A method for analyzing fetal electrocardio signal dynamics and predicting fetal status based on topographic map analysis

Abstract

The application discloses a kind of based on topographic map analysis fetal electrocardiosignal dynamic analysis and fetal state and prediction method, collect clinical pregnant woman abdominal electric signal, obtain fetal electrocardiosignal by pre-processing and signal separation technique, construct signal dynamic analysis benchmark data set.It designs topographic map analysis algorithm suitable for fetal electrocardiosignal, constructs signal state template, generates the time series change graph of fetal electrocardiosignal state, with the state statistical parameter sequence with annotation as the benchmark data set of fetal state prediction;Construct decision tree model, train the training data set of state statistical parameter sequence, obtain trained model;The test set of state statistical parameter sequence is input into decision tree model, and fetal state is predicted.The present application solves the limitation of existing non-invasive fetal electrocardiosignal monitoring in interpretation, visualization and fetal state prediction, significantly optimizes the readability, interpretability and fetal state prediction performance of visualization result.

Description

Technical Field

[0001] The present invention relates to the technical field of fetal heart monitoring, and specifically to a method for analyzing the dynamics of fetal electrocardiogram (ECG) signal states and predicting adverse fetal states, and is specifically a method for analyzing the dynamics of ECG signal states and predicting states based on topographic map analysis. Background Art

[0002] Fetal heart rate monitoring technology is used to collect fetal heart rate changes, reflecting the fetal heart status. It can be used to identify whether the fetus is in adverse conditions such as hypoxia and asphyxia. The fetal heart rate characteristics displayed by the monitoring can also assist in the early diagnosis of diseases such as fetal acidemia and hypoxic-ischemic encephalopathy. Therefore, the dynamic display of fetal ECG monitoring data can more intuitively reflect the fetal status. The prediction of adverse conditions can facilitate early medical intervention, reduce neonatal morbidity and mortality, and improve delivery outcomes.

[0003] In actual clinical practice, fetal heart rate monitoring methods have many limitations. First, the commonly used fetal heart rate monitoring method is cardiotocography (CTG), but it has the disadvantages of poor interpretability, lack of detailed information on fetal heart rate variability, and unsuitability for long-term monitoring. Other monitoring methods are not widely used due to reasons such as high difficulty of operation, high cost, and invasiveness.

[0004] To overcome the limitations of these methods and provide a safer means of monitoring, a fetal monitoring technology that has recently attracted considerable attention is non-invasive fetal electrocardiography (NI-FECG). This method uses electrodes placed on the pregnant woman's abdomen to collect mixed maternal and fetal ECG signals. Signal processing and feature recognition methods are then used to isolate the electrical signals of fetal cardiac activity. NI-FECG offers the advantages of being non-invasive, non-destructive, and radiation-free. However, it also presents technical challenges such as low signal-to-noise ratio and susceptibility to interference from noise and maternal ECG signals. To address the limitations of NI-FECG, current research has focused on extending adult ECG signal processing methods to fetal ECG signal extraction and feature recognition, overcoming some of these challenges. However, existing approaches primarily focus on optimizing signal processing results, without providing rational solutions for signal readability, visualization, or the correlation between signals and birth outcomes. Summary of the Invention

[0005] The purpose of the present invention is to propose a method for dynamic analysis of fetal ECG signals and prediction of fetal status based on topographic map analysis, so as to address the limitations of existing non-invasive fetal ECG monitoring in terms of result interpretation, visualization and fetal status prediction indicators, improve the display of fetal ECG signals, and enhance the accuracy and reliability of prediction of adverse fetal conditions.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] 1. A method for dynamic analysis of fetal ECG signals and prediction of fetal status based on topographical analysis, such as Figure 1 As shown, the following steps are included:

[0008] In S1, electrodes of the NI-FECG device are placed on the abdomen of pregnant women to collect real mixed abdominal electrical signals and collect birth outcomes. The collected abdominal electrical signals are preprocessed and separated to obtain fetal ECG signals. These abdominal electrical signal data are annotated according to birth outcome categories, and the annotated abdominal fetal ECG signals are constructed into a benchmark dataset for fetal ECG dynamic analysis.

[0009] S2 designs a topographic analysis algorithm suitable for dynamic analysis of fetal ECG signals. This algorithm, based on the inherent characteristics of multi-lead abdominal electrodes and the temporal variation characteristics of fetal ECG signals, extends the application of microstate analysis of EEG signals to abdominal fetal electrical signals with similar characteristics. Using the topographic analysis algorithm, a state template for the fetal ECG signal is constructed, a time series diagram of fetal ECG signal states is generated, and state statistical parameters are calculated. This state statistical parameter sequence corresponds to the annotated data collected in S1, and this sequence is identically annotated to form a benchmark dataset for fetal state prediction.

[0010] S3, building a decision tree model, training the training data of the statistical parameters of the topographic map state with the fetal state annotation collected in S2, and obtaining the trained model parameters.

[0011] S4, extracting the test data of the statistical parameter sequence of the topographic map state with the fetal state annotation collected by S2 and inputting it into the decision tree model, and using the parameters obtained by S3 to predict the fetal state.

[0012] 2. In S1, the same NI-FECG device and electrodes were used to collect abdominal electrical signals from pregnant women at different gestational ages. The signal data covered a variety of abdominal electrical signal samples from pregnant women at different gestational ages and in different physical conditions.

[0013] 3. In S1, after the various abdominal electrical signals are collected and preprocessed and separated to obtain fetal ECG signals, the data are classified and labeled as adverse birth outcomes or favorable birth outcomes based on the degree of neonatal hypoxia and acidosis recorded in the actual birth outcomes.

[0014] 4. Topographic analysis includes state segmentation, fetal ECG data state fitting, and state statistical parameter calculation;

[0015] The state segmentation is used for analyzing energy distribution of the fetal electrocardio signal to generate a template topographic map; the fetal electrocardio data state fitting is used for matching the template topographic map with the electrocardio signal; and the state statistical parameter calculation can obtain statistical characteristics of matching of each template topographic map with the signal, which can potentially represent a specific state mechanism.

[0016] 5. The state segmentation part is used for analyzing energy distribution of the fetal electrocardio signal to generate a template topographic map, and the process comprises:

[0017] The instantaneous distribution potential topographic map of the fetal electrocardio signal is calculated to obtain a local maximum topographic map curve, and a clustering technique is applied to determine an optimal clustering number to obtain the template topographic map.

[0018] 6. The fetal electrocardio data state fitting is used for matching the template topographic map with the electrocardio signal, and the process comprises:

[0019] The topological structure of the fetal electrocardio signal at each time point in the time domain is calculated, and similarity with the template topographic map is calculated, and the template topographic map with high similarity is matched with the signal at the time point.

[0020] 7. The state statistical parameter calculation obtains statistical characteristics of matching of each template topographic map with the signal, and the process for potentially representing a specific state mechanism comprises:

[0021] Statistical distribution characteristics of each template topographic map in the signal are calculated, including a duration, an occurrence rate, a coverage rate and a transfer rate, and the statistical parameters are used for S3 and S4 to predict the fetal state.

[0022] 8. In S3, a binary classification decision tree model is constructed, and a state statistical parameter sequence with a label collected in S2 is divided into a training set and a test set, and the training set is used for S3 to perform model training.

[0023] 9. In S4, the process for testing the model performance comprises:

[0024] The test data set divided in S3 is input into the model trained in S3 to predict the fetal state, and a prediction accuracy, a sensitivity, a precision, a specificity, an F1 score and a ROC curve of a performance index are calculated.

[0025] The present application has the following beneficial effects:

[0026] In the present invention's method for dynamic fetal ECG signal analysis and fetal status prediction based on topographical analysis, mixed abdominal electrical signals from various types of pregnant women are first collected. Fetal ECG signals are obtained through preprocessing and signal separation, and fetal ECG data containing specific categories are labeled according to delivery outcomes to serve as a benchmark dataset for dynamic fetal ECG signal analysis. A topographical analysis algorithm suitable for fetal ECG signals is designed to construct a state template for the fetal ECG signals, generate a time series change graph of the fetal ECG signal states, and obtain a labeled state statistical parameter sequence, which serves as a benchmark dataset for fetal status prediction. A decision tree model is constructed and trained with a training dataset of the collected state statistical parameter sequence to obtain a trained model. A test set of the collected state statistical parameter sequence is input into the decision tree model to predict fetal status. The present invention utilizes topographical analysis and a decision tree model suitable for fetal ECG signals to address the limitations of non-invasive fetal ECG monitoring in terms of result interpretability, visualization, and fetal status prediction indicators. Compared with existing methods, the visualization of monitoring results is significantly improved, making them clearer and easier to understand, increasing interpretability, and significantly optimizing fetal status prediction performance. This invention establishes a new benchmark dataset based on specific fetal status categories and fetal ECG datasets, providing an alternative perspective for fetal status prediction. This dataset includes a statistical description of the dynamic characteristics of fetal ECG signals, links changes in fetal ECG signals to fetal status, and offers a new perspective for fetal status prediction. Therefore, this invention can address the limitations of existing non-invasive fetal ECG monitoring in terms of result interpretability, visualization, and fetal status prediction indicators, improving the readability of fetal ECG monitoring results, the comprehensiveness of fetal status prediction indicators, and the accuracy of prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the implementation of the present invention or the existing technical solutions, a brief introduction to the drawings required for describing the prior art is given below.

[0028] Figure 1 A method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis is shown.

[0029] Figure 2 This is a diagram of a method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis in an example of the invention. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. 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.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] like Figure 2 As shown, the embodiment of the present invention discloses a method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis, comprising the following specific steps:

[0033] S101: The four-conductor electrodes of the NI-FECG device were placed on the abdomen of pregnant women to collect mixed abdominal electrical signals and to collect birth outcomes. A total of 100 cases of abdominal electrocardiogram (ECG) data were collected. The collected abdominal electrical signals were preprocessed and separated to obtain fetal ECG signals. These abdominal electrical signal data were annotated according to birth outcome categories. This annotated abdominal fetal ECG data set was constructed as a benchmark dataset for fetal ECG dynamics analysis.

[0034] S102: Design a topographical analysis algorithm suitable for dynamic analysis of fetal ECG signals, construct a fetal ECG signal state template, generate a time series diagram of fetal ECG signal states, and calculate state statistical parameters. The state statistical parameter sequence corresponds to the annotated data collected in S101, and the sequence is annotated identically to construct a benchmark dataset for fetal state prediction.

[0035] S103: Construct a decision tree model, train the training data of the statistical parameters of the topographic map state with the fetal state annotation collected in S102, and obtain trained model parameters.

[0036] S104: Extract the test data of the statistical parameter sequence of the topographic map state with the fetal state annotation collected in S102 and input it into the decision tree model, and use the parameters obtained in S103 to predict the fetal state.

[0037] As a preferred embodiment of the present invention, in step S101, the process of data collection is as follows:

[0038] aaa: The same NI-FECG device and electrodes were used to collect 4-lead abdominal electrical signals from pregnant women at different gestational ages. The signal data covered a variety of abdominal electrical signal samples from pregnant women at different gestational ages and physical conditions over 35 weeks.

[0039] bbb: Data related to delivery outcomes were collected, including umbilical artery blood gas pH, whole blood base excess, lactate, oxygen tension, carbon dioxide tension, and Apgar score;

[0040] ccc: pre-process and signal separation of the collected abdominal electrical signals to obtain fetal electrocardiogram signals;

[0041] ddd: based on the degree of neonatal hypoxia and acidosis recorded in the real delivery outcome, the fetal electrocardiogram data is classified and labeled as 0 representing adverse delivery outcome and 1 representing good delivery outcome.

[0042] In the step S102, the topographic map analysis is based on the characteristics of the multi-lead abdominal electrode itself and the time-varying characteristics of the fetal electrocardiogram signal, which is an application of microstate analysis of electroencephalogram signals on fetal abdominal electrical signals with the same characteristics. The topographic map analysis includes state segmentation, fetal electrocardiogram data state fitting, and state statistical parameter calculation. The state segmentation is used to analyze the energy distribution of the fetal electrocardiogram signal to generate a template topographic map; the fetal electrocardiogram data state fitting is used to match the template topographic map with the electrocardiogram signal; and the state statistical parameter calculation can obtain the statistical characteristics of the matching of each template topographic map and the signal, which can potentially represent a specific state mechanism.

[0043] Further, the state segmentation part of step S102 includes the following specific steps:

[0044] aaa: calculate the instantaneous distribution potential topographic map of the fetal electrocardiogram signal;

[0045] bbb: get the local maximum topographic map curve;

[0046] ccc: apply K-means clustering technology to determine the optimal cluster number as 10 and obtain the template topographic map.

[0047] Further, the fetal electrocardiogram data state fitting part of step S102 includes the following specific steps:

[0048] aaa: calculate the topological structure of the fetal electrocardiogram signal at each time point in the time domain;

[0049] bbb: calculate the global topographic map dissimilarity with the template topographic map;

[0050] ccc: dissimilarity of 0 indicates that the template topographic map and the topographic map at this time point are completely matched, and dissimilarity of 2 indicates that the template topographic map and the topographic map at this time point are opposite in polarity but the same in structure. According to the matching result, a topographic map time sequence change map is constructed.

[0051] Further, the state statistical parameter calculation of step S102 calculates the statistical distribution characteristics of each template topographic map in the signal, including duration, occurrence rate, coverage rate and transfer rate, and constructs a state statistical parameter sequence for predicting the fetal state.

[0052] In the step S103, a binary classification decision tree model is constructed, and the annotated state statistical parameter sequence collected in S102 is divided into a training set and a test set according to an 8:2 ratio, and the training set is used for model training in S103.

[0053] In step S104, the process of testing the model performance includes:

[0054] aaa: Input the test data set divided by S103 into the model trained by S103 to predict the fetal status;

[0055] bbb: calculate the prediction accuracy, sensitivity, precision, specificity, and F1 score of the performance indicators;

[0056] ccc: draws the ROC curve.

[0057] The above description of the disclosed embodiments is intended to enable those skilled in the art to utilize the present invention. Numerous modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein. Those skilled in the art may also make various equivalent modifications or substitutions without departing from the spirit of the present invention.

Claims

1. A method for analyzing the dynamics of fetal electrocardiogram signals and predicting fetal status based on topographical analysis, comprising the following steps: S1: Place the electrodes of the NI-FECG device on the abdomen of a pregnant woman to collect real abdominal mixed electrical signals and collect information about the delivery outcomes of the pregnant woman. The collected abdominal electrical signals are preprocessed and separated to obtain fetal ECG signals. The abdominal electrical signal data are annotated according to the delivery outcome classification, and the abdominal fetal ECG signals containing the annotated information are constructed as a benchmark dataset for fetal ECG signal dynamic analysis. S2: Design a topographic analysis algorithm suitable for dynamic analysis of fetal ECG signals. This algorithm is based on the inherent characteristics of multi-lead abdominal electrodes and the temporal variation characteristics of fetal ECG signals. It is an extension of the microstate analysis of EEG signals to the abdominal fetal electrical signals with the same characteristics. Using the topographic analysis algorithm, a state template of the fetal ECG signal is constructed, a time series variation diagram of the fetal ECG signal state is generated, and state statistical parameters are calculated. The state statistical parameter sequence corresponds to the annotated data collected in S1, and this sequence is annotated identically to construct a benchmark data set for fetal state prediction. S3, building a decision tree model, training the training data of the topographic state statistical parameters with fetal state annotation collected in S2, and obtaining the trained model parameters; S4, extracting the test data of the statistical parameter sequence of the topographic map state with the fetal state annotation collected by S2 and inputting it into the decision tree model, and using the parameters obtained by S3 to predict the fetal state.

2. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 1, characterized in that: In S1, the same NI-FECG device and electrodes were used to collect abdominal electrical signals from pregnant women at different gestational weeks. The signal data covered a variety of abdominal electrical signal samples from pregnant women at different gestational weeks and in different physical conditions.

3. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 1, characterized in that: In S1, after the various types of abdominal electrical signals collected are preprocessed and separated to obtain fetal ECG signals, the data are classified and labeled as adverse delivery outcomes or good delivery outcomes based on the degree of neonatal hypoxia and acidosis recorded in the actual delivery outcomes.

4. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographical analysis according to claim 1, characterized in that: Topographic analysis includes state segmentation, fetal ECG data state fitting, and state statistical parameter calculation; Among them, state segmentation is used to analyze the energy distribution of fetal ECG signals and generate template topography; fetal ECG data state fitting is used to match the template topography with the ECG signal; state statistical parameter calculation can obtain the statistical characteristics of the matching between each template topography and the signal, which can potentially represent a specific state mechanism.

5. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 4, characterized in that: The state segmentation part analyzes the energy distribution of the fetal ECG signal and generates a template topographic map. The process includes: The instantaneous distribution potential topography of the fetal electrocardiogram signal is calculated to obtain the local maximum topography curve. The clustering technology is applied to determine the optimal number of clusters to obtain the template topography.

6. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 4, characterized in that: The process of fetal ECG data state fitting to match the template topography with the ECG signal includes: The topological structure of the fetal ECG signal at each time point is calculated in the time domain, and the similarity is calculated with the template topographic map. The template topographic map with high similarity matches the signal at that time point.

7. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 4, characterized in that: The state statistical parameters are calculated to obtain the statistical characteristics of each template topography and signal matching, which are used to potentially characterize a specific state mechanism. The process includes: The statistical distribution characteristics of each template topography in the signal, including duration, incidence, coverage and transfer rate, were calculated, and these statistical parameters were used to predict the fetal status in S3 and S4.

8. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 1, characterized in that: In S3, a binary classification decision tree model is constructed, and the annotated state statistical parameter sequence collected by S2 is divided into a training set and a test set. The training set is used for model training in S3.

9. The method for dynamic analysis of fetal electrocardiogram signals and prediction of fetal status based on topographic map analysis according to claim 1, characterized in that: In S4, the process of testing model performance includes: The test data set divided by S3 is input into the model trained by S3 to predict the fetal status, and the prediction accuracy, sensitivity, precision, specificity, F1 score of the performance indicators are calculated, and the ROC curve is drawn.

Images