Premature delivery prediction method and system for pregnant woman based on uterine electromyographic signals

Abstract

The invention relates to the field of medicine, in particular to a pregnant woman premature delivery prediction system based on uterus electromyographic signals. Body surface electrodes are used for detecting the uterus electromyographic signals of the abdomen of a puerpera; the analog / digital converter is electrically connected with the body surface electrode, and the analog / digital converter receives electric signals of the body surface electrode and converts the electric signals into digital signals. And the memory is arranged on the microprocessor and is used for storing data. The microprocessor is electrically connected with the analog / digital converter, receives data of the analog / digital converter and stores the data into the memory; a computer program is stored on the memory, and the following method for predicting the premature delivery of the pregnant woman based on the uterine electromyographic signals is realized when the computer program is executed by the processing device.

Description

Technical Field

[0001] The present invention relates to the field of medicine, and particularly to a preterm birth prediction system for pregnant women based on uterine electromyogram signals. Background Art

[0002] Preterm birth is a disorder of abnormal pregnancy duration and is one of the most common adverse pregnancy outcomes; in China, preterm birth is defined as delivery occurring between 28 completed weeks and less than 37 completed weeks of gestation, and is divided into spontaneous preterm birth and medical preterm birth.

[0003] The disadvantages of the prior art are as follows: The current preterm birth prediction methods have certain limitations, mostly relying on clinical symptoms, medical history and some biomarkers, but the accuracy and timeliness need to be improved. The evaluation of clinical symptoms and medical history is highly subjective, and different doctors may have different judgments. The detection of some biomarkers has problems such as invasiveness, high cost, and complex operation, and is not suitable for large-scale screening. Moreover, the existing methods are difficult to predict the risk of preterm birth early and accurately, which is not conducive to taking timely intervention measures to reduce the incidence of preterm birth and its adverse effects. Summary of the Invention

[0004] The present invention provides a preterm birth prediction system for pregnant women based on uterine electromyogram signals to solve the above-mentioned disadvantages.

[0005] In order to achieve the above-mentioned invention purpose, the technical solution adopted by the present invention is as follows:

[0006] A preterm birth prediction method for pregnant women based on uterine electromyogram signals, comprising the following steps:

[0007] Step S100, arranging detection electrodes on the abdomen of the pregnant woman with the navel as the center to obtain uterine electromyogram signals;

[0008] Step S200, setting a frequency f s , and obtaining the time series signal of the uterine electromyogram signal by means of the frequency f s ;

[0009] Step S300, decomposing the frequency components of the time series signal containing 0 - 5 Hz into different frequency components y(t) through short-time Fourier transform;

[0010] Step S400, calculating the approximate entropy feature and sample entropy feature

[0011] where ApEn represents the calculation of approximate entropy, SampEn represents the calculation of sample entropy, the integer m = 2, the real number r = 0.2 * std, and std represents the standard deviation of the frequency component sequence.

[0012] Further, in step S500, the recorded time t is utilized r to adjust the approximate entropy features and sample entropy features respectively;

[0013]

[0014] In step S600, principal component analysis technology is applied to all the obtained features to select the optimal number of principal components to obtain features. Principal component analysis technology is applied to all the obtained features to select the optimal number of principal components to obtain features.

[0015] Further, step S600 specifically includes the following:

[0016] (1) First, standardize all the entropy feature data so that each feature has zero mean and unit variance;

[0017] (2) Calculate the covariance matrix based on the standardized data;

[0018] (3) Solve the eigenvalues and corresponding eigenvectors of the covariance matrix;

[0019] (4) Select the principal components whose cumulative contribution rate reaches a certain threshold.

[0020] Further, in step S700, a preterm birth prediction model is constructed based on deep machine learning. The features, features features are input into the preterm birth prediction model, and the preterm birth prediction model outputs the preterm birth probability.

[0021] Further, sample data for training is prepared. The sample data includes features, features features and information on whether it is a preterm birth.

[0022] Further, an electrode placement grid is constructed with the navel as the center. The shape of the grid is 1 cm * 1 cm, and the detection electrodes are placed at the grid points of the framework grid.

[0023] On the other hand, a pregnant woman preterm birth prediction system based on uterine electromyogram signals

[0024] The body surface electrodes are used to detect the uterine electromyogram signals of the parturient's abdomen;

[0025] The analog / digital converter is electrically connected to the body surface electrodes. The analog / digital converter receives the electrical signals of the body surface electrodes and converts the electrical signals into digital signals;

[0026] The memory is installed on the microprocessor, and the memory is used to store data;

[0027] The microprocessor is electrically connected to the analog / digital converter. The microprocessor receives the data from the analog / digital converter and stores it in the memory; a computer program is stored in the memory. When the computer program is executed by the processing device, it realizes the method for predicting premature birth of pregnant women based on uterine electromyogram signals as described in any one of claims 1 to 6.

[0028] Furthermore, the screen is electrically connected to the microprocessor, and the screen is used to display the prediction result.

[0029] Furthermore, the screen is installed on the surface of the housing, and the memory, the analog / digital converter, and the microprocessor are located inside the housing;

[0030] A storage bin is provided on the back of the housing. The top of the storage bin is a cover plate that can be opened, and the body surface electrodes are stored in the storage bin.

[0031] Furthermore, the button, the power connection terminal of the microprocessor, and the positive and negative power supplies form a series circuit, and the button is used to control the power on and off of the microprocessor.

[0032] The beneficial effects of the present invention are:

[0033] The present invention extends the analysis range of EHG signals from the time domain to the time-frequency domain, constructs an entropy feature system in the time-frequency domain, and further optimizes the time-frequency domain entropy features according to the principle of biological evolution to achieve an accurate description of the pregnancy process. At the same time, the present invention uses principal component analysis technology to effectively process the noise in the EHG signals. The experimental results show that the performance of the proposed new entropy features in premature birth prediction has been significantly improved. Description of the Drawings

[0034] Figure 1 It is a structural schematic diagram of the prediction system;

[0035] Figure 2 It is a logic block diagram of the electrical components of the prediction system.

[0036] Housing 10, screen 20, storage bin 40, button 50, memory 60, analog / digital converter 80, body surface electrode 90, microprocessor 100. Detailed Embodiments

[0037] The following further illustrates the detailed embodiments of the present invention with reference to the drawings. The same components are denoted by the same reference numerals.

[0038] It should be noted that the terms "front", "rear", "left", "right", "up" and "down" used in the following description refer to the directions in the accompanying drawings, and the terms "inner" and "outer" refer to the directions towards or away from the geometric center of a specific component respectively.

[0039] In order to make the content of the present invention easier to be clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0040] On the one hand, referring to Figure 1 , Figure 2 , a premature birth prediction system for pregnant women based on uterine myoelectric signals includes a housing 10, a screen 20, a storage bin 40, a button 50, a memory 60, an analog / digital converter 80, a body surface electrode 90, and a microprocessor 100;

[0041] The body surface electrode 90 is used to detect the uterine myoelectric signals on the abdomen of the parturient.

[0042] The analog / digital converter 80 is electrically connected to the body surface electrode 90. The analog / digital converter 80 receives the electrical signal of the body surface electrode 90 and converts the electrical signal into a digital signal.

[0043] The memory 60 is installed on the microprocessor 100, and the memory 60 is used to store data.

[0044] The microprocessor 100 is electrically connected to the analog / digital converter 80. The microprocessor 100 receives the data of the analog / digital converter 80 and stores it in the memory 60; a computer program is stored on the memory 60, and when the computer program is executed by the processing device, the following premature birth prediction method for pregnant women based on uterine myoelectric signals is realized.

[0045] Furthermore, the screen 20 is electrically connected to the microprocessor 100, and the screen 20 is used to display the prediction result.

[0046] Furthermore, the screen 20 is installed on the surface of the housing 10, and the memory 60, the analog / digital converter 80, and the microprocessor 100 are located inside the housing 10;

[0047] A storage bin 40 is arranged on the back of the housing 10. The top of the storage bin 40 is a cover plate that can be opened, and the body surface electrode 90 is stored in the storage bin 40; when in use, the body surface electrode 90 is taken out from the storage bin 40.

[0048] Furthermore, the button 50, the power connection terminal of the microprocessor 100, and the positive and negative power supplies form a series circuit, and the button 50 is used to control the power on and off of the microprocessor 100.

[0049] On the other hand, a method for predicting preterm birth in pregnant women based on uterine electromyogram signals includes the following steps:

[0050] Step S100: Set detection electrodes on the abdomen of the pregnant woman with the navel as the center to obtain uterine electromyogram signals. A grid for electrode placement is constructed with the navel as the center. The shape of the grid is 1 cm * 1 cm, and the detection electrodes are placed at the grid points of the constructed grid.

[0051] Step S200: Set the frequency fs, and collect the uterine electromyogram signals to obtain a time series signal; the frequency fs is the number of times of sampling the continuous signal per unit time.

[0052] Step S300: Decompose the time series signal containing frequencies from 0 to 5 Hz into different frequency components y(t) through short-time Fourier transform;

[0053] Step S400: Calculate the approximate entropy feature and sample entropy feature at the corresponding frequency according to the frequency component y(t) and sample entropy feature

[0054] where ApEn represents the calculation of approximate entropy, SampEn represents the calculation of sample entropy, the integer m = 2, and the real number r = 0.2 * std, where std represents the standard deviation of the frequency component sequence.

[0055] Step S500: Use the recorded time t r to adjust the approximate entropy feature and sample entropy feature respectively;

[0056] where 37 weeks is the dividing line between preterm birth and non-preterm birth;

[0057] Step S600: Apply the principal component analysis technique to all the features obtained, select the optimal number of principal components to obtain features, apply the principal component analysis technique to all the features obtained, and select the optimal number of principal components to obtain features; this step helps the new features to remove the influence of different frequency noises in the EHG signal and at the same time lock the information related to uterine contraction activities in the EHG signal. The core principle of the principal component analysis technique is to linearly transform the original data and convert it into a set of mutually orthogonal principal components. In the present invention, its main purpose is to perform principal component analysis on all the time-frequency domain entropy features (including the improved approximate entropy feature and sample entropy feature ) Analyze and process. These entropy features reflect the characteristics of the electrohysterogram (EHG) to a certain extent, but there may be information redundancy and interference from noise. The PCA technique aims to extract the most crucial and representative information from these complex features, reduce the dimensionality of the data, and remove the influence of noise, thereby improving the accuracy and efficiency of the subsequent preterm birth prediction model.

[0058] Specifically, (1) First, standardize all the entropy feature data so that each feature has a zero mean and unit variance. This step helps ensure that the weights of different features are relatively fair in subsequent calculations and avoids some features having too much influence on the results due to differences in feature dimensions.

[0059] (2) Calculate the covariance matrix based on the standardized data. The covariance matrix reflects the correlation between each feature. The elements on the diagonal are the variances of each feature, and the off-diagonal elements represent the covariance between features. By analyzing the covariance matrix, the linear relationship between features can be understood.

[0060] (3) Solve the eigenvalues and corresponding eigenvectors of the covariance matrix. The eigenvalues represent the amount of information contained in each principal component, and the eigenvectors determine the direction of the principal components. Usually, the eigenvectors are sorted in descending order of eigenvalues.

[0061] (4) Select the optimal number of principal components according to the set criteria. A common method is to select the first few principal components whose cumulative contribution rate reaches a certain threshold (such as 85% or 90%). In the present invention, after applying the PCA technique to all features and features, the optimal number of principal components is selected respectively to obtain features and features. These selected principal components can retain the information of the original data to the greatest extent, reduce the dimensionality of the data, and reduce the interference of noise on subsequent analysis.

[0062] Step S700: Construct a preterm birth prediction model based on deep machine learning, features, features features are input into the preterm birth prediction model, and the preterm birth prediction model outputs the preterm birth probability.

[0063] Specifically, prepare the sample data for training. The sample data includes features, features features and the information on whether it is a preterm birth; preterm birth is represented by 1 or non-preterm birth is represented by 2.

[0064] The selected machine learning model is trained using sample data. During the training process, the model learns the internal relationship pattern between features and preterm birth based on the feature values and the corresponding preterm or non-preterm birth labels in the sample data.

[0065] In summary, the present invention extends the analysis of EHG signals from the time domain to the time-frequency domain, constructs entropy features within the time-frequency domain, and further optimizes the time-frequency domain entropy features according to the characteristics of biological evolution to accurately depict the pregnancy process. At the same time, the principal component analysis technique is used to effectively process the noise in the EHG signals. The effect of the proposed new entropy features in preterm birth prediction has been significantly improved.

[0066] The above are only the preferred embodiments of the present invention for patent, and are not intended to limit the present invention for patent. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention for patent shall be included within the protection scope of the present invention for patent.

Claims

1. A system for predicting premature birth in pregnant women based on uterine myoelectric signals, characterized in that: include: The body surface electrodes are used to detect uterine myoelectric signals in the abdomen of the parturient; The analog / digital converter is electrically connected to the body surface electrode, and receives the electrical signal from the body surface electrode and converts the electrical signal into a digital signal; The memory is installed on the microprocessor, and the memory is used to store data; The microprocessor is electrically connected to the analog / digital converter, and the microprocessor receives data from the analog / digital converter and stores the data in a memory; The storage device stores a computer program, which, when executed by the processing device, implements a method for predicting premature birth of pregnant women based on uterine myoelectric signals.

2. The system for predicting premature birth of pregnant women based on uterine myoelectric signals according to claim 1, characterized in that: The screen is electrically connected to the microprocessor, and the screen is used to display the prediction results.

3. The system for predicting premature birth of pregnant women based on uterine myoelectric signals according to claim 2, characterized in that: A screen is mounted on the surface of the casing, and a storage device, an analog / digital converter, and a microprocessor are located inside the casing; A storage bin is provided on the back of the casing, and a cover plate is provided on the top of the storage bin that can be opened, and the body surface electrodes are stored in the storage bin.

4. The system for predicting premature birth of pregnant women based on uterine myoelectric signals according to claim 3, characterized in that: The button, the power connection terminal of the microprocessor, and the positive and negative electrodes of the power supply form a series circuit, and the button is used to control the power on and off of the microprocessor.

5. A method for predicting premature birth in pregnant women based on uterine myoelectric signals, characterized in that: The following steps are involved: Step S100, setting detection electrodes on the abdomen of the pregnant woman with the navel as the center to obtain uterine myoelectric signals; Step S200: Setting frequency f s , through the frequency f s A time sequence signal of the uterine electromyographic signal is obtained from the uterine electromyographic signal; Step S300, decomposing the frequency of the time series signal including 0-5 Hz into different frequency components y(t) by short-time Fourier transform; Step S400: Calculate the approximate entropy feature at the corresponding frequency according to the frequency component y(t) And sample entropy characteristics Wherein, ApEn represents the calculation of approximate entropy, SampEn represents the calculation of sample entropy, integer m=2, real number r=0.2*std, and std represents the standard deviation of the frequency component sequence.

6. The method for predicting premature birth in pregnant women based on uterine myoelectric signals according to claim 5, characterized in that: Step S500: using the recording time t r The approximate entropy features And sample entropy characteristics Make adjustments; Step S600: The principal component analysis technique is used to select the optimal number of principal components to obtain Features, for all obtained The principal component analysis technique is used to select the optimal number of principal components. feature.

7. The method for predicting premature birth of pregnant women based on uterine myoelectric signals according to claim 6, characterized in that: Step S600 specifically includes the following: (1) First, all entropy feature data are standardized so that each feature has zero mean and unit variance; (2) Calculate the covariance matrix based on the standardized data; (3) Solve the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; (4) Select the principal components whose cumulative contribution rate reaches a certain threshold.

8. The method for predicting premature birth of pregnant women based on uterine myoelectric signals according to claim 6, characterized in that: Step S700: construct a premature birth prediction model based on deep machine learning. Features, Features The features are input into the premature birth prediction model, and the premature birth prediction model outputs the probability of premature birth.

Images