Premature delivery prediction method and device based on uterine vector electromyographic signals and processing equipment

Abstract

According to the premature delivery prediction method and device based on the uterus vector electromyographic signals and the processing equipment, through a specially-designed novel electrode arrangement scheme, the collected uterus vector electromyographic signals are different from traditional uterus surface electromyographic signals, the spatial characteristics of uterus contraction activities can be reflected more comprehensively, and the premature delivery prediction accuracy is improved. Compared with the prior art, the premature delivery detection method has the advantages that the premature delivery detection method contains more signal characteristics related to premature delivery, so that high premature delivery prediction precision can be achieved under the condition of fewer measuring electrodes, and clinical application value is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of non-invasive uterine electromyographic signal detection, and in particular to a method, device and processing equipment for predicting premature birth based on uterine vector electromyographic signals. Background Art

[0002] Premature birth screening helps to identify risks early and reduce the occurrence of premature fetal complications. Currently, the main method of premature birth prediction is based on uterine electromyography monitoring. The uterine electromyography monitoring method actually places electrodes on the abdominal surface of the pregnant woman to record the electrical activity transmitted from the uterine myometrium to the abdominal surface to obtain a uterine surface electromyogram.

[0003] Uterine surface electromyography is a non-invasive monitoring method that collects electromyographic signals from the abdominal surface to reflect the electrical signal characteristics of the uterine myometrial cell contraction process. Subsequently, the electromyographic signal-related indicators, including time domain indicators, frequency domain indicators, electromyographic conduction velocity and nonlinear indicators, are analyzed to judge the true and false contractions of the uterus and the degree of contraction, thereby realizing the prediction of premature birth.

[0004] However, the inventors of this application have found that the current methods for predicting premature birth have the following problems: 1) Signal feature extraction for premature birth prediction is complex. Uterine surface electromyography is susceptible to interference, including movement, fetal movement, cardiac electrical activity, and breathing. Extracting time-domain, frequency-domain, and electromyographic conduction velocity metrics from the collected electromyographic signals requires complex filtering and signal denoising. These signal processing steps can result in the loss of useful signal features, reducing prediction accuracy to a certain extent. 2) Uterine surface electromyography (SEM) cannot monitor the full range of uterine electrophysiological activity. Uterine contractions are anisotropic, three-dimensional spatial activities. Uterine electromyographic signals can be considered the spatial summation of the electrical activity of myometrial cells. However, SEM electrodes are placed on the maternal abdomen. The collected SEM signals can only characterize the localized uterine contraction characteristics, not the full range of spatial activity. 3) Some proposals have proposed the use of multiple electrodes (16-64) to collect uterine myoelectric signals from a larger area, thereby obtaining richer uterine electrophysiological activity characteristics. However, this electrode arrangement is too complicated, and the labor and data acquisition costs are too high, which is not conducive to clinical application and promotion. Summary of the Invention

[0005] The present application provides a method, apparatus, and processing equipment for predicting premature birth based on uterine vector electromyographic signals. Through a specially designed novel electrode arrangement scheme, the collected uterine vector electromyographic signals are different from traditional uterine surface electromyographic signals, and can more comprehensively reflect the spatial characteristics of uterine contraction activity and contain more signal features related to premature birth. Therefore, with fewer measuring electrodes, a higher prediction accuracy for premature birth can be obtained, which has clinical application value.

[0006] In a first aspect, the present application provides a device for predicting premature birth based on uterine vector electromyographic signals, the device comprising: Obtaining uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs, the uterine vector myoelectric signals specifically including uterine myoelectric signals in three orthogonal directions corresponding to the three electrode pairs, and the uterine vector myoelectric signals also including a differential signal obtained by differentiating two signals from each electrode pair; Inputting the uterine vector electromyographic signals into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to a preset electrode arrangement, and the prediction model is used to predict the corresponding premature birth based on the uterine vector electromyographic signals input to the model; Extract the premature birth prediction results output by the prediction model.

[0007] In a second aspect, the present application provides a device for predicting premature birth based on uterine vector electromyographic signals, the device comprising: an acquisition unit, configured to acquire uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs, and the uterine vector myoelectric signals specifically include uterine myoelectric signals in three orthogonal directions corresponding to the three electrode pairs, and also include a differential signal obtained by differentiating two signals from each electrode pair; a prediction unit, configured to input the uterine vector electromyographic signal into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to a preset electrode arrangement, and the prediction model is configured to predict a corresponding premature birth based on the uterine vector electromyographic signal input to the model; The extraction unit is used to extract the premature birth prediction result output by the prediction model.

[0008] In a third aspect, the present application provides a processing device comprising a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method provided in the first aspect of the present application or any possible implementation of the first aspect of the present application is executed.

[0009] In a fourth aspect, the present application provides a computer-readable storage medium, which stores multiple instructions, and the instructions are suitable for a processor to load to execute the method provided in the first aspect of the present application or any possible implementation of the first aspect of the present application.

[0010] From the above content, it can be concluded that this application has the following beneficial effects: Aiming at the goal of predicting premature birth based on uterine vector electromyographic signals, this application uses a specially designed novel electrode arrangement scheme. The collected uterine vector electromyographic signals are different from traditional uterine surface electromyographic signals. They can more comprehensively reflect the spatial characteristics of uterine contraction activities and contain more signal features related to premature birth. Therefore, with fewer measuring electrodes, higher prediction accuracy for premature birth can be achieved, which has clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of a method for predicting premature birth based on uterine vector electromyographic signals in this application; Figure 2 A schematic diagram of a scenario of the electrode arrangement scheme of this application; Figure 3 This is a structural diagram of a premature birth prediction device based on uterine vector electromyographic signals in this application; Figure 4 This is a structural diagram of the processing equipment for this application. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0014] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can be changed in the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0015] The division of modules in this application is a logical division. In actual application, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. Moreover, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed into multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this application.

[0016] Before introducing the premature birth prediction method based on uterine vector electromyographic signals provided by this application, the background content involved in this application is first introduced.

[0017] The premature birth prediction method, device and computer-readable storage medium based on uterine vector electromyographic signals provided in this application can be applied to processing equipment. Through a specially designed novel electrode arrangement scheme, the collected uterine vector electromyographic signals are different from traditional uterine surface electromyographic signals, and can more comprehensively reflect the spatial characteristics of uterine contraction activities and contain more signal features related to premature birth. Therefore, under the condition of fewer measuring electrodes, a higher prediction accuracy for premature birth can be obtained, which has clinical application value.

[0018] The method for predicting premature birth based on uterine vector electromyographic signals, as described in this application, can be implemented by a premature birth prediction device based on uterine vector electromyographic signals, or by various types of processing devices, such as a server, physical host, or user equipment (UE) that integrates the premature birth prediction device based on uterine vector electromyographic signals. The premature birth prediction device based on uterine vector electromyographic signals can be implemented using hardware or software. The UE can specifically be a terminal device such as a smartphone, tablet computer, laptop computer, desktop computer, or personal digital assistant (PDA). The processing device can be configured as a device cluster.

[0019] Next, the method for predicting premature birth based on uterine vector electromyographic signals provided by the present application will be introduced.

[0020] First, see Figure 1 , Figure 1 A flowchart of a method for predicting premature birth based on uterine vector electromyographic signals according to the present application is shown. The method for predicting premature birth based on uterine vector electromyographic signals provided by the present application may specifically include the following steps S101 to S103: Step S101: Acquire uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs. The uterine vector myoelectric signals specifically include uterine myoelectric signals in three orthogonal directions corresponding to the three electrode pairs, and also include a differential signal obtained by differentiating two signals from each electrode pair. It can be understood that the present application focuses on the goal of lighter and higher-performance premature birth prediction, and specially designs a novel electrode arrangement scheme. The preset electrode arrangement scheme specifically involves only 6 measurement electrodes, and these 6 electrodes arranged on the maternal skin specifically form the measurement conditions of electrode pairs arranged orthogonally in pairs. In this way, the uterine electromyographic signals corresponding to the three electrode pairs and in three orthogonal directions can be measured, and then combined with the differential signal to further enhance the spatial characteristics and form a uterine vector electromyographic signal with spatial attributes. This special uterine vector electromyographic signal can provide corresponding data input for the premature birth prediction logic based on uterine vector electromyographic signals specially designed by this application.

[0021] That is to say, under the design of the present application, theoretically, only 6 measuring electrodes are needed to complete the signal acquisition of uterine vector electromyographic signals, and the clinical operability is significantly enhanced.

[0022] Among them, the measuring electrodes can usually be electrodes of an existing multi-channel electromyography collector, or in some cases, they can also be electrodes of a multi-channel electromyography collector adapted to the solution of this application.

[0023] In addition, it should be understood that, corresponding to the application of the measuring electrode, a ground electrode, or a reference electrode, is usually also configured, which can be specifically configured at a specific location such as the surface of the mother's thigh.

[0024] Furthermore, in specific operations, the acquisition and processing of the uterine vector electromyographic signal can be either real-time acquisition and processing or extraction and processing of existing signals, which are both possible in actual situations.

[0025] Among them, as mentioned above, in addition to using the electromyographic signals measured by the six measuring electrodes as subsequent data input, this application also performs further processing based on these six basic electromyographic signals to obtain electromyographic signal characteristics of new dimensions to provide richer data input and data reference.

[0026] Specifically, the overall uterine vector electromyographic signal also includes a differential signal obtained by differentiating the two signals of each electrode pair.

[0027] Among them, difference is a mature mathematical operation.

[0028] That is to say, after obtaining the uterine electromyographic signals in three orthogonal directions corresponding to the basic three electrode pairs, the two signals of each electrode pair can be differentiated to obtain the differential signal obtained by differentiating the two signals of each electrode pair. The differential signal and the uterine electromyographic signals in three orthogonal directions corresponding to the basic three electrode pairs constitute the overall uterine vector electromyographic signal. In this way, compared with the original 6-channel uterine electromyographic signal, a richer 9-channel uterine electromyographic signal is formed.

[0029] At the same time, as a specific embodiment, refer to Figure 2 A schematic diagram of a scenario of the electrode arrangement scheme of the present application is shown. For the three electrode pairs involved in the preset electrode arrangement scheme of the present application, specifically: 1) The first electrode pair consists of the first electrode (i.e., electrode 1 or point 1) and the fourth electrode (i.e., electrode 4 or point 4). The first electrode is placed slightly above the center of the mother's abdominal wall, and the fourth electrode is placed slightly below the center of the mother's back. The preset distance between the first electrode and the horizontal line of the mother's waist (or the waist reference line, the horizontal line where the navel is located) (that is, the specific distance when the first electrode is arranged above the center of the mother's abdominal wall) and the preset distance between the fourth electrode and the horizontal line of the mother's waist (that is, the specific distance when the fourth electrode is arranged below the center of the mother's back) are not necessarily equal. This is due to the asymmetry of the abdominal wall and back itself, as well as the body shape differences between different individuals.

[0030] Furthermore, the following limiting conditions may also exist: The first electrode is located above the waistline, generally not exceeding the lowest rib plane of the chest, and the second electrode is located below the waistline on the back, generally not exceeding the end of the coccyx.

[0031] 2) The second electrode pair consists of the second electrode (i.e., electrode 2 or point 2) and the third electrode (i.e., electrode 3 or point 3). The second electrode is placed slightly above the middle of the mother's back, and the third electrode is placed slightly below the middle of the mother's abdominal wall. The line connecting the first electrode pair and the line connecting the second electrode pair form a perpendicular relationship. Among them, the preset distance between the second electrode and the horizontal horizontal line of the mother's waist (that is, the specific distance at which the second electrode is arranged above the center of the waist and back), and the preset distance between the third electrode and the horizontal horizontal line of the mother's waist (that is, the specific distance at which the third electrode is arranged below the center of the mother's abdominal wall) also have the characteristics mentioned above that they are not necessarily equal.

[0032] Moreover, it can be clearly seen from the figure that the line connecting points 1 and 4, and the line connecting points 2 and 3, are perpendicular to each other, that is, they have an angle of 90°.

[0033] 3) The third electrode pair consists of the fifth electrode (i.e., electrode 5 or point 5) and the sixth electrode (i.e., electrode 6 or point 6). The fifth electrode and the sixth electrode are located along the horizontal line of the mother's waist and are located on the left and right sides of the abdominal wall, respectively. The connecting line of the third electrode pair forms an orthogonal relationship with the connecting line of the first electrode pair and the connecting line of the second electrode pair, respectively.

[0034] Wherein, on the horizontal line at the maternal waist level, the preset distance between the fifth electrode and the navel is equal to the preset distance between the sixth electrode and the navel.

[0035] Moreover, it can be clearly seen from the figure that the line connecting points 1 and 4, and the line connecting points 2 and 3, are both perpendicular to the line connecting points 5 and 6. In other words, the plane in which points 1, 2, 3 and 4 are located (the plane where the navel is located, perpendicular to the horizontal plane) is perpendicular to the line connecting points 5 and 6.

[0036] In addition, corresponding to the data supplement of the differential signal obtained by differentiating the two signals of the above electrode pair, the present application can also perform new differential processing on two signals that do not belong to the same electrode pair in terms of details to obtain signal characteristics of a new direction (or dimension) and continue to enhance the corresponding vector signal content and quality.

[0037] In this regard, as an exemplary embodiment, the uterine vector electromyographic signal may further include an auxiliary differential signal obtained by differentiating two signals that do not belong to the same electrode pair (or belong to different electrode pairs).

[0038] As an example, differential processing may be performed on the two signals measured by the first electrode (belonging to the first electrode pair) and the second electrode (belonging to the second electrode pair) to obtain a new differential signal.

[0039] It is understandable that under the new differential signal configuration mechanism here, more dimensional signal features can be flexibly formed to promote subsequent more delicate and accurate premature birth prediction processing.

[0040] Step S102: Inputting the uterine vector electromyographic signal into a pre-configured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to a preset electrode arrangement, and the prediction model is used to predict the corresponding premature birth based on the uterine vector electromyographic signal input to the model; It can be understood that, as previously introduced, under the preset electrode arrangement scheme specially designed for this application, this application has pre-trained a corresponding prediction model, which can also be called a premature birth prediction model. Its training samples, that is, the labeled sample uterine vector electromyographic signals, are obtained by configuring the preset electrode scheme.

[0041] Among them, the sample uterine vector electromyographic signal is usually collected by configuring the corresponding measuring electrodes under the preset electrode scheme, and in some cases, it may also be extracted from some ready-made signals, that is, it is not ruled out that the sample uterine vector electromyographic signal corresponding to the individual measuring electrode itself has been collected in advance in response to some signal acquisition requirements, or, in some cases, it may also be a signal processed on the basis of the obtained uterine vector electromyographic signal to further increase the sample size, and even, in some cases, it may be a directly constructed uterine vector electromyographic signal. This setting is convenient for obtaining corresponding sample uterine vector electromyographic signals for special groups or special symptoms.

[0042] The labeling of sample uterine vector electromyographic signals, that is, the labeling of the theoretical model prediction results - the premature birth prediction results, can usually be done manually, or it can be completed by corresponding automated labeling tools. The automated labeling tools require the corresponding automated labeling logic to be pre-configured.

[0043] In this way, after obtaining the labeled sample uterine vector electromyographic signals, the initial model can be trained under the corresponding model training scheme and the specific loss function configuration, and the loss function can be quantified with the help of the model prediction results and sample labels. The weight parameters involved in the relevant model structure are continuously optimized and recalculated in combination with the loss function calculation results under the back propagation algorithm. In this way, when the model training requirements such as training time, number of training times, and prediction accuracy are met, the training can be completed and a prediction model that can be put into practical use can be obtained.

[0044] It can also be seen that the present application may also involve data processing in the preliminary model training phase. In this regard, as an exemplary embodiment, the method of the present application may also include: obtaining a sample uterine vector electromyographic signal corresponding to a preset electrode arrangement scheme; Configure corresponding annotations for the sample uterine vector electromyographic signals; The prediction model is trained using labeled sample uterine vector electromyographic signals corresponding to the preset electrode layout scheme.

[0045] In addition, it is understandable that, in some cases, the differential signal in the uterine vector electromyographic signal can also be obtained by performing differential analysis on the basis of inputting 6 basic uterine electromyographic signals by the prediction model.

[0046] At this time, after obtaining the uterine vector electromyographic signal of the user whose premature birth situation needs to be predicted, it can be input into the prediction model to carry out the corresponding premature birth prediction process.

[0047] In addition, this application also provides a specific implementation plan for the model structure involved in the prediction model.

[0048] Correspondingly, as an exemplary embodiment, the prediction model of the present application may specifically include a signal encoding module and a signal decoding module, and the signal encoding module and the signal decoding module are configured by a BiLSTM model with separately set parameters.

[0049] In other words, the prediction model of this application is designed with an encoding and decoding structure based on the BiLSTM model. The BiLSTM model, or Bi-directional Long Short-Term Memory (BiLSTM), is a relatively mature existing model with the following main advantages: 1. Considering both past and future information: BiLSTM runs two LSTM layers simultaneously, one processing data in forward order and the other in reverse order. This allows the output of each time step to include information from both the previous and next time steps, thereby better capturing dynamic features and contextual relationships in sequence data. 2. Improved information utilization: Through forward and backward propagation, BiLSTM can more comprehensively utilize the forward and backward feature information of the sequence, significantly improving the modeling capabilities of sequence data. This bidirectional structure enables the model to better understand the semantics and contextual relationships in sequence data. 3. Applicable to complex sequence data: BiLSTM is particularly suitable for time-varying and complex data. It can effectively process nonlinear and non-stationary signals and capture dynamic features in sequence data.

[0050] More specifically, the signal encoding module may include the following processing during its operation: 1) Extract features from each channel of the uterine vector electromyographic signal input to the model and output the corresponding initial one-dimensional time series features; In which, corresponding to the previous embodiment, the uterine vector electromyographic signal input to the model can specifically be a 6-channel basic uterine electromyographic signal, or can be a 6-channel basic uterine electromyographic signal and a 3-channel differential signal.

[0051] This corresponds to the feature conversion link on the input side of the model, which converts the initial signal into the corresponding initial one-dimensional time series feature. In layman's terms, it is to convert the initial signal into the corresponding coded output through signal coding processing.

[0052] 2) Vertically concatenate each initial one-dimensional time series feature to form a two-dimensional feature matrix; It can be understood that after obtaining the initial one-dimensional time series features of each channel, vertical splicing or channel splicing can be performed to form a two-dimensional feature matrix with the specifications of the number of channels × encoder input and output size (such as 6×S, 9×S).

[0053] 3) Use a multi-layer convolutional neural network to perform feature transformation and dimension conversion, and output the target one-dimensional time series features.

[0054] After the two-dimensional feature matrix is ​​obtained by splicing, convolution processing can be continued to achieve feature transformation and dimension conversion, thereby outputting a one-dimensional feature matrix with a specification of 1× encoder output size, that is, the target one-dimensional time series feature.

[0055] Specifically, as an example, the signal encoding process of the signal encoding module, i.e., the encoder, is configured as follows: The number of hidden layer units of the BiLSTM network is set to 128, and the number of layers is set to 2; the input size corresponding to each signal processed separately is set to 1; the output size of the BiLSTM is set to S, which is consistent with the sequence length of the signal segment.

[0056] The BiLSTM network model is used to independently encode each channel in the input signal segment; the encoded output of each channel is spliced ​​by channel to form a 9×S two-dimensional feature matrix; then a 1×1 convolutional neural network is used to convolve the two-dimensional feature matrix according to the channel dimension, and the dimension is reduced to output a 1×S one-dimensional feature matrix.

[0057] On the other hand, the signal decoding module may include the following processing contents during its operation: The target one-dimensional time series features are used to predict premature birth. Premature birth or pregnancy outcomes include three types: Type 1, premature birth (delivery before 37 weeks of gestation); Type 2, full-term and normal delivery; The third type is full-term, but induced labor or cesarean section.

[0058] It can be understood that the signal decoding module directly decodes the target one-dimensional time series features transmitted by the previous signal encoding module to fuse them into the final model output result, that is, the corresponding label of the specific premature birth type.

[0059] The first type of tag can be specifically configured as 0, the second type of tag can be specifically configured as 1, and the third type of tag can be specifically configured as 2.

[0060] Specifically, as an example, the number of hidden layer units of the signal decoding module, i.e., the decoder, is set to 64, the number of layers is set to 1, and the output size is set to 2. The decoded output vector is first calculated by the Softmax function, and then the maximum value is set to 1 and the others are set to zero to form a binary output vector corresponding to the above three labels.

[0061] It can be understood that this application is based on the above-mentioned encoding and decoding structure design, or in other words, this application is based on the above-mentioned encoding-decoding signal processing method. Without involving or requiring a complex filtering process, the powerful feature extraction capability of the neural network can be directly used to perform subtle feature extraction and multi-dimensional conversion of uterine electromyographic signals and / or differential signals. Among them, the subtle feature extraction can include more nonlinear signal features, and the multi-dimensional conversion is conducive to fully identifying the time domain and frequency domain characteristics. These measures can effectively improve the accuracy of the current method for predicting premature birth.

[0062] In addition, this application also provides a specific implementation plan for acquiring sample uterine vector electromyographic signals.

[0063] Specifically, as an exemplary embodiment, obtaining a sample uterine vector electromyographic signal corresponding to a preset electrode arrangement scheme may include: 1) Obtain initial sample uterine vector electromyographic signals from M cases with complete pregnancy outcomes corresponding to the preset electrode arrangement scheme, where the initial sample uterine vector electromyographic signals are at least 30 minutes long, the signal sampling rate is no less than 360 Hz, and M is greater than 200; Among them, for cases, or initial sample uterine vector electromyographic signals, they can be specifically limited to no more than 26 weeks of gestation, corresponding to the actual situation where there is a need for premature birth prediction.

[0064] The specific sample content can involve three types of premature births or pregnancy outcomes, as set above, and indicated by labels 0, 1, and 2.

[0065] 2) The initial sample uterine vector electromyographic signal is sliced ​​into segments of 1 to 3 seconds in length to form N groups of signal segments, where N is greater than 600 and each signal segment contains 9 sequence data of length S, where S is the standard sequence length of the model input (or signal encoding module input) and is greater than 360 sampling points; It can be understood that the specific slicing process here is conducive to standardized and more delicate sample configuration effects in practical applications.

[0066] 3) Merge the N groups of signal fragments from each of the M cases to form M×N groups of signal fragments, which are then randomly shuffled to form the corresponding batch of sequence data, where the batch size is 16-128 groups (signal fragments).

[0067] It can be understood that the randomly generated sequence data of the same batch size or different batch sizes can be used as the final, highly diversified training samples, and sent to the signal encoding module in batches to promote specific model training processing.

[0068] In addition, as an exemplary embodiment, a loss function that can be used in the training process may be a cross entropy function, which is an existing loss function.

[0069] In this case, the labeled data mentioned above can be one-hot encoded to form a corresponding vector for the cross entropy function to use to calculate the loss error between the model output and the true label value.

[0070] Step S103: extracting the premature birth prediction result output by the prediction model.

[0071] It is easy to understand that when the prediction model completes the premature birth prediction process and outputs the corresponding premature birth prediction result, the premature birth prediction result can be extracted to complete a premature birth prediction task for the current user.

[0072] The premature birth prediction result indicates the specific type of premature birth that is suitable for the situation, and further provides the corresponding probability.

[0073] At the same time, corresponding to data application needs, further data application links may also be involved. For example, local storage, remote storage, result display, result printing, result push or result analysis (disease diagnosis, etc.) can be performed based on the premature birth prediction results.

[0074] It is understandable that specific data application processing can be adaptively adjusted according to pre-configured and real-time configured data application strategies, which is relatively flexible in actual situations.

[0075] Finally, regarding the above scheme content, in general, for the goal of predicting premature birth based on uterine vector electromyographic signals, this application uses a specially designed novel electrode arrangement scheme. The collected uterine vector electromyographic signals are different from traditional uterine surface electromyographic signals, and can more comprehensively reflect the spatial characteristics of uterine contraction activities and contain more signal features related to premature birth. Therefore, under the condition of fewer measuring electrodes, a higher prediction accuracy for premature birth can be obtained, which has clinical application value.

[0076] The above is an introduction to the premature birth prediction method based on uterine vector electromyographic signals provided by this application. In order to facilitate better implementation of the premature birth prediction method based on uterine vector electromyographic signals provided by this application, this application also provides a premature birth prediction device based on uterine vector electromyographic signals from the perspective of functional modules.

[0077] See Figure 3 , Figure 3This is a schematic diagram of the structure of a premature birth prediction device based on uterine vector electromyographic signals in this application. In this application, the premature birth prediction device 300 based on uterine vector electromyographic signals may specifically include the following structure: An acquisition unit 301 is configured to acquire uterine vector myoelectric signals collected from the mother using a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs. The uterine vector myoelectric signals specifically include uterine myoelectric signals in three orthogonal directions corresponding to the three electrode pairs, and also include a differential signal obtained by differentiating two signals from each electrode pair. Prediction unit 302, configured to input the uterine vector electromyographic signal into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to a preset electrode arrangement, and the prediction model is configured to predict the corresponding premature birth based on the uterine vector electromyographic signal input to the model; The extraction unit 303 is used to extract the premature birth prediction result output by the prediction model.

[0078] As an exemplary embodiment, for three electrode pairs, there are: The first electrode pair consists of the first electrode and the fourth electrode. The first electrode is placed at the upper middle of the mother's abdominal wall, and the fourth electrode is placed at the lower middle of the mother's back. The second electrode pair is composed of the second electrode and the third electrode. The second electrode is arranged at the upper middle of the mother's waist and back, and the third electrode is arranged at the lower middle of the mother's abdominal wall. The connecting line of the first electrode pair and the connecting line of the second electrode pair form a perpendicular relationship. The third electrode pair consists of the fifth electrode and the sixth electrode. The fifth electrode and the sixth electrode are along the horizontal line of the mother's waist and are located on the left and right sides of the abdominal wall respectively. The connecting line of the third electrode pair forms an orthogonal relationship with the connecting line of the first electrode pair and the connecting line of the second electrode pair respectively.

[0079] As another exemplary embodiment, the uterine vector electromyographic signal also includes an auxiliary differential signal obtained by differentiating two signals that do not belong to the same electrode pair. As another exemplary embodiment, the prediction model includes a signal encoding module and a signal decoding module, and the signal encoding module and the signal decoding module are configured by a BiLSTM model with independently set parameters; The signal coding module includes the following processing during its operation: Extract features from each channel of the uterine vector electromyographic signal input to the model and output the corresponding initial one-dimensional time series features; Each initial one-dimensional time series feature is vertically spliced ​​to form a two-dimensional feature matrix; A multi-layer convolutional neural network is used to perform feature transformation and dimension conversion, outputting the target one-dimensional time series features; The signal decoding module includes the following processing during its operation: The target one-dimensional time series features are used to predict premature births. There are three types of premature births, corresponding to: Type 1, premature birth; Type 2, full-term and normal delivery; The third type is full-term, but induced labor or cesarean section.

[0080] As another exemplary embodiment, the apparatus further includes a training unit 304, configured to: obtaining a sample uterine vector electromyographic signal corresponding to a preset electrode arrangement scheme; Configure corresponding annotations for the sample uterine vector electromyographic signals; The prediction model is trained using labeled sample uterine vector electromyographic signals corresponding to the preset electrode layout scheme.

[0081] As another exemplary embodiment, the training unit 304 is specifically configured to: Obtain initial sample uterine vector electromyographic signals of M cases with complete pregnancy outcomes corresponding to the preset electrode arrangement scheme, wherein the initial sample uterine vector electromyographic signals are at least 30 minutes long, the signal sampling rate is not less than 360 Hz, and M is greater than 200; The initial sample uterine vector electromyographic signal is sliced ​​into time segments of 1 to 3 seconds to form N groups of signal segments, where N is greater than 600 and each signal segment contains 9 sequence data of length S, where S is the standard sequence length of the model input and S is greater than 360 sampling points. The N groups of signal fragments from each of the M cases were merged to form M×N groups of signal fragments, which were then randomly shuffled to form the corresponding batch of sequence data, where the batch size was 16-128 groups.

[0082] As another exemplary embodiment, the loss function in the training process specifically adopts the cross entropy function, and the labeled data is one-hot encoded to form a corresponding vector for use in the cross entropy function calculation.

[0083] This application also provides a processing device from the perspective of hardware structure, see Figure 4 , Figure 4 The schematic diagram of the structure of the processing device of the present application is shown. Specifically, the processing device of the present application may include a processor 401, a memory 402 and an input / output device 403. The processor 401 is used to execute the computer program stored in the memory 402 to implement the following Figure 1 The steps of the method for predicting premature birth based on uterine vector electromyographic signals in the corresponding embodiment; or, the processor 401 is used to execute the computer program stored in the memory 402 to implement the following Figure 3 The memory 402 is used to store the functions of each unit in the embodiment, and the processor 401 executes the above Figure 1 The computer program required for the premature birth prediction method based on uterine vector electromyographic signals in the corresponding embodiment.

[0084] For example, the computer program may be divided into one or more modules / units, one or more of which are stored in memory 402 and executed by processor 401 to implement the present application. One or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a computer device.

[0085] The processing device may include, but is not limited to, a processor 401, a memory 402, and an input / output device 403. Those skilled in the art will appreciate that the illustrations are merely examples of processing devices and do not limit the processing device. The processing device may include more or fewer components than shown, or a combination of certain components, or different components. For example, the processing device may also include a network access device, a bus, etc., and the processor 401, the memory 402, the input / output device 403, etc. are connected via the bus.

[0086] Processor 401 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the processing device and connects various parts of the entire device using various interfaces and lines.

[0087] Memory 402 can be used to store computer programs and / or modules. Processor 401 implements various functions of the computer device by running or executing computer programs and / or modules stored in memory 402 and accessing data stored in memory 402. Memory 402 may primarily include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function, and the data storage area may store data generated based on the use of the processing device. Furthermore, memory may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0088] When the processor 401 is used to execute the computer program stored in the memory 402, it can specifically implement the following functions: Obtaining uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs, the uterine vector myoelectric signals specifically including uterine myoelectric signals in three orthogonal directions corresponding to the three electrode pairs, and the uterine vector myoelectric signals also including a differential signal obtained by differentiating two signals from each electrode pair; Inputting the uterine vector electromyographic signals into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to a preset electrode arrangement, and the prediction model is used to predict the corresponding premature birth based on the uterine vector electromyographic signals input to the model; Extract the premature birth prediction results output by the prediction model.

[0089] Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working process of the above-described premature birth prediction device based on uterine vector electromyographic signal, processing equipment and corresponding units can refer to the following. Figure 1 The description of the premature birth prediction method based on uterine vector electromyographic signals in the corresponding embodiment will not be repeated here.

[0090] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be accomplished by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0091] To this end, the present application provides a computer-readable storage medium, which stores a plurality of instructions, which can be loaded by a processor to execute the present application as follows: Figure 1 The steps of the method for predicting premature birth based on uterine vector electromyographic signals in the corresponding embodiment, the specific operations can be referred to as follows Figure 1 The description of the premature birth prediction method based on uterine vector electromyographic signals in the corresponding embodiment will not be repeated here.

[0092] The computer-readable storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0093] Due to the instructions stored in the computer readable storage medium, the present application can be executed as follows: Figure 1 The steps of the method for predicting premature birth based on uterine vector electromyographic signals in the corresponding embodiment can thus be realized as follows: Figure 1 The beneficial effects that can be achieved by the premature birth prediction method based on uterine vector electromyographic signals in the corresponding embodiment are detailed in the previous description and will not be repeated here.

[0094] The above is a detailed introduction to the premature birth prediction method, device, processing equipment and computer-readable storage medium based on uterine vector electromyographic signals provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the core idea of ​​the present application; at the same time, for technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting premature birth based on uterine vector electromyographic signals, characterized in that: The method comprises: Obtaining uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs, the uterine vector myoelectric signals specifically including uterine myoelectric signals corresponding to three of the electrode pairs in three orthogonal directions, and the uterine vector myoelectric signals also including a differential signal obtained by differentiating two signals from each electrode pair; Inputting the uterine vector electromyographic signal into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector electromyographic signals corresponding to the preset electrode arrangement, and the prediction model is used to predict corresponding premature birth based on the uterine vector electromyographic signal input into the model; Extract the premature birth prediction result output by the prediction model.

2. The method according to claim 1, characterized in that For the three electrode pairs, we have: The first electrode pair consists of a first electrode and a fourth electrode, wherein the first electrode is arranged at the upper middle of the maternal abdominal wall, and the fourth electrode is arranged at the lower middle of the maternal back; The second electrode pair is composed of a second electrode and a third electrode, wherein the second electrode is arranged at the upper middle of the mother's waist and back, and the third electrode is arranged at the lower middle of the mother's abdominal wall, and the line connecting the first electrode pair and the line connecting the second electrode pair form a perpendicular relationship; The third electrode pair is composed of the fifth electrode and the sixth electrode. The fifth electrode and the sixth electrode are along the horizontal line of the mother's waist and are located on the left and right sides of the abdominal wall respectively. The connecting line of the third electrode pair forms an orthogonal relationship with the connecting line of the first electrode pair and the connecting line of the second electrode pair respectively.

3. The method according to claim 2, characterized in that The uterine vector myoelectric signal also includes an auxiliary differential signal obtained by differentiating two signals that do not belong to the same electrode pair.

4. The method according to claim 1, wherein The prediction model includes a signal encoding module and a signal decoding module, and the signal encoding module and the signal decoding module are configured by a BiLSTM model with individually set parameters; The signal encoding module includes the following processing contents during operation: Extracting features from each channel of the uterine vector electromyographic signal input into the model, and outputting corresponding initial one-dimensional time series features; Vertically splicing each of the initial one-dimensional time series features to form a two-dimensional feature matrix; A multi-layer convolutional neural network is used to perform feature transformation and dimension conversion, outputting the target one-dimensional time series features; The signal decoding module includes the following processing contents during operation: The target one-dimensional time series feature is used to predict the premature birth situation. The premature birth situation specifically includes three types, corresponding to: Type 1, premature birth; Type 2, full-term and normal delivery; The third type is full-term, but induced labor or cesarean section.

5. The method according to claim 1, wherein The method further comprises: Acquiring the sample uterine vector electromyographic signal corresponding to the preset electrode arrangement scheme; Configuring corresponding annotations for the sample uterine vector electromyographic signals; The prediction model is trained using labeled sample uterine vector electromyographic signals corresponding to the preset electrode arrangement scheme.

6. The method according to claim 5, characterized in that The obtaining of the sample uterine vector electromyographic signal corresponding to the preset electrode arrangement scheme includes: Obtaining initial sample uterine vector electromyographic signals of M cases with complete pregnancy outcomes corresponding to the preset electrode arrangement scheme, wherein the initial sample uterine vector electromyographic signals are no less than 30 minutes, the signal sampling rate is no less than 360 Hz, and M is greater than 200; The initial sample uterine vector electromyographic signal is sliced ​​into time segments of 1 to 3 seconds to form N groups of signal segments, where N is greater than 600, and each signal segment contains 9 sequence data of length S, where S is the standard sequence length of the model input, and S is greater than 360 sampling points; The N groups of signal segments of each of the M cases are merged to form M×N groups of signal segments, which are then randomly shuffled to form corresponding batches of sequence data, wherein the batch size is 16-128 groups.

7. The method according to claim 5, characterized in that The loss function in the training process specifically adopts the cross entropy function, and the labeled data is one-hot encoded to form a corresponding vector for use in the calculation of the cross entropy function.

8. A device for predicting premature birth based on uterine vector electromyographic signals, characterized in that: The device comprises: an acquisition unit, configured to acquire uterine vector myoelectric signals collected from the mother under a preset electrode arrangement scheme, wherein the preset electrode arrangement scheme involves six electrodes, specifically configured as electrode pairs arranged orthogonally in pairs, the uterine vector myoelectric signals specifically including uterine myoelectric signals corresponding to three of the electrode pairs and located in three orthogonal directions, and the uterine vector myoelectric signals also including a differential signal obtained by differentiating two signals from each electrode pair; a prediction unit, configured to input the uterine vector myoelectric signal into a preconfigured prediction model, wherein the prediction model is specifically trained using labeled sample uterine vector myoelectric signals corresponding to the preset electrode arrangement, and the prediction model is configured to predict a corresponding premature birth based on the uterine vector myoelectric signal input into the model; The extraction unit is used to extract the premature birth prediction result output by the prediction model.

9. A processing device, characterized in that The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method according to any one of claims 1 to 7 is executed.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 7.

Images