Interactive CSEP intraoperative hemorrhage prediction method and system
Through screening and modeling of multiple machine learning methods, an interactive CSEP intraoperative large bleeding prediction system is constructed, which solves the problems that are not considered in variable relationships in the existing technology, and achieves more accurate predictions and safer clinical decisions.
Patent Information
- Application Number
- CN202411041805.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-15
AI Technical Summary
Existing methods for predicting large bleeding intraoperative CSEP fail to effectively consider potential covariance or nonlinear relationships between variables, and lack accurate prediction systems based on machine learning, affecting the accuracy and safety of predictions.
Variable screening and modeling are adopted using a variety of machine learning methods, including random forests, LASSO regression, Boruta feature selection and XGBoost. Combined with algorithms such as NB classifiers, an interactive prediction system is built to provide accurate prediction results for intraoperative large bleeding through cloud deployment. .
It improves the accuracy and safety of prediction of major bleeding intraoperative CSEP, provides clinicians with targeted preoperative evaluation and decision-making support, and enhances the ability to identify major bleeding risks.
Smart Images

Figure CN118969310A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of machine learning prediction of massive bleeding during CSEP surgery, and in particular to an interactive CSEP intraoperative massive bleeding prediction method and system. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] Cesarean scar ectopic pregnancy (CSEP) is a complication of a cesarean section-born woman's next pregnancy, characterized by early pregnancy implanting in the scar tissue formed by a previous cesarean section. As the gestational age increases, if it is not diagnosed and treated surgically as soon as possible, late uterine rupture, massive bleeding, and even life-threatening conditions may occur. However, regardless of the surgical method used by CSEP patients, the degree of intraoperative bleeding will significantly affect the success of the operation.
[0004] Although some studies on the influencing factors of intraoperative bleeding in CSEP patients have been published, most of them use logistic regression methods to conduct multivariate analysis on the basis of univariate factors to explore the influencing factors. The problem with this type of analysis is that it ignores the potential covariance or nonlinear relationship between variables, reducing its effectiveness in solving nonlinear problems.
[0005] Currently, there is no risk factor screening method for massive bleeding (bleeding volume ≥ 300 m3) during CSEP surgery based on machine learning (ML), nor is there an interactive risk prediction Internet system based on ML to accurately predict massive bleeding during CSEP surgery. Summary of the invention
[0006] In order to solve the above problems, the present invention proposes an interactive CSEP intraoperative massive bleeding prediction method and system, which uses a variety of machine learning methods to screen variables and considers various potential relationships between variables to screen out accurate prediction factors. Eight machine learning methods are selected for modeling and used to predict CSEP intraoperative massive bleeding, ensuring the effectiveness of the prediction model construction and the accuracy of the prediction. At the same time, the model is deployed to the cloud to help clinicians quickly make targeted preoperative evaluations and clinical decisions.
[0007] In order to achieve the above object, the present invention adopts the following technical solution:
[0008] In a first aspect, the present invention provides an interactive CSEP intraoperative massive bleeding prediction method, comprising:
[0009] Obtain historical case data and extract the influencing factors of massive bleeding during CSEP surgery from the historical case data;
[0010] Use at least two machine learning methods to screen the influencing factors, and determine the predictor variables after taking the intersection of the variable screening results;
[0011] Based on the predictive variables, several machine learning methods were used to build models and predict massive bleeding during CSEP. The optimal prediction model was obtained by evaluating the prediction results of each model.
[0012] Receive parameter information of the prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
[0013] As an optional implementation method, the random forest method, LASSO regression method, Boruta feature selection method and XGBoost method are used to screen the variables of the influencing factors respectively.
[0014] As an optional embodiment, the determined predictive variables include: gestational age, duration of preoperative vaginal bleeding, average diameter of gestational sac, thinnest thickness of anterior uterine wall, preoperative HGB level, preoperative HCG level, presence of uterine arteriovenous fistula, presence of early pregnancy placenta accreta and blood flow grade.
[0015] As an optional implementation, several machine learning methods used in modeling include NB classifier, artificial neural network, decision tree, KNN classification, logistic regression, random forest, support vector machine and XGBoost method; the optimal prediction model obtained after evaluation is the prediction model constructed using the NB classifier.
[0016] As an optional implementation, when several machine learning methods are used for modeling, each model uses ten-fold internal cross validation.
[0017] As an optional implementation, when evaluating the prediction results of each model, accuracy, AUC (Area Under Curve, defined as the area under the ROC curve and the coordinate axis) and decision curve are used as evaluation indicators.
[0018] As an optional implementation, after the optimal prediction model is obtained, the sensitivity, specificity, Matthews correlation coefficient and F1 score are used to perform internal and external validation on the optimal prediction model.
[0019] As an optional implementation method, the optimal prediction model is deployed to the cloud to receive parameter information of the prediction variables input by the interactive interface and feed back the prediction results to the interactive interface; the deployment process includes: model visualization based on the iBreakDown component, and assigning the screened prediction variables and the optimal prediction model to the graphical interface and server for cloud deployment based on Shiny and ShinyDashboard components.
[0020] In a second aspect, the present invention provides an interactive CSEP intraoperative massive bleeding prediction system, comprising:
[0021] An acquisition module is configured to acquire historical case data and extract influencing factors affecting massive bleeding during CSEP surgery from the historical case data;
[0022] A variable screening module is configured to use at least two machine learning methods to screen the influencing factors, and determine the prediction variable after taking the intersection of the variable screening results;
[0023] A modeling module is configured to use several machine learning methods to respectively model and predict massive bleeding during CSEP based on the predictive variables, and obtain the optimal prediction model by evaluating the prediction results of each model;
[0024] The prediction module is configured to receive parameter information of prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
[0025] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.
[0026] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.
[0027] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.
[0028] Compared with the prior art, the present invention has the following beneficial effects:
[0029] The present invention proposes an interactive CSEP intraoperative massive hemorrhage prediction variable screening method, which determines 20 influencing factors affecting CSEP intraoperative massive hemorrhage from historical case data, including demographics, reproductive history, medical history, clinical symptoms and ultrasound examination characteristics; then, the random forest method, LASSO regression method, Boruta feature selection method and extreme gradient boosting tree XGBoost method are used to screen the variables of the 20 influencing factors respectively; finally, the variables screened by the above four machine learning methods are intersected to determine 9 predictive variables for model modeling and prediction. Variable screening is performed by multiple machine learning methods, and various potential relationships between variables are considered, so as to screen accurate predictive factors and ensure the effectiveness of subsequent prediction model construction and the accuracy of prediction.
[0030] Abstract: The present invention proposes an interactive CSEP intraoperative massive bleeding prediction method, selects 8 machine learning models for modeling and uses them to predict CSEP intraoperative massive bleeding. In order to optimize the prediction model, all models are developed using ten-fold internal cross-validation, and accuracy, AUC and decision curve are used as model evaluation indicators. Sensitivity, specificity, Matthews correlation coefficient and F1 score are used to carry out internal and external validation of the optimal prediction model, so as to improve the ability of the prediction model in identifying CSEP patients with intraoperative massive bleeding, and improve the prediction accuracy and sensitivity of the model, so as to provide theoretical and methodological support for the research direction of clinical medicine in the prediction of intraoperative massive bleeding in CSEP.
[0031] Based on the interactive CSEP intraoperative massive bleeding prediction method, the present invention develops and externally verifies an ML-based interpretable CSEP intraoperative massive bleeding risk prediction system, and deploys it to the cloud. Thus, after receiving the parameter information of the prediction variables input by the interactive interface, the CSEP intraoperative massive bleeding prediction results are obtained based on the optimal prediction model, and the prediction results are fed back to the interactive interface, which can help clinicians quickly make targeted preoperative evaluations and clinical decisions.
[0032] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0034] Figure 1A flow chart of the interactive CSEP intraoperative massive bleeding prediction method provided in Example 1 of the present invention;
[0035] Figure 2 An impact diagram of the effect of the predictor variables provided in Example 1 of the present invention on massive bleeding during CSEP surgery;
[0036] Figure 3 The importance ranking of predictor variables screened by random forest RF provided in Example 1 of the present invention;
[0037] Figure 4 The predictor variables screened by the LASSO regression method provided in Example 1 of the present invention;
[0038] Figure 5 The predictor variables selected by the Boruta feature selection method provided in Example 1 of the present invention;
[0039] Figure 6 The prediction variables selected by the XGBoost method provided in Example 1 of the present invention;
[0040] Figure 7 The nine prediction variables determined after taking the intersection provided in Embodiment 1 of the present invention;
[0041] Figure 8 The accuracy of the eight machine learning models provided in Example 1 of the present invention;
[0042] Fig. 9 ROC curves of 8 machine learning models provided in Example 1 of the present invention;
[0043] Fig.10 DCA curves of 8 machine learning models provided in Example 1 of the present invention;
[0044] Figures 11(a)-(c) are enlarged schematic diagrams of the interactive cloud interface provided in Example 1 of the present invention and the prediction results and control panel part of the interface. DETAILED DESCRIPTION
[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0046] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0047] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "comprise" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0048] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0049] Terminology explanation:
[0050] RF (Random Forest) is an integrated concept based on decision trees (DT). It contains multiple DT classifiers, has high accuracy and robustness, and can handle large-scale data sets. It is widely used in the field of machine learning. In classification problems, each DT will vote to give the final classification result. RF introduces randomness when constructing DT. It constructs a machine learning model by randomly sampling sample data and finding the optimal solution for splitting, which is conducive to discovering different nonlinear relationships between feature values.
[0051] LASSO regression (LASSO Regression, Least Absolute Convergence and Selection Operator Algorithm) is a regression analysis technique used for variable selection and regularization. As an alternative to the traditional least squares regression method, L1 regularization is introduced in the process of stochastic gradient descent of the linear model loss function to penalize the weight of the features. The weight of some features will be reduced or even completely reduced to 0, which can reduce the complexity and multicollinearity of the model, thereby avoiding overfitting and enhancing the generalization ability of the model. The characteristic is that feature screening and complexity adjustment are performed while fitting the model.
[0052] NB classifier (Naive Bayes) belongs to the family of probabilistic classifiers based on Bayesian theory and has been widely used in various fields. It mainly assumes that attributes are conditionally independent based on Bayesian theorem, calculates the conditional probability of the sample under different categories, and then classifies it as the one with the highest probability, which is why it is called "naive". Because of this, the parameters can be learned separately, which is simpler and faster.
[0053] ANN (Artificial Neural Network) is a mathematical model in machine learning algorithms that is relatively close to the characteristics of biological neural networks. It simulates the structure and function of biological neural networks, uses input layers, hidden layers, and output layers to map input data to appropriate outputs, and is composed of a large number of nodes (neurons) and interconnected connections. Its functions are similar to those of the human brain and can be used to model complex relationships between data. It is an abstraction and simulation of several basic characteristics of natural neural networks to discover nonlinear characteristics between data.
[0054] DT (Decision Tree) is an efficient and reliable decision-making algorithm with the characteristics of high classification accuracy and simple visualization of collected data. It has basically no requirements for the selection of data and can handle both data-type and conventional attributes. The DT algorithm is a tree structure in which each internal node represents a test on an attribute, each branch represents a test output, and each leaf node represents a category. It can handle collinearity problems between independent variables and display interactions between independent variables.
[0055] The KNN classification algorithm (K Nearest Neighbor) is a non-parametric supervised learning classifier that uses the proximity to other data points to predict the grouping of a single data point. It calculates the distance between the input sample and each sample in the training set, selects the first k samples with the closest distance as neighbors, and determines the category of the sample to be classified based on the classification labels of the k neighbor samples. If most of the K neighbors belong to the same category, the sample to be classified also belongs to that category.
[0056] LR (Logistic Regression) is a generalized linear regression analysis model used for classification problems by mapping features to classification targets and predicting the probability of new cases belonging to the target category. The probability of an event is estimated based on a given set of independent variable data, and the range of the dependent variable is between 0 and 1. Independent variables can be either continuous or categorical. It is used to find risk factors and predict the probability of a disease or condition under different independent variable conditions.
[0057] SVM (Support Vector Machine) is a supervised learning method proposed for binary classification problems. Its learning goal is to separate samples of two different categories by finding an optimal hyperplane and to ensure that the interval between samples of the two categories is the largest. It adopts the principle of structural risk minimization and measures structured risk using the VC dimension. SVM seeks an optimal compromise between model complexity and learning ability to achieve optimal generalization performance, and can use optimization theory to ensure its global optimal solution.
[0058] XGBoost (eXtreme Gradient Boosting) combines the additive model with the forward step-by-step algorithm, generates a gradient boosting tree model through continuous iteration, and combines many tree models to improve the accuracy of calculations. Its advantage is that it uses the second-order Taylor expansion to approximate the loss function, and then minimizes the loss function to find the optimal tree structure and leaf node value. It is good at solving sample predictions with high-dimensional factors and a large number of data features.
[0059] Boruta (Forest God) is different from the general feature selection algorithm in that it is a wrapper-based feature selection algorithm that aims to select the feature set most relevant to the dependent variable rather than the minimum compact feature set that best suits a specific model. It iteratively deletes features with low relevance, thereby reducing signal noise and achieving a stable classification effect.
[0060] Example 1
[0061] As a branch of artificial intelligence, ML is a scientific discipline that studies how computers learn and mine information from data. It has a stronger ability to extract information from data and has been continuously applied in the field of medical research in recent years. It is good at first training a part of the data, then predicting and judging other data, and can continuously improve the performance of the algorithm through empirical learning. Therefore, with its powerful autonomous learning ability, ML can extract predictive variables (influencing factors, eigenvalues) related to the risk of intraoperative massive bleeding in CSEP patients from clinical data, explore the hidden laws between the two, and build a precise multi-parameter combination model, so as to conduct personalized assessments of the probability of occurrence of clinical events in patients.
[0062] Therefore, this embodiment proposes an interactive CSEP intraoperative massive bleeding prediction method, develops and externally verifies an interpretable CSEP intraoperative massive bleeding risk prediction system based on ML methods, and deploys it to the cloud to help clinicians quickly make targeted preoperative evaluations and clinical decisions.
[0063] like Figure 1 As shown, specifically including:
[0064] Obtain historical case data and extract the influencing factors of massive bleeding during CSEP surgery from the historical case data;
[0065] Use at least two machine learning methods to screen the influencing factors, and determine the predictor variables after taking the intersection of the variable screening results;
[0066] Based on the predictive variables, several machine learning methods were used to build models and predict massive bleeding during CSEP. The optimal prediction model was obtained by evaluating the prediction results of each model.
[0067] Receive parameter information of the prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
[0068] In this embodiment, historical case data of 1,680 CSEP patients treated in a hospital between January 1, 2008 and December 31, 2023 are obtained, and then the data are preprocessed, categorical variables are treated as dummy variables, near-zero variance variables are removed, numerical variables are standardized, and missing data are processed using the median interpolation method.
[0069] Also, after obtaining the historical case data, the historical case data were cleaned according to the preset exclusion rules to obtain the patients who could finally be included in the research analysis, including the group with major bleeding risk and the group without major bleeding risk, and they were randomly divided into a training set of 80% for training and a test set of 20% for testing (internal validation).
[0070] Among them, the exclusion rules include: excluding patients who have received pretreatment such as drug therapy, uterine artery embolization, curettage, etc., excluding patients who cannot exclude ectopic pregnancy and gestational trophoblastic disease, excluding patients who cannot tolerate surgery, and excluding patients with other malignant tumors. Figure 2 Impact diagram of the main predictor variables on intraoperative massive bleeding.
[0071] In this example, 20 influencing factors affecting massive bleeding during CSEP were determined from historical case data, including demographics, reproductive history, medical history, clinical symptoms, and ultrasound examination characteristics;
[0072] Then, the random forest method, LASSO regression method, Boruta feature selection method and extreme gradient boosting tree XGBoost method were used to screen the variables of the 20 influencing factors.
[0073] Among them, Figure 3 The figure shows the ranking of the importance of the predictive variables screened by the random forest method. The horizontal axis is from left to right: average diameter of the gestational sac, thinnest thickness of the anterior uterine wall, gestational age, preoperative HCG level, preoperative HGB level, time since the last cesarean section, age of onset, duration of preoperative vaginal bleeding, number of pregnancies, blood flow grade, number of miscarriages, whether there is placenta accreta in early pregnancy, number of days of abdominal pain, number of previous cesarean sections, symptoms before admission, whether there is uterine arteriovenous fistula, type of gestational sac or mass, presence of fetal bud, presence of fetal heart rate, and whether there is a history of CSEP.
[0074] like Figure 4-6 Shown are schematic diagrams of importance ranking of predictor variables selected by LASSO regression method, Boruta feature selection method and XGBoost method.
[0075] Finally, the variables screened by the four machine methods were intersected to determine the nine predictive variables used for model building prediction, including: gestational age (ie, gestational age), duration of vaginal bleeding before surgery, average diameter of gestational sac, thinnest thickness of anterior uterine wall, preoperative HGB level, preoperative HCG level, whether there is uterine arteriovenous fistula, whether there is placenta accreta in early pregnancy, and blood flow grade. Figure 7 shown.
[0076] In this embodiment, 8 types of machine learning are selected for modeling and used to predict massive bleeding during CSEP surgery, including: NB classifier, artificial neural network ANN, decision tree DT, KNN classification algorithm, logistic regression LR method, random forest RF, support vector machine SVM and extreme gradient boosting tree XGBoost method.
[0077] Based on the 9 predictive variables obtained through screening, the above 8 machine learning methods were used to build models and predict massive bleeding during CSEP surgery, such as Figure 8-10 The following are the accuracy, ROC curve, and DCA curve diagrams of 8 machine learning models. In order to optimize the prediction model, all models were developed using 10-fold internal cross-validation.
[0078] At the same time, the prediction results of the eight models were evaluated and compared using the evaluation indicators of accuracy, AUC and decision curve, thereby obtaining the optimal prediction model;
[0079] Finally, internal and external validation of the optimal prediction model were performed using sensitivity, specificity, Matthews correlation coefficient, and F1 score.
[0080] In the training set of 8 machine learning models, from the perspective of accuracy comparison, the NB model has the largest accuracy of 0.876 (95% [CI] 0.854-0.898), and the LR model is 0.817 (95% [CI] 0.781-0.837); from the perspective of AUC value comparison, the NB model is 0.882 (95% [CI] 0.860-0.904), slightly lower than the maximum value of 0.889 (95% [CI] 0.867-0.911) of the RF model; from the perspective of net yield comparison, the maximum threshold probability of the NB model is 0.41, and the maximum yield is 7.01%, which is higher than any other model.
[0081] In internal validation, the NB model achieved an accuracy of 0.869 (95% [CI] 0.847–0.891), an AUC of 0.821 (95% [CI] 0.800–0.842), a sensitivity of 0.938, a specificity of 0.422, a Matthews correlation coefficient of 0.392, and an F1 score of 0.925 on the test set.
[0082] In external validation, the NB model had an accuracy of 0.936 (95% [CI] 0.946-0.994), an AUC of 0.853 (95% [CI] 0.832-0.874), a sensitivity of 0.954, a specificity of 0.5, a Matthews correlation coefficient of 0.365, and an F1 score of 0.966 on the external cohort.
[0083] This shows that in external validation, the model has excellent ability to identify CSEP patients with intraoperative hemorrhage. CSEP intraoperative hemorrhage is a serious complication of surgery. Once diagnosed, active measures should be taken to treat it. Therefore, the false negative (missed diagnosis rate) or false positive (misdiagnosis rate) results of the prediction model are equally important, which requires the model to have a higher accuracy. It can be seen that the NB model has better performance. When the cutoff value is 0.968 in the training set, the sensitivity is 0.769, the specificity is 0.868, and the Youden index is 0.637.
[0084] In this embodiment, based on the above interactive CSEP intraoperative massive bleeding prediction method, an interpretable CSEP intraoperative massive bleeding risk prediction system based on ML method is developed and externally verified, and deployed to the cloud, whereby, after receiving the parameter information of the prediction variable input by the interactive interface, the prediction result of CSEP intraoperative massive bleeding is obtained based on the optimal prediction model, and the prediction result is fed back to the interactive interface, as shown in Figure 11 (a)-(c).
[0085] The specific deployment process of the interactive risk prediction system includes:
[0086] iBreakDown supports variable interaction and interactive interpretation with D3.js visualization. By generating a model interpreter, attaching an importance coefficient to each predictor variable, the sum of which is close to the model's prediction, and using the input predictor variables to explain the model's results, not only can the prediction of the specified individual be made, but also the partial dependence diagram (PDP) of the specified variable can be generated to visualize the importance of the predictor variable.
[0087] Shiny and ShinyDashboard components can automatically and reactively bind inputs and outputs of R code, and assign the selected 9 prediction variables and the optimal prediction model to the graphical interface (UI) and server for cloud deployment.
[0088] Therefore, clinical staff can open the interactive system address through a browser when connected to the Internet, and enter the parameter information of the nine predictive variables of a specific CSEP patient according to the system prompts. The system can then display the patient's probability of massive bleeding during CSEP surgery in real time.
[0089] It should be noted that all data is obtained in compliance with laws and regulations and user consent, and the data is used legally.
[0090] Example 2
[0091] This embodiment provides an interactive CSEP intraoperative massive bleeding prediction system, including:
[0092] An acquisition module is configured to acquire historical case data and extract influencing factors affecting massive bleeding during CSEP surgery from the historical case data;
[0093] A variable screening module is configured to use at least two machine learning methods to screen the influencing factors, and determine the prediction variable after taking the intersection of the variable screening results;
[0094] A modeling module is configured to use several machine learning methods to respectively model and predict massive bleeding during CSEP based on the predictive variables, and obtain the optimal prediction model by evaluating the prediction results of each model;
[0095] The prediction module is configured to receive parameter information of prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
[0096] As an optional implementation method, the random forest method, LASSO regression method, Boruta feature selection method and XGBoost method are used to screen the variables of the influencing factors respectively.
[0097] As an optional embodiment, the determined predictive variables include: gestational age, duration of preoperative vaginal bleeding, average diameter of gestational sac, thinnest thickness of anterior uterine wall, preoperative HGB level, preoperative HCG level, presence of uterine arteriovenous fistula, presence of early pregnancy placenta accreta and blood flow grade.
[0098] As an optional implementation, several machine learning methods used in modeling include NB classifier, artificial neural network, decision tree, KNN classification, logistic regression, random forest, support vector machine and XGBoost method; the optimal prediction model obtained after evaluation is the prediction model constructed using the NB classifier.
[0099] As an optional implementation, when several machine learning methods are used for modeling, each model uses ten-fold internal cross validation.
[0100] As an optional implementation, when evaluating the prediction results of each model, accuracy, AUC and decision curve are used as evaluation indicators.
[0101] As an optional implementation, after the optimal prediction model is obtained, the sensitivity, specificity, Matthews correlation coefficient and F1 score are used to perform internal and external validation on the optimal prediction model.
[0102] As an optional implementation method, the optimal prediction model is deployed to the cloud to receive parameter information of the prediction variables input by the interactive interface and feed back the prediction results to the interactive interface; the deployment process includes: model visualization based on the iBreakDown component, and assigning the screened prediction variables and the optimal prediction model to the graphical interface and server for cloud deployment based on Shiny and ShinyDashboard components.
[0103] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.
[0104] In further embodiments, there is also provided:
[0105] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.
[0106] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0107] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.
[0108] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.
[0109] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.
[0110] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.
[0111] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.
[0112] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.
[0113] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.
[0114] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0115] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. An interactive CSEP intraoperative massive bleeding prediction method, characterized in that: include: Obtain historical case data and extract the influencing factors of massive bleeding during CSEP surgery from the historical case data; Use at least two machine learning methods to screen the influencing factors, and determine the predictor variables after taking the intersection of the variable screening results; Based on the predictive variables, several machine learning methods were used to build models and predict massive bleeding during CSEP. The optimal prediction model was obtained by evaluating the prediction results of each model. Receive parameter information of the prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
2. The interactive CSEP intraoperative massive bleeding prediction method according to claim 1, characterized in that: The random forest method, LASSO regression method, Boruta feature selection method and XGBoost method were used to screen the influencing factors.
3. The interactive CSEP intraoperative massive bleeding prediction method according to claim 1, characterized in that: The identified predictive variables included gestational age, duration of preoperative vaginal bleeding, mean diameter of gestational sac, thinnest thickness of anterior uterine wall, preoperative HGB level, preoperative HCG level, presence of uterine arteriovenous fistula, presence of early placenta accreta, and blood flow grade.
4. The interactive CSEP intraoperative massive bleeding prediction method according to claim 1, characterized in that: Several machine learning methods used in modeling include NB classifier, artificial neural network, decision tree, KNN classification, logistic regression, random forest, support vector machine and XGBoost method; the optimal prediction model obtained after evaluation is the prediction model built using NB classifier.
5. The interactive CSEP intraoperative massive bleeding prediction method according to claim 1, characterized in that: When modeling with several machine learning methods, each model used ten-fold internal cross-validation; When evaluating the prediction results of each model, accuracy, AUC and decision curve were used as evaluation indicators; After obtaining the optimal prediction model, sensitivity, specificity, Matthews correlation coefficient and F1 score were used to perform internal and external validation of the optimal prediction model.
6. The interactive CSEP intraoperative massive bleeding prediction method according to claim 1, characterized in that: Deploy the optimal prediction model to the cloud to receive parameter information of the prediction variables input by the interactive interface and feed back the prediction results to the interactive interface; The deployment process includes: model visualization based on the iBreakDown component, and assigning the selected prediction variables and the optimal prediction model to the graphical interface and server for cloud deployment based on the Shiny and ShinyDashboard components.
7. An interactive CSEP intraoperative massive bleeding prediction system, characterized in that: include: An acquisition module is configured to acquire historical case data and extract influencing factors affecting massive bleeding during CSEP surgery from the historical case data; A variable screening module is configured to use at least two machine learning methods to screen the influencing factors, and determine the prediction variable after taking the intersection of the variable screening results; A modeling module is configured to use several machine learning methods to respectively model and predict massive bleeding during CSEP based on the predictive variables, and obtain the optimal prediction model by evaluating the prediction results of each model; The prediction module is configured to receive parameter information of prediction variables input by the interactive interface, obtain the prediction result of massive bleeding during CSEP surgery based on the optimal prediction model, and feed back the prediction result to the interactive interface.
8. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 6 is completed.
9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 6.
10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.
Citation Information
Patent Citations
A biomedicine determinant attribute selecting method
CN107169284A
Visualization system for prediction of gynecological neoplastic disease risks
CN111899888A
Model and system for predicting massive hemorrhage in uterine scar pregnancy, and terminal equipment
CN117912681A
Method and system for constructing pulmonary infection risk prediction model based on machine learning
CN118039177A
Disease key molecule and module screening method for integrating multi-omics data based on deep learning
CN118173254A
