformer architecture. Specifically, before inputting the signal into the Transformer architecture, statistical analysis can be performed on the signal within each window to calculate its error. This analysis can then be used to obtain statistics such as the mean and variance of the error, as well as advanced statistical features such as the signal's skewness, crest factor, peak factor, and impulse factor.

[0062] The threshold that minimizes the cost function is found through optimization technology to define the threshold for anomaly detection. During this process, the algorithm will use historical data and the characteristics of the current signal to dynamically adjust the threshold to adapt to the anomaly detection needs in different situations. The abnormal sequence identification and evaluation part is responsible for identifying error sequences that exceed the threshold, calculating the maximum error value of each abnormal sequence, and pruning possible false positives based on the percentage change between the maximum errors. An anomaly score is calculated for each pruned abnormal sequence to indicate its degree of abnormality. Identify error sequences that exceed the threshold and consider a certain "abnormal filling" strategy to expand these sequences to more accurately capture the full picture of abnormal events. Calculate the maximum error value of each abnormal sequence and prune possible false positives based on the percentage change between the maximum errors and the length, frequency and other characteristics of the sequence. The accuracy of anomaly detection is improved by pruning false positives. An anomaly score is calculated for each pruned abnormal sequence. This score comprehensively considers multiple factors such as error value, sequence characteristics, and the model's confidence in the anomaly, and can more comprehensively represent its degree of abnormality. In addition, the algorithm will provide doctors with early warnings and prompts about the trend of changes in fetal status based on the distribution and changes of the anomaly score. Assume that n The reference health status time series of the sampling points is X = [ x 1, x 2,…, x n ], the time series of real-time monitoring is , then the calculation formula for the abnormal score of the monitoring data is as follows:

[0063] Among them, Cov() is used to calculate the covariance, Var() is used to calculate the variance, is the mean value of the time series.

[0064] In the anomaly monitoring and assessment neural network architecture, optimization techniques are used to find the threshold that minimizes the cost function, which is used to define the anomaly detection threshold. This dynamic threshold adjustment strategy can automatically adjust the threshold based on the characteristics of different signals, improving the accuracy and robustness of anomaly detection.

[0065] An anomaly score is calculated for each pruned abnormal sequence, taking into account multiple factors to more comprehensively represent the degree of abnormality. Based on the distribution and changes in the anomaly score, doctors are provided with early warnings and alerts regarding trends in fetal status, helping to promptly identify and address potential fetal health risks.

[0066] In one embodiment, the method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture provided by the present invention may further include the steps of: The fetal physiological status classification results obtained by the classification and recognition neural network architecture in the dual neural network architecture and the abnormality recognition results of the fetal status of pregnant women obtained by the abnormality monitoring and evaluation neural network architecture are integrated to obtain the fetal heart monitoring results.

[0067] By integrating the outputs of the two networks, a complete AI-assisted fetal heart rate monitoring diagnosis report is generated. The report can include information such as the fetal real-time status, abnormal monitoring results, and abnormality assessment, providing doctors with a comprehensive diagnosis basis.

[0068] This example uses the CTU-UHB public medical dataset for model training and validation. This dataset, jointly constructed by the Department of Biomedical Engineering at the Czech Technical University in Prague (CTU) and the Department of Obstetrics and Gynecology at the University Hospital in Brno (UHB), contains complete monitoring records of 552 pregnant women with singleton pregnancies, gestational age ≥36 weeks, and no known developmental defects. Data collection strictly adheres to medical ethics standards. Each sample includes two-lead fetal heart rate (FHR) and uterine contraction pressure (UC) signals collected simultaneously at a 4Hz sampling rate, with signal durations ranging from 20 to 60 minutes. The data includes 18 clinical indicators, including maternal age, gestational age, and parity. The training set (387 cases) and the test set (165 cases) were randomly divided into a 7:3 ratio to ensure that the distribution of maternal age and gestational age did not differ significantly between the datasets. The positional encoding uses a learnable sinusoidal positional encoding with a dimension of 128, and a two-layer MLP Transformer encoder to achieve multi-scale feature interaction and feature fusion. The model performance of this method on the test set (165 cases) is as follows:

[0069] The present invention provides a method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture. The method adopts a time-aligned dual-signal sampling strategy, and inputs dual-lead one-dimensional time series data of fetal heart rate and uterine contraction in parallel. A dual-branch neural network architecture is adopted, and each branch independently processes the signal of a lead, which can make full use of the complementary information between different lead signals. The trained fetal heart monitoring auxiliary interpretation model can realize accurate identification of fetal physiological status and abnormal fetal status of pregnant women, and greatly improve the recognition accuracy of the fetal heart monitoring auxiliary interpretation model for complex working conditions such as fetal hypoxia and uterine contraction compression, and time-dependent pathological conditions such as late deceleration and variable deceleration.

[0070] The present invention provides a method and system for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and a dual neural network architecture, which can bring at least the following beneficial effects: (1) By inputting the dual-lead one-dimensional time series data of fetal heart rate and uterine contraction in parallel, the signal coupling interference problem in traditional single-channel monitoring is solved, and the recognition accuracy of the fetal heart monitoring auxiliary interpretation model for complex working conditions such as fetal hypoxia and uterine contraction compression is greatly improved.

[0071] (2) The time-aligned dual-signal sampling strategy is adopted to retain the time correlation characteristics of the original monitoring waveform, which can improve the prediction accuracy of the model in time-dependent pathological states such as late deceleration and variable deceleration.

[0072] (3) A dual-branch neural network architecture is adopted, where each branch processes the signal of a lead independently, making full use of the complementary information between signals of different leads.

[0073] (4) Using the classification recognition neural network architecture, feature extraction and dimensionality reduction are performed through the feature extraction module, the first embedding layer, and the classification recognition module. The features extracted by each branch are spliced ​​in the channel dimension to form a comprehensive feature vector, which improves the richness and accuracy of feature expression. The comprehensive feature vector is mapped to the output category through the fully connected layer to achieve accurate classification and recognition of the fetal physiological state.

[0074] (5) Using the anomaly monitoring and evaluation neural network architecture, the second embedding layer, the anomaly monitoring module based on the Transformer architecture, and the projection layer are used to calculate the signal error, statistically analyze the error distribution, and use optimization technology to find the anomaly detection threshold. The degree of abnormality of the signal is comprehensively evaluated, and the error sequence is found to achieve accurate identification of abnormal fetal status in fetal heart monitoring signals.

[0075] The fetal heart monitoring auxiliary system provided by the present invention is described below. The fetal heart monitoring auxiliary system described below and the construction method of the fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signal and dual neural network architecture described above can be referenced to each other.

[0076] The present invention provides a fetal heart monitoring auxiliary system, which may include: The data receiving module is used to receive fetal heart monitoring data of the pregnant woman to be tested from at least one terminal; A fetal heart monitoring auxiliary module is configured to obtain a fetal heart monitoring result of the pregnant woman to be tested based on the fetal heart monitoring data of the pregnant woman to be tested and using a fetal heart monitoring auxiliary interpretation model obtained by any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture; The data output module is used to output the fetal heart monitoring results of the pregnant woman to be tested to at least one terminal.

[0077] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3As shown, the electronic device may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to perform the following steps: receiving fetal heart monitoring data of a pregnant woman to be tested from at least one terminal; According to the fetal heart monitoring data of the pregnant woman to be tested, the fetal heart monitoring auxiliary interpretation model obtained by any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is used to obtain the fetal heart monitoring result of the pregnant woman to be tested; Outputting the fetal heart monitoring result of the pregnant woman to be tested to at least one terminal.

[0078] Furthermore, the logic instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0079] In another aspect, the present invention further provides a computer program product, comprising a computer program, which may be stored on a non-transitory computer-readable storage medium, and wherein when the computer program is executed by a processor, the computer is capable of performing the following steps: receiving fetal heart monitoring data of a pregnant woman to be tested from at least one terminal; According to the fetal heart monitoring data of the pregnant woman to be tested, the fetal heart monitoring auxiliary interpretation model obtained by any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is used to obtain the fetal heart monitoring result of the pregnant woman to be tested; Outputting the fetal heart monitoring result of the pregnant woman to be tested to at least one terminal.

[0080] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is configured to execute the following steps when executed by a processor: receiving fetal heart monitoring data of a pregnant woman to be tested from at least one terminal; According to the fetal heart monitoring data of the pregnant woman to be tested, the fetal heart monitoring auxiliary interpretation model obtained by any of the above-mentioned methods for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture is used to obtain the fetal heart monitoring result of the pregnant woman to be tested; Outputting the fetal heart monitoring result of the pregnant woman to be tested to at least one terminal.

[0081] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0082] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture, characterized in that: include: Obtain fetal heart monitoring data and corresponding monitoring result label data of a group of pregnant women, wherein the fetal heart monitoring data includes one-dimensional time series data of fetal heart rate and one-dimensional time series data of uterine contraction, and the monitoring result label data includes fetal physiological state labels and abnormal fetal state interval labels of pregnant women; Based on the fetal heart monitoring data of pregnant women and the corresponding monitoring result label data, a dual neural network architecture is used to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data, and a fetal heart monitoring auxiliary interpretation model is trained.

2. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 1 is characterized in that: The fetal physiological state includes any one or any combination of the following: fetal hypoxia state, blood pH state, blood sample state, blood carbon dioxide state; Abnormal fetal status of pregnant women includes any one of the following or any combination thereof: fetal distress, bradycardia, tachycardia, uterine contraction; The dual neural network architecture includes a classification and recognition neural network architecture and an anomaly monitoring and evaluation neural network architecture, wherein the classification and recognition neural network architecture includes a feature extraction module, a classification and recognition module, and a first embedding layer connecting the feature extraction module and the classification and recognition module, the feature extraction module includes a one-dimensional convolutional neural network sub-module, a global average pooling sub-module, an encoder, and a projector, the classification and recognition module includes a fully connected layer, a batch normalization sub-module and a Softmax activation function, and wherein the anomaly monitoring and evaluation neural network architecture includes a second embedding layer, an anomaly monitoring module based on a Transformer architecture, and a projection layer, the second embedding layer includes a mask embedding sub-module and a position encoding sub-module, the anomaly monitoring module based on the Transformer architecture includes a first encoder layer, a second encoder layer and a third encoder layer, and the first encoder layer, the second encoder layer and the third encoder layer respectively include a multi-head attention layer and a multi-layer perceptron.

3. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 2 is characterized in that: The method uses a dual neural network architecture to learn the relationship between fetal heart monitoring data and corresponding monitoring result label data based on the fetal heart monitoring data of a group of pregnant women, and trains a fetal heart monitoring auxiliary interpretation model, including: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, feature extraction is performed through the feature extraction module of the classification and recognition neural network architecture in the dual neural network architecture to obtain comprehensive features; According to the comprehensive features, a comprehensive feature vector is obtained through the first embedding layer of the classification recognition neural network architecture in the dual neural network architecture; According to the comprehensive feature vector, the comprehensive feature vector is mapped to the output category through the classification recognition module of the classification recognition neural network architecture in the dual neural network architecture to realize the classification of the fetal physiological state.

4. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 3 is characterized in that: The fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data are extracted by the feature extraction module of the classification and recognition neural network architecture in the dual neural network architecture to obtain comprehensive features, including: Based on the fetal heart monitoring data of pregnant women and the corresponding monitoring result label data, the one-dimensional convolutional neural network submodule of the feature extraction module uses different sizes and step sizes of convolution kernels to capture the different frequency components and time series relationships in the fetal heart monitoring data and the corresponding monitoring result label data, and obtain multiple convolution features; According to the multiple convolved features, the encoder of the feature extraction module is used to encode them respectively to obtain multiple encoded features; According to the multiple encoded features, the encoded features are mapped to the classification space through the projector of the feature extraction module, and the multiple mapped features are spliced ​​in the channel dimension to obtain the comprehensive features.

5. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 3 is characterized in that: The method uses a dual neural network architecture to learn the relationship between fetal heart monitoring data and corresponding monitoring result label data based on the fetal heart monitoring data of a group of pregnant women, and trains a fetal heart monitoring auxiliary interpretation model, including: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to obtain the time series features and location information; According to the timing characteristics and location information, the abnormality monitoring module based on the Transformer architecture of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture calculates the attention weight and association difference to identify abnormal points in the timing characteristics and location information, thereby realizing the abnormal identification of the fetal status of pregnant women.

6. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 5, characterized in that: The method of obtaining time series features and location information based on the fetal heart monitoring data of the pregnant women group and the corresponding monitoring result label data through the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture includes: Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the mask embedding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture is used to obtain the time series features of the corresponding monitoring result labels; Based on the fetal heart monitoring data of the pregnant women and the corresponding monitoring result label data, the position information of the corresponding monitoring result label is obtained through the position encoding submodule of the second embedding layer of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture; Based on the time series features and location information, the abnormality monitoring module based on the Transformer architecture of the abnormality monitoring and evaluation neural network architecture in the dual neural network architecture calculates attention weights and association differences to identify abnormal points in the time series features and location information, thereby realizing abnormal status recognition of pregnant women and fetuses, including: Based on time series features and location information, the multi-head attention layer in the anomaly monitoring module based on the Transformer architecture is combined with a sliding window method and a minimization optimization function to identify error sequences. The error sequences are then expanded through anomaly filling, and continuous or overlapping error sequences are merged to form an abnormality recognition signal for the maternal and fetal status. Through the multi-layer perceptron in the anomaly monitoring module based on the Transformer architecture, feature extraction is performed based on the output of the corresponding multi-head attention layer to obtain the anomaly recognition signal features; Through the projection layer in the abnormality monitoring module based on the Transformer architecture, the abnormality recognition signal features are projected and mapped to the required output dimension, which is the abnormality recognition result of the pregnant woman's fetal status.

7. The method for constructing a fetal heart monitoring auxiliary interpretation model based on one-dimensional time series signals and dual neural network architecture according to claim 6, characterized in that: The method further includes: using a dual neural network architecture to learn the relationship between the fetal heart monitoring data and the corresponding monitoring result label data based on the fetal heart monitoring data of the pregnant women group and the corresponding monitoring result label data, and training a fetal heart monitoring auxiliary interpretation model. The fetal physiological status classification results obtained by the classification and recognition neural network architecture in the dual neural network architecture and the abnormality recognition results of the fetal status of pregnant women obtained by the abnormality monitoring and evaluation neural network architecture are integrated to obtain the fetal heart monitoring results.

8. A fetal heart monitoring auxiliary system, characterized in that: include: The data receiving module is used to receive fetal heart monitoring data of the pregnant woman to be tested from at least one terminal; A fetal heart monitoring auxiliary module, configured to obtain a fetal heart monitoring result of the pregnant woman to be tested based on the fetal heart monitoring data of the pregnant woman to be tested and using a fetal heart monitoring auxiliary interpretation model obtained by the method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture according to any one of claims 1 to 7; The data output module is used to output the fetal heart monitoring results of the pregnant woman to be tested to at least one terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the following steps are implemented: receiving fetal heart monitoring data of a pregnant woman to be tested from at least one terminal; Obtaining a fetal heart monitoring result of the pregnant woman to be tested based on the fetal heart monitoring data of the pregnant woman to be tested, using a fetal heart monitoring auxiliary interpretation model obtained by the method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture according to any one of claims 1 to 7; Outputting the fetal heart monitoring result of the pregnant woman to be tested to at least one terminal.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the following steps are implemented: receiving fetal heart monitoring data of a pregnant woman to be tested from at least one terminal; Obtaining a fetal heart monitoring result of the pregnant woman to be tested based on the fetal heart monitoring data of the pregnant woman to be tested, using a fetal heart monitoring auxiliary interpretation model obtained by the method for constructing a fetal heart monitoring auxiliary interpretation model based on a one-dimensional time series signal and a dual neural network architecture according to any one of claims 1 to 7; Outputting the fetal heart monitoring result of the pregnant woman to be tested to at least one terminal.

Images