A risk prediction system and method for severe postpartum hemorrhage

Abstract

The present application provides a risk prediction system and method for severe postpartum hemorrhage. First, principal component analysis is performed on all postpartum hemorrhage risk data to obtain the principal component characteristics of severe postpartum hemorrhage risk; then, all postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, thereby determining the data constraints of each cluster; further, feature extraction is performed on all prenatal examination data to obtain a data feature graph, thereby determining the confidence boundary of each cluster; then, secondary clustering is performed on all clusters to obtain multiple secondary clusters; a risk prediction model for severe postpartum hemorrhage is determined based on all secondary clusters, and the risk prediction model is used to evaluate the risk value of severe hemorrhage in parturients. The scheme of the present application can reduce the impact of data imbalance on the clustering of postpartum hemorrhage risk data, thereby improving the accuracy of severe postpartum hemorrhage risk prediction.

Description

Technical Field

[0001] The present application relates to the technical field of data clustering, and more specifically, to a risk prediction system and method for severe postpartum hemorrhage. Background Art

[0002] The application of data processing in the medical field covers a large number of tasks, from disease diagnosis to the formulation of treatment plans, covering multiple links such as data collection, cleaning, analysis, modeling and visualization. First, through sensors and imaging technology, medical data can be collected in real time, including patients' physiological parameters, medical images, genomic data, etc. Then, these data need to be cleaned and preprocessed to remove noise and incomplete information to ensure data quality. Then, using technologies such as machine learning and deep learning, tasks such as medical image analysis, disease prediction and personalized treatment can be achieved. In addition, data processing technology can also integrate patients' historical medical records through electronic health record systems, conduct big data analysis, and explore potential health risk factors and early signs of diseases. Data visualization technology helps doctors understand complex medical data in an intuitive way and assists clinical decision-making. In general, data processing technology improves the efficiency, accuracy and personalization of medical services, and promotes the digital transformation of the medical and health industry.

[0003] In the processing of severe postpartum hemorrhage risk data, clustering is mainly used to identify patient groups with similar risk characteristics, thereby helping doctors to provide early warning and personalized treatment. Through cluster analysis, patients can be divided into different risk groups based on clinical data (such as age, mode of delivery, stage of labor, blood test results, etc.), and potential high-risk patients can be identified. However, in postpartum hemorrhage risk data, cases of severe bleeding are relatively rare, resulting in highly unbalanced data. Data imbalance may cause clustering algorithms to tend to ignore less common categories (such as severe bleeding risk groups). Therefore, how to reduce the impact of data imbalance on clustering of postpartum hemorrhage risk data and thus improve the accuracy of severe postpartum hemorrhage risk prediction has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a system and method for predicting the risk of severe postpartum hemorrhage, which can reduce the impact of data imbalance on postpartum hemorrhage risk data clustering, thereby improving the accuracy of severe postpartum hemorrhage risk prediction.

[0005] In a first aspect, the present application provides a method for predicting the risk of severe postpartum hemorrhage, the method comprising the following steps:

[0006] Obtain data on antenatal care and postpartum hemorrhage risk for different mothers;

[0007] All postpartum hemorrhage risk data were subjected to principal component analysis to obtain the principal component characteristics of severe postpartum hemorrhage risk;

[0008] Clustering all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and determining the data constraint condition of each cluster by the maximum redundancy of each cluster and the principal component characteristics;

[0009] Perform feature extraction on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and determine a confidence boundary of each cluster according to the data feature graph and a cluster label of each cluster;

[0010] Perform secondary clustering on all clusters through the data constraints of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters;

[0011] A risk prediction model for severe postpartum hemorrhage is determined based on all secondary clusters, and the risk value of severe maternal hemorrhage is evaluated by the risk prediction model.

[0012] In some embodiments, principal component analysis is performed on all postpartum hemorrhage risk data to obtain principal component characteristics of severe postpartum hemorrhage risk, specifically including:

[0013] A covariance matrix was constructed based on all PPH risk data;

[0014] Performing eigenvalue decomposition on the covariance matrix to obtain a plurality of principal components;

[0015] The principal component characteristics of the risk of severe postpartum hemorrhage were determined by the loading of each principal component and the variance contribution of each principal component.

[0016] In some embodiments, all postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, specifically including:

[0017] Correlation analysis was performed on different postpartum hemorrhage risk data to obtain characteristic correlation coefficients;

[0018] Determine a mapping association feature of each postpartum hemorrhage risk data according to the mapping relationship between the feature correlation coefficient and different postpartum hemorrhage risk data;

[0019] Cluster analysis was performed on the amount of postpartum hemorrhage in each postpartum hemorrhage risk data using all the mapped association features to obtain multiple clusters.

[0020] In some embodiments, determining the data constraint condition of each cluster by the maximum redundancy of each cluster and the principal component feature specifically includes:

[0021] Perform linear fitting on the maximum redundancy of each cluster to obtain the cluster redundancy curve;

[0022] The data constraint condition of each cluster is determined according to the cluster redundancy curve and the principal component characteristics.

[0023] In some embodiments, feature extraction is performed on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, specifically including:

[0024] Preprocessing all ultrasound images in each prenatal examination data to obtain multiple preprocessed ultrasound images;

[0025] Load pre-trained machine learning models;

[0026] Extracting the output result of each preprocessed ultrasound image at the convolutional layer according to the machine learning model;

[0027] Using a filter to filter the output result of each preprocessed ultrasound image in the convolution layer to obtain multiple filter vectors;

[0028] The feature representation composed of all filter vectors is used as the data feature map during prenatal examination.

[0029] In some embodiments, determining the confidence boundary of each cluster according to the data feature graph and the cluster label of each cluster specifically includes:

[0030] Select a cluster as the selected cluster;

[0031] Determining the closeness of each bleeding association value in the selected cluster according to the maximum similarity of each filter vector in the data feature graph and the cluster label of the selected cluster;

[0032] The confidence boundary of the selected cluster is determined by the closeness of all bleeding association values ​​within the selected cluster;

[0033] Continue to determine confidence bounds for the remaining clusters.

[0034] In some embodiments, performing secondary clustering on all clusters by using the data constraint condition of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters specifically includes:

[0035] Determine the characteristic center of each cluster;

[0036] Determining a plurality of core bleeding association values ​​of each cluster according to the feature center of each cluster and the confidence boundary of each cluster;

[0037] Each cluster is re-clustered according to the data constraint conditions of each cluster and multiple core bleeding association values ​​of each cluster to obtain a secondary cluster of each cluster.

[0038] In a second aspect, the present application provides a severe postpartum hemorrhage risk prediction system, the severe postpartum hemorrhage risk prediction system comprising a risk data processing unit, the risk data processing unit comprising:

[0039] The acquisition module is used to obtain the prenatal examination data and postpartum hemorrhage risk data of different parturients;

[0040] A processing module is used to perform principal component analysis on all postpartum hemorrhage risk data to obtain the principal component characteristics of severe postpartum hemorrhage risk;

[0041] The processing module is further used to cluster all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and determine the data constraint conditions of each cluster through the maximum redundancy of each cluster and the principal component characteristics;

[0042] The processing module is further used to extract features from all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and determine a confidence boundary of each cluster cluster according to the data feature graph and a cluster label of each cluster cluster;

[0043] The processing module is further used to perform secondary clustering on all clusters according to the data constraint conditions of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters;

[0044] The evaluation module is used to determine a risk prediction model for severe postpartum hemorrhage based on all secondary clusters, and to evaluate the risk value of severe hemorrhage of the parturient through the risk prediction model.

[0045] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned risk prediction method for severe postpartum hemorrhage.

[0046] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned risk prediction method for severe postpartum hemorrhage is implemented.

[0047] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0048] In the severe postpartum hemorrhage risk prediction system and method provided in the present application, first, prenatal examination data and postpartum hemorrhage risk data of different parturients are obtained; principal component analysis is performed on all postpartum hemorrhage risk data to obtain principal component characteristics of severe postpartum hemorrhage risk; all postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and the data constraints of each cluster are determined by the maximum redundancy of each cluster and the principal component characteristics; feature extraction is performed on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and the confidence boundary of each cluster is determined according to the data feature graph and the cluster label of each cluster; secondary clustering is performed on all clusters according to the data constraint conditions of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters; a risk prediction model for severe postpartum hemorrhage is determined based on all secondary clusters, and the risk value of severe bleeding of parturients is evaluated by the risk prediction model.

[0049] It can be seen that in the present application, a risk prediction model for severe postpartum hemorrhage can be determined based on all secondary clustering clusters; wherein, firstly, the postpartum hemorrhage risk data is subjected to dimensionality reduction processing through principal component analysis, the most representative principal component features are extracted, and redundant and noise information is removed. This process reduces the complexity of the data and provides a clearer and more efficient input for subsequent clustering analysis. The extraction of principal components not only retains information that is crucial to the prediction of severe bleeding risk, but also provides a basis for setting data constraints in the clustering process, thereby ensuring the effective use of key features in the clustering analysis process; then, based on the clustering analysis method, all postpartum hemorrhage risk data are preliminarily clustered, and by clustering the mapping relationships of different postpartum hemorrhage risk data, samples with similar risks can be classified into the same clustering cluster in the case of data imbalance, thereby reducing the dispersion of a few high-risk samples, enabling the model to concentrate on learning on these high-risk data, and avoiding the interference of low-risk samples on the results. In the clustering process, the calculation of the maximum redundancy and the combination of principal component features ensure the accuracy of the clustering results, and also set reasonable data constraints for each clustering cluster. , making the clustering results more stable; in addition, in the feature extraction stage of prenatal examination data, based on machine learning models (such as support vector machines, random forests, etc.), deep feature learning is performed on the prenatal examination data to extract key features that are helpful for risk prediction, and combined with the clustering labels of each cluster cluster, the confidence boundary of each cluster is further set. The purpose of this process is to determine the risk area of ​​each cluster cluster, so as to accurately distinguish the high-risk group in the subsequent secondary clustering. The combination of clustering labels and confidence boundaries ensures the accurate identification of high-risk groups; finally, based on the results of secondary clustering analysis, a prediction model for the risk of severe postpartum hemorrhage is constructed. The model can comprehensively consider the characteristic information, confidence boundaries and data constraints of the cluster clusters, and ultimately improve the recognition rate of high-risk pregnant women. Compared with the traditional classification method, this scheme not only improves the accuracy of the prediction model through precise clustering and data constraints, but also significantly reduces the negative impact of data imbalance on model training; in summary, the scheme of the present application can reduce the impact of data imbalance on the clustering of postpartum hemorrhage risk data, thereby improving the accuracy of severe postpartum hemorrhage risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is an exemplary flow chart of a method for predicting the risk of severe postpartum hemorrhage according to some embodiments of the present application;

[0051] Figure 2 is a schematic diagram of the structure of a machine learning model according to some embodiments of the present application;

[0052] Figure 3 is a schematic diagram of a process for determining secondary clustering clusters in some embodiments of the present application;

[0053] Figure 4 is a schematic diagram of the structure of a risk data processing unit in some embodiments of the present application;

[0054] Figure 5 It is a structural schematic diagram of a computer device for implementing a risk prediction method for severe postpartum hemorrhage according to some embodiments of the present application. DETAILED DESCRIPTION

[0055] In order to better understand the technical solution of the present application, the technical solution of the present application will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0056] refer to Figure 1 , which is an exemplary flow chart of a method for predicting the risk of severe postpartum hemorrhage according to some embodiments of the present application. The method 100 for predicting the risk of severe postpartum hemorrhage mainly includes the following steps:

[0057] In step 101, prenatal examination data and postpartum hemorrhage risk data of different parturients are obtained.

[0058] In specific implementation, the prenatal examination data and postpartum hemorrhage risk data of different mothers are obtained from the database of the hospital's electronic health record system.

[0059] It should be noted that the prenatal examination data described in this application refers to the data of the mother's pre-delivery examination, and the prenatal examination data is composed of ultrasound images of ultrasound examinations performed at different stages of pregnancy.

[0060] In addition, it should be noted that the postpartum hemorrhage risk data described in this application are data consisting of the age of the pregnant woman, number of pregnancies and parities, maternal body mass index, postpartum blood pressure, previous bleeding events (1 for occurrence, 0 for non-occurrence), mode of delivery (1 for vaginal delivery, 0 for cesarean section), time of delivery, fetal weight, birth canal injury (1 for loss and 0 for no injury), placental retention time, uterine recovery time and the amount of postpartum bleeding at different time points.

[0061] In step 102, principal component analysis is performed on all postpartum hemorrhage risk data to obtain principal component characteristics of severe postpartum hemorrhage risk.

[0062] In some embodiments, principal component analysis is performed on all postpartum hemorrhage risk data to obtain principal component characteristics of severe postpartum hemorrhage risk by using the following steps:

[0063] A covariance matrix was constructed based on all PPH risk data;

[0064] Performing eigenvalue decomposition on the covariance matrix to obtain a plurality of principal components;

[0065] The principal component characteristics of the risk of severe postpartum hemorrhage were determined by the loading of each principal component and the variance contribution of each principal component.

[0066] It should be noted that the principal component in this application represents the most representative features extracted from all postpartum hemorrhage risk data.

[0067] In the specific implementation, first, the NumPy library in Python is used to calculate the covariance matrix of all postpartum hemorrhage risk data; secondly, Scikit-learn in the machine learning library is used to perform eigenvalue decomposition on the covariance matrix and extract multiple principal components; then, Scikit-learn is used to obtain the load and variance contribution rate of each principal component, the load of each principal component is multiplied by the variance contribution rate of each principal component, and then all the multiplied values ​​are summed, and the summed value is used as the principal component feature of severe postpartum hemorrhage risk. Other methods can also be used in other embodiments, which are not limited here.

[0068] It should be noted that the principal component features described in this application represent the characteristic values ​​of the main factors affecting the risk of postpartum hemorrhage. The principal component features are used to enhance the recognition ability of minority data, thereby helping the clustering algorithm to better identify minority data.

[0069] In step 103, all postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and the data constraint condition of each cluster is determined by the maximum redundancy of each cluster and the principal component characteristics.

[0070] In some embodiments, clustering all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters can be achieved by the following steps:

[0071] Correlation analysis was performed on different postpartum hemorrhage risk data to obtain characteristic correlation coefficients;

[0072] Determine a mapping association feature of each postpartum hemorrhage risk data according to the mapping relationship between the feature correlation coefficient and different postpartum hemorrhage risk data;

[0073] Cluster analysis was performed on the amount of postpartum hemorrhage in each postpartum hemorrhage risk data using all the mapped association features to obtain multiple clusters.

[0074] It should be noted that the mapping relationship in the present application represents the mapping value between postpartum hemorrhage risk data and maternal hemorrhage risk prediction value. The historical postpartum hemorrhage risk data and the historical maternal hemorrhage risk prediction value can be trained by the random forest algorithm, and then the postpartum hemorrhage risk data is input into the model obtained after the training, and the model output result is used as the mapping relationship of the postpartum hemorrhage risk data.

[0075] In the specific implementation, first, the Pearson correlation coefficient between each feature in all postpartum hemorrhage risk data is calculated, and then all Pearson correlation coefficients are summed, and the summed value is used as the feature correlation coefficient; secondly, the mapping relationship of all postpartum hemorrhage risk data is Min-Max standardized, and the values ​​obtained after standardization are used as the mapping standard values ​​of each postpartum hemorrhage risk data; then, the feature correlation coefficient is multiplied by the mapping standard value of each postpartum hemorrhage risk data, and the multiplied value is used as the mapping association feature of each postpartum hemorrhage risk data; finally, the postpartum bleeding volume in each postpartum hemorrhage risk data is divided by the mapping association feature of each postpartum hemorrhage risk data, and the divided values ​​are used as the bleeding association value, and all the bleeding association values ​​are further clustered using a density-based clustering algorithm (DBSCAN) to obtain cluster clusters. In other embodiments, other methods can also be used to achieve this, which will not be repeated here.

[0076] It should be noted that the cluster described in the present application is composed of multiple bleeding association values, wherein the bleeding association value represents the correlation parameter between the amount of maternal bleeding and the characteristics that affect the amount of bleeding, the larger the bleeding association value, the higher the correlation between the amount of maternal bleeding and the characteristics that affect the amount of bleeding, and each bleeding association value corresponds to a parturient; the characteristic correlation coefficient represents an indicator of the strength of the linear relationship between different postpartum hemorrhage risk data; the mapping association feature represents the association strength feature of the mapping between postpartum hemorrhage risk data and maternal bleeding risk.

[0077] In some embodiments, determining the data constraint condition of each cluster by the maximum redundancy of each cluster and the principal component feature can be implemented by the following steps:

[0078] Perform linear fitting on the maximum redundancy of each cluster to obtain the cluster redundancy curve;

[0079] The data constraint condition of each cluster is determined according to the cluster redundancy curve and the principal component characteristics.

[0080] It should be noted that the maximum redundancy in the present application represents the parameter value of the repetition of feature information within the cluster, and the maximum redundancy reflects the feature diversity and information richness within the cluster.

[0081] In the specific implementation, first, an existing linear fitting algorithm (such as a least squares support vector machine algorithm) is used to perform linear fitting on the maximum redundancy of all cluster clusters, and the fitted curve is used as a cluster redundancy curve, wherein the values ​​on the cluster redundancy curve are all used as maximum redundancy fitting values, and each maximum redundancy fitting value corresponds to a maximum redundancy, and each maximum redundancy corresponds to a cluster cluster; secondly, the maximum redundancy corresponding to the maximum redundancy fitting value is subtracted from each maximum redundancy fitting value on the cluster redundancy curve, and the subtracted values ​​are multiplied by the principal component features, and then the multiplied values ​​are used as data constraints for each cluster cluster corresponding to the maximum redundancy. Other methods may also be used in other embodiments, which are not limited here.

[0082] It should be noted that the data constraints described in the present application represent characteristic parameters that constrain different bleeding association values ​​within the clustering clusters. The data constraints can guide the clustering algorithm to pay more attention to important risk features, especially by placing greater weights on data of minority classes (such as severe bleeding), thereby reducing the negative impact of unbalanced data on the clustering results.

[0083] In step 104, feature extraction is performed on all prenatal examination data based on a preset machine learning model to obtain a data feature graph of the prenatal examination, and the confidence boundary of each cluster is determined based on the data feature graph and the cluster label of each cluster.

[0084] In some embodiments, feature extraction is performed on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, which can be achieved by the following steps:

[0085] Preprocessing all ultrasound images in each prenatal examination data to obtain multiple preprocessed ultrasound images;

[0086] Load pre-trained machine learning models;

[0087] Extracting the output result of each preprocessed ultrasound image at the convolutional layer according to the machine learning model;

[0088] Using a filter to filter the output result of each preprocessed ultrasound image in the convolution layer to obtain multiple filter vectors;

[0089] The feature representation composed of all filter vectors is used as the data feature map during prenatal examination.

[0090] In the specific implementation, first, all ultrasound images in each prenatal examination data are normalized and enhanced. Image enhancement includes contrast stretching, image sharpening, and denoising to improve image quality and prepare for subsequent feature extraction. Secondly, a pre-trained machine learning model is loaded, and a convolutional neural network is usually selected as the model framework. Figure 2 As shown, this figure is a structural schematic diagram of the machine learning model shown in some embodiments of the present application, which is used to process ultrasound images, extract features through the convolution layer, and then reduce the dimension through the pooling layer. After multiple feature extractions, it is mapped to the output through the fully connected layer, and finally the output layer gives the result. The machine learning model has been trained on a large number of medical image data sets, so it can extract important features in the image; then, after the machine learning model is loaded, all preprocessed ultrasound images are used as input, and the output result of each preprocessed ultrasound image in the convolution layer is extracted through the pretrained machine learning model; next, a high-pass filter is used to perform sliding filtering on the output result of each preprocessed ultrasound image in the convolution layer, and the vectors obtained after the sliding filtering are used as filter vectors. The sliding filtering can effectively highlight the high-frequency details of the image, such as edge and texture information; finally, all the vectors obtained by filtering are integrated to form a comprehensive feature representation, and the obtained feature representation is used as the data feature map during prenatal examination. Other methods can also be used in other embodiments, which will not be repeated here.

[0091] It should be noted that the data feature graph described in this application represents a high-dimensional feature representation of all prenatal examination data that reflects the health status of the mother.

[0092] In some embodiments, determining the confidence boundary of each cluster cluster according to the data feature graph and the cluster label of each cluster cluster can be implemented by the following steps:

[0093] Select a cluster as the selected cluster;

[0094] Determining the closeness of each bleeding association value in the selected cluster according to the maximum similarity of each filter vector in the data feature graph and the cluster label of the selected cluster;

[0095] The confidence boundary of the selected cluster is determined by the closeness of all bleeding association values ​​within the selected cluster;

[0096] Continue to determine confidence bounds for the remaining clusters.

[0097] It should be noted that the cluster label described in the present application represents an identifier assigned to each bleeding association value of the cluster, and the identifier is a discrete category number.

[0098] In specific implementation, first, the cosine similarity between each filter vector in the data feature graph and other filter vectors is calculated, and the maximum cosine similarity is used as the maximum similarity of each filter vector, the maximum maximum similarity and the minimum maximum similarity are selected from all the maximum similarities, and the difference between the maximum similarity and the minimum maximum similarity is used as the similarity difference, and further all values ​​in the cluster label of the selected cluster cluster are summed, and the summed value is used as the global identification feature value, and then the identifier of each bleeding association value in the selected cluster cluster is divided by the global identification feature value and multiplied by the similarity difference, and the multiplied value is used as the compactness of each bleeding association value in the selected cluster cluster; secondly, the compactness of all bleeding association values ​​in the selected cluster cluster is fitted by a density estimation method (such as Gaussian mixture model, k-nearest neighbor density estimation, etc., which are not limited here), to obtain the boundary of the selected cluster cluster, and the obtained boundary is used as the confidence boundary of the selected cluster cluster. In other embodiments, other methods can also be used for implementation, which are not limited here.

[0099] It should be noted that the confidence boundary described in the present application represents the boundary that defines the valid range of bleeding association values ​​within the cluster. Through the precise division of the confidence boundary, low-density, high-discrete clusters can be effectively isolated, reducing the impact of data imbalance on the results, thereby better reflecting the characteristics of the minority class.

[0100] In step 105, all clusters are secondary clustered according to the data constraint condition of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters.

[0101] In some embodiments, reference Figure 3 As shown, this figure is a schematic diagram of the process of determining secondary clustering clusters in some embodiments of the present application. In this embodiment, all clustering clusters are secondary clustered by the data constraints of each clustering cluster and the confidence boundary of each clustering cluster. The following steps can be used to obtain multiple secondary clustering clusters:

[0102] First, in step 1051, the feature center of each cluster is determined;

[0103] Secondly, in step 1052, multiple core bleeding association values ​​of each cluster are determined according to the feature center of each cluster and the confidence boundary of each cluster;

[0104] Then, in step 1053, each cluster is re-clustered according to the data constraint conditions of each cluster and the multiple core bleeding association values ​​of each cluster, so as to obtain a secondary cluster of each cluster.

[0105] In the specific implementation, first, the characteristic center of each cluster is extracted using the Gaussian mixture model, where the characteristic center represents the center of all bleeding association values ​​in the cluster; secondly, the bleeding association values ​​outside the confidence boundary of all clusters are removed, and the set of bleeding association values ​​remaining after removing the bleeding association values ​​from each cluster is used as the temporary cluster of each cluster, and then a cluster is further selected as the selected cluster, and the distances from all bleeding association values ​​in the temporary cluster of the selected cluster to the characteristic center of the selected cluster are calculated, and then the bleeding association values ​​whose distances are less than the mean of all distances are selected from the temporary cluster of the selected cluster, and the selected bleeding association values ​​are used as the temporary cluster. The associated values ​​are all used as core bleeding associated values, and multiple core bleeding associated values ​​of the remaining clusters are further determined; then, the bleeding associated values ​​of the non-core bleeding associated values ​​in each cluster are multiplied by the data constraints of each cluster, and then the DBSCAN clustering algorithm is used to cluster all the multiplied values ​​and all the core bleeding associated values ​​of each cluster, and the clusters obtained by clustering are used as secondary clusters, wherein all data points in the secondary clusters are used as severe bleeding risk values, and each severe bleeding risk value corresponds to a parturient, thereby obtaining a secondary cluster for each cluster. Other methods can also be used to achieve this in other embodiments, which will not be repeated here.

[0106] In step 106, a risk prediction model for severe postpartum hemorrhage is determined based on all secondary clusters, and the risk value of severe hemorrhage of the parturient is evaluated by the risk prediction model.

[0107] In some embodiments, determining a risk prediction model for severe postpartum hemorrhage based on all secondary clustering clusters may be implemented by the following steps:

[0108] Obtain the postpartum hemorrhage risk data of the parturient corresponding to each severe bleeding risk value in each secondary cluster;

[0109] All severe bleeding risk values ​​in each secondary cluster are combined into severe bleeding labels;

[0110] The prediction model is trained using all the postpartum hemorrhage risk data and the severe bleeding labels, and the obtained prediction model is used as a risk prediction model for severe postpartum hemorrhage.

[0111] In the specific implementation, all postpartum hemorrhage risk data and severe bleeding labels are used as training data for the model, and the training data is trained using a supervised learning model (for example, support vector machine, random forest or XGBoost, not limited here), and the model obtained after the training is used as a risk prediction model for severe postpartum hemorrhage. In other embodiments, other methods can also be used to implement it, which will not be repeated here.

[0112] It should be noted that the severe bleeding risk value described in the present application indicates the risk level of severe bleeding in a parturient after delivery. The larger the severe bleeding risk value, the higher the risk level of severe bleeding in a parturient after delivery, and vice versa.

[0113] In specific implementation, the risk prediction model can be used to evaluate the risk value of severe maternal bleeding in the following manner: after the parturient is admitted to the hospital, the postpartum hemorrhage risk data of the parturient is collected, the obtained postpartum hemorrhage risk data is input into the risk prediction model, and the output result of the risk prediction model is used as the risk value of severe maternal bleeding, the risk value is divided into different risk levels, and finally the risk level of the parturient is used as the final evaluation result. Other methods can also be used in other embodiments, which are not limited here.

[0114] In addition, in another aspect of the present application, in some embodiments, the present application provides a severe postpartum hemorrhage risk prediction system, the severe postpartum hemorrhage risk prediction system includes a risk data processing unit, referring to Figure 4 , which is a schematic diagram of the structure of a risk data processing unit according to some embodiments of the present application, the risk data processing unit 400 includes: an acquisition module 401, a processing module 402 and an evaluation module 403, which are described as follows:

[0115] Acquisition module 401, in this application, acquisition module 401 is mainly used to obtain prenatal examination data and postpartum hemorrhage risk data of different parturients;

[0116] Processing module 402, in this application, processing module 402 is mainly used to perform principal component analysis on all postpartum hemorrhage risk data to obtain principal component characteristics of severe postpartum hemorrhage risk;

[0117] It should be noted that the processing module 402 in the present application is also used to cluster all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and determine the data constraint conditions of each cluster by the maximum redundancy of each cluster and the principal component characteristics;

[0118] It should be noted that the processing module 402 in the present application is also used to extract features from all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and determine the confidence boundary of each cluster cluster according to the data feature graph and the cluster label of each cluster cluster;

[0119] In addition, the processing module 402 in the present application is also used to perform secondary clustering on all clusters through the data constraint conditions of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters;

[0120] Evaluation module 403: In the present application, evaluation module 403 is mainly used to determine a risk prediction model for severe postpartum hemorrhage based on all secondary clusters, and to evaluate the risk value of severe postpartum hemorrhage through the risk prediction model.

[0121] In addition, the present application also provides a computer device, which includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned risk prediction method for severe postpartum hemorrhage.

[0122] In some embodiments, reference Figure 5 , which is a schematic diagram of the structure of a computer device for implementing a risk prediction method for severe postpartum hemorrhage according to some embodiments of the present application. The risk prediction method for severe postpartum hemorrhage in the above embodiment can be Figure 5 The computer device 500 shown in the figure is implemented, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503 and at least one communication interface 504.

[0123] Processor 501 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more processors for controlling the execution of the risk prediction method for severe postpartum hemorrhage in the present application.

[0124] The communication bus 502 may be used to transmit information between the above-mentioned components.

[0125] The memory 503 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compressed optical disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 503 may exist independently and be connected to the processor 501 via the communication bus 502. The memory 503 may also be integrated with the processor 501.

[0126] The memory 503 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 501. The processor 501 is used to execute the program code stored in the memory 503. The program code may include one or more software modules. The method described in the above method embodiment can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0127] The communication interface 504 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0128] In a specific implementation, as an embodiment, a computer device may include multiple processors, each of which may be a single-CPU processor or a multi-CPU processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0129] The above-mentioned computer device may be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of computer device.

[0130] In addition, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned risk prediction method for severe postpartum hemorrhage.

[0131] In summary, in the severe postpartum hemorrhage risk prediction system and method disclosed in the embodiments of the present application, first, prenatal examination data and postpartum hemorrhage risk data of different parturients are obtained; principal component analysis is performed on all postpartum hemorrhage risk data to obtain the principal component characteristics of severe postpartum hemorrhage risk; all postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and the data constraints of each cluster are determined by the maximum redundancy of each cluster and the principal component characteristics; feature extraction is performed on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and the confidence boundary of each cluster is determined according to the data feature graph and the cluster label of each cluster; all clusters are secondary clustered according to the data constraint conditions of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters; a risk prediction model for severe postpartum hemorrhage is determined based on all secondary clusters, and the risk value of severe bleeding of parturients is evaluated by the risk prediction model. The solution of the present application can reduce the impact of data imbalance on postpartum hemorrhage risk data clustering, thereby improving the accuracy of severe postpartum hemorrhage risk prediction.

[0132] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0133] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for predicting the risk of severe postpartum hemorrhage, characterized in that: The steps include: Obtain data on antenatal care and postpartum hemorrhage risk for different mothers; All postpartum hemorrhage risk data were subjected to principal component analysis to obtain the principal component characteristics of severe postpartum hemorrhage risk; Clustering all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and determining the data constraint condition of each cluster by the maximum redundancy of each cluster and the principal component characteristics; Perform feature extraction on all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and determine a confidence boundary of each cluster according to the data feature graph and a cluster label of each cluster; Perform secondary clustering on all clusters through the data constraints of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters; Determining a risk prediction model for severe postpartum hemorrhage based on all secondary clusters, and evaluating the risk value of severe hemorrhage in parturient women through the risk prediction model; Among them, the risk prediction model for severe postpartum hemorrhage determined based on all secondary clusters specifically includes: Obtain the postpartum hemorrhage risk data of the parturient corresponding to each severe bleeding risk value in each secondary cluster; All severe bleeding risk values ​​in each secondary cluster are combined into severe bleeding labels; The prediction model is trained using all the postpartum hemorrhage risk data and the severe bleeding labels, and the obtained prediction model is used as a risk prediction model for severe postpartum hemorrhage.

2. The method according to claim 1, characterized in that Principal component analysis was performed on all postpartum hemorrhage risk data, and the principal component characteristics of severe postpartum hemorrhage risk were obtained, including: A covariance matrix was constructed based on all PPH risk data; Performing eigenvalue decomposition on the covariance matrix to obtain a plurality of principal components; The principal component characteristics of the risk of severe postpartum hemorrhage were determined by the loading of each principal component and the variance contribution of each principal component.

3. The method according to claim 1, characterized in that All postpartum hemorrhage risk data are clustered according to the mapping relationship of different postpartum hemorrhage risk data, and multiple clusters are obtained, including: Correlation analysis was performed on different postpartum hemorrhage risk data to obtain characteristic correlation coefficients; Determine a mapping association feature of each postpartum hemorrhage risk data according to the mapping relationship between the feature correlation coefficient and different postpartum hemorrhage risk data; Cluster analysis was performed on the amount of postpartum hemorrhage in each postpartum hemorrhage risk data using all the mapped association features to obtain multiple clusters.

4. The method according to claim 1, characterized in that The data constraint conditions of each cluster are determined by the maximum redundancy of each cluster and the principal component characteristics, specifically including: Perform linear fitting on the maximum redundancy of each cluster to obtain the cluster redundancy curve; The data constraint condition of each cluster is determined according to the cluster redundancy curve and the principal component characteristics.

5. The method according to claim 1, characterized in that Based on the preset machine learning model, all prenatal examination data are feature extracted to obtain the data feature graph during prenatal examination, including: Preprocessing all ultrasound images in each prenatal examination data to obtain multiple preprocessed ultrasound images; Load pre-trained machine learning models; Extracting the output result of each preprocessed ultrasound image at the convolutional layer according to the machine learning model; Using a filter to filter the output result of each preprocessed ultrasound image in the convolution layer to obtain multiple filter vectors; The feature representation composed of all filter vectors is used as the data feature map during prenatal examination.

6. The method according to claim 1, characterized in that Determining the confidence boundary of each cluster according to the data feature graph and the cluster label of each cluster specifically includes: Select a cluster as the selected cluster; Determining the closeness of each bleeding association value in the selected cluster according to the maximum similarity of each filter vector in the data feature graph and the cluster label of the selected cluster; The confidence boundary of the selected cluster is determined by the closeness of all bleeding association values ​​within the selected cluster; Continue to determine confidence bounds for the remaining clusters.

7. The method according to claim 1, characterized in that All clusters are clustered again by using the data constraints of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters, including: Determine the characteristic center of each cluster; Determining a plurality of core bleeding association values ​​of each cluster according to the feature center of each cluster and the confidence boundary of each cluster; Each cluster is re-clustered according to the data constraint conditions of each cluster and multiple core bleeding association values ​​of each cluster to obtain a secondary cluster of each cluster.

8. A severe postpartum hemorrhage risk prediction system, the severe postpartum hemorrhage risk prediction system comprising a risk data processing unit, characterized in that: The risk data processing unit comprises: The acquisition module is used to obtain the prenatal examination data and postpartum hemorrhage risk data of different parturients; A processing module is used to perform principal component analysis on all postpartum hemorrhage risk data to obtain the principal component characteristics of severe postpartum hemorrhage risk; The processing module is further used to cluster all postpartum hemorrhage risk data according to the mapping relationship of different postpartum hemorrhage risk data to obtain multiple clusters, and determine the data constraint conditions of each cluster through the maximum redundancy of each cluster and the principal component characteristics; The processing module is further used to extract features from all prenatal examination data based on a preset machine learning model to obtain a data feature graph during prenatal examination, and determine a confidence boundary of each cluster cluster according to the data feature graph and a cluster label of each cluster cluster; The processing module is further used to perform secondary clustering on all clusters according to the data constraint conditions of each cluster and the confidence boundary of each cluster to obtain multiple secondary clusters; An evaluation module, used to determine a risk prediction model for severe postpartum hemorrhage based on all secondary clusters, and to evaluate the risk value of severe hemorrhage of parturient women through the risk prediction model; Among them, determining the risk prediction model for severe postpartum hemorrhage based on all the secondary clustering clusters specifically includes: obtaining the postpartum hemorrhage risk data of the parturient corresponding to each severe bleeding risk value in each secondary clustering cluster, forming a severe bleeding label with all the severe bleeding risk values ​​in each secondary clustering cluster, training the prediction model through all the postpartum hemorrhage risk data and the severe bleeding labels, and using the obtained prediction model as the risk prediction model for severe postpartum hemorrhage.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the risk prediction method for severe postpartum hemorrhage according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the risk prediction method for severe postpartum hemorrhage according to any one of claims 1 to 7 is implemented.

Images