Method and device for predicting thrombolysis risk of myocardial infarction patient and processing equipment

Abstract

The invention provides a myocardial infarction patient thrombolysis risk prediction method, a myocardial infarction patient thrombolysis risk prediction device and processing equipment, which are used for creating a set of complete myocardial infarction patient thrombolysis risk prediction model processing scheme from two stages before and after thrombolysis, and continuing to optimize from other details, so that the thrombolysis risk prediction accuracy is improved. Therefore, the thrombolysis risk of the myocardial infarction patient can be efficiently and precisely predicted, precise and reliable data reference is provided for clinical work of the myocardial infarction patient in a hospital application scene, and better medical quality and safety guarantee can be provided for the myocardial infarction patient.

Description

Technical Field

[0001] The present application relates to the field of medical technology, and in particular to a method, device and processing equipment for predicting the risk of thrombolysis in patients with myocardial infarction. Background Art

[0002] In the hospital setting, thrombolytic therapy is a major treatment method in clinical work for patients with myocardial infarction (short for myocardial infarction). Commonly used thrombolytic therapy drugs include urokinase and plasmin, etc., to dissolve and eliminate the formed blood clots.

[0003] However, thrombolytic therapy may cause side effects, which may even lead to death in severe cases. Therefore, when thrombolytic therapy is performed on patients with myocardial infarction, the corresponding thrombolytic risk needs to be assessed. For patients with myocardial infarction who are at higher risk, whether to perform thrombolytic therapy obviously needs to be treated with caution.

[0004] The inventors of the present application have discovered that, in the current clinical work on myocardial infarction patients, thrombolytic risk assessment generally relies on the work experience of medical staff, which obviously has unstable problems in terms of efficiency and accuracy. The thrombolytic risk prediction model that follows the idea of ​​configuring a machine learning model to predict the thrombolytic risk of myocardial infarction patients and is configured based on sample data of myocardial infarction patients who have received thrombolytic treatment also has the problem of limited accuracy. Summary of the invention

[0005] The present application provides a method, device and processing equipment for predicting the thrombolytic risk of myocardial infarction patients, which are used to create a complete processing solution for the thrombolytic risk prediction model of myocardial infarction patients starting from the two stages before and after thrombolysis, and continue to optimize from other details. In this way, the thrombolytic risk of myocardial infarction patients can achieve efficient and high-precision prediction goals, provide accurate and reliable data references for the clinical work of myocardial infarction patients in hospital application scenarios, and thus provide better medical quality and safety guarantees for myocardial infarction patients.

[0006] In a first aspect, the present application provides a method for predicting the risk of thrombolysis in patients with myocardial infarction, the method comprising:

[0007] Obtain the target characteristic variables of the current thrombolysis stage of the myocardial infarction patient whose thrombolysis risk is to be predicted, wherein the current thrombolysis stage is the pre-thrombolysis stage or the post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables;

[0008] Inputting the target characteristic variable into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by the characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used to predict the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model;

[0009] Based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients, corresponding risk reminders are output.

[0010] In a second aspect, the present application provides a device for predicting the risk of thrombolysis in patients with myocardial infarction, the device comprising:

[0011] An acquisition unit is used to acquire a target characteristic variable of a myocardial infarction patient whose thrombolysis risk is to be predicted at a current thrombolysis stage, wherein the current thrombolysis stage is a pre-thrombolysis stage or a post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables;

[0012] A prediction unit, used for inputting the target characteristic variables into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used for predicting the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model;

[0013] The reminder unit is used to output corresponding risk reminders based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients.

[0014] In a third aspect, the present application provides a processing device including a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method provided in the first aspect of the present application is executed.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a plurality of instructions suitable for loading by a processor to execute the method provided in the first aspect of the present application.

[0016] From the above content, it can be concluded that the present application has the following beneficial effects:

[0017] In response to the thrombolytic risk assessment needs of myocardial infarction patients, this application has created a complete treatment plan for the thrombolytic risk prediction model for myocardial infarction patients, starting from the two stages before and after thrombolysis, and continued to optimize other details. In this way, the thrombolytic risk of myocardial infarction patients can achieve efficient and high-precision prediction goals, and provide accurate and reliable data references for the clinical work of myocardial infarction patients in hospital application scenarios, thereby providing better medical quality and safety guarantees for myocardial infarction patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A schematic diagram of a process for predicting the thrombolytic risk in patients with myocardial infarction in this application;

[0020] Figure 2 A schematic diagram of the structure of a device for predicting thrombolytic risk in patients with myocardial infarction according to the present application;

[0021] Figure 3 A schematic diagram of the structure of the processing equipment of this application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0023] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0024] The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.

[0025] Before introducing the method for predicting thrombolytic risk in patients with myocardial infarction provided by this application, the background content involved in this application is first introduced.

[0026] The method, device and computer-readable storage medium for predicting thrombolytic risk in myocardial infarction patients provided in the present application can be applied to processing equipment to create a complete processing solution for thrombolytic risk prediction model for myocardial infarction patients starting from the two stages before and after thrombolysis, and continue to optimize other details. In this way, the thrombolytic risk of myocardial infarction patients can achieve efficient and high-precision prediction goals, provide accurate and reliable data reference for the clinical work of myocardial infarction patients in hospital application scenarios, and thus provide better medical quality and safety guarantees for myocardial infarction patients.

[0027] The method for predicting the risk of thrombolysis in patients with myocardial infarction mentioned in this application can be implemented by a device for predicting the risk of thrombolysis in patients with myocardial infarction, or a server, physical host, or user equipment (UE) and other different types of processing equipment that integrate the device for predicting the risk of thrombolysis in patients with myocardial infarction. Among them, the device for predicting the risk of thrombolysis in patients with myocardial infarction can be implemented in hardware or software, and the UE can be a terminal device such as a smart phone, tablet computer, laptop computer, desktop computer, or personal digital assistant (PDA), and the processing device can be set in the form of a device cluster.

[0028] It is understandable that, considering that the present application scheme is usually a data processing carried out on the basis of existing patient data (characteristic variables), the processing equipment for executing the method for predicting the risk of thrombolysis for myocardial infarction patients provided by the present application, or the processing equipment equipped with the application service of the method for predicting the risk of thrombolysis for myocardial infarction patients provided by the present application, only needs to have the required data processing capabilities in actual applications, and can be any type of processing equipment such as a server, a physical host or even a UE.

[0029] If the specific collection of patient data (characteristic variables) is also involved, the processing equipment needs to involve the equipment configuration of the corresponding collection part, such as integrating other instruments / equipment / systems, or configuring the corresponding data collection module.

[0030] In addition, the present application also involves the application of a thrombolytic risk prediction model for myocardial infarction patients, and behind this, it may also involve the preliminary / preliminary training of the thrombolytic risk prediction model for myocardial infarction patients. In this regard, the processing device may also be configured with a first device part for model training and a second device part for model application based on the application needs of splitting the model training link and the model application link.

[0031] For example, the first device part can be background equipment such as laboratories in hospital application scenarios. After the model training is completed, it can be transmitted to the equipment on the medical staff side, so that the corresponding prediction work of thrombolytic risk of myocardial infarction patients can be performed on-site to meet the risk prediction needs on-site.

[0032] It can be seen from this that the specific equipment deployment form and equipment type involved in processing equipment are relatively flexible in actual situations and can be flexibly configured according to actual needs.

[0033] Next, we will begin to introduce the method for predicting thrombolytic risk in patients with myocardial infarction provided by this application.

[0034] First, see Figure 1 , Figure 1 A schematic diagram of a process of predicting the thrombolytic risk of patients with myocardial infarction in the present application is shown. The method for predicting the thrombolytic risk of patients with myocardial infarction provided in the present application may specifically include the following steps S101 to S103:

[0035] Step S101, obtaining target characteristic variables of the current thrombolysis stage of the myocardial infarction patient whose thrombolysis risk is to be predicted, wherein the current thrombolysis stage is the pre-thrombolysis stage or the post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables;

[0036] It can be understood that, corresponding to the model input requirements of the thrombolytic risk prediction model for myocardial infarction patients configured in the present application, when the thrombolytic risk prediction processing of the myocardial infarction patient whose thrombolytic risk is to be predicted is carried out through the thrombolytic risk prediction model for myocardial infarction patients, it is necessary to obtain the characteristic variables of the current thrombolytic node. For the convenience of explanation, the characteristic variables obtained here are recorded as target characteristic variables.

[0037] Among them, the characteristic variables targeted by the thrombolytic risk prediction model for myocardial infarction patients in this application may involve the two stages before and after thrombolysis, and the specific characteristic variables involved in different stages are also different. In this way, in both the pre-thrombolytic stage and the post-thrombolytic stage, efficient and high-precision thrombolytic risk prediction effects can be achieved for patients of this type or disease type such as myocardial infarction patients.

[0038] Specifically, on the one hand, the pre-thrombolysis stage may involve 19 first characteristic variables, and on the other hand, the post-thrombolysis stage may involve 35 characteristic variables, that is, in addition to the 19 first characteristic variables involved in the pre-thrombolysis stage, it also involves 16 second characteristic variables.

[0039] Furthermore, it can be understood that these characteristic variables are usually ready-made data recorded in hospital-related systems, and the data acquisition method behind these characteristic variables can obviously be adaptively adjusted according to the specific content of the characteristic variables.

[0040] Specifically, data acquisition methods may involve demographics, medical history, vital signs, chief complaints, laboratory tests, ultrasound examinations, coronary angiography, drug treatment, or clinical evaluation.

[0041] Step S102, inputting the target characteristic variable into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used to predict the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model;

[0042] After obtaining the corresponding target characteristic variables of the current myocardial infarction patient, it is obvious that they can be input / imported into the thrombolytic risk prediction model for myocardial infarction patients pre-configured in this application, so that it can carry out the corresponding thrombolytic risk prediction processing for myocardial infarction patients.

[0043] It is understandable that the pre-configuration work of the thrombolytic risk for myocardial infarction patients mainly involves the training of the model.

[0044] In model training work, it usually involves:

[0045] Configure the characteristic variables of sample myocardial infarction patients;

[0046] Label the characteristic variables of sample myocardial infarction patients. The labeling results are the theoretical prediction results of thrombolysis risk in myocardial infarction patients, which guide the model training and use;

[0047] Based on the characteristic variables of the labeled sample myocardial infarction patients, the initial model is trained, and the trained model is the thrombolytic risk prediction model for myocardial infarction patients. In each round of model training, a set of labeled sample myocardial infarction patients' characteristic variables are input into the model for thrombolytic risk prediction processing for myocardial infarction patients to achieve forward propagation. Then, based on the thrombolytic risk prediction results of myocardial infarction patients output by the model and the corresponding annotation results, the loss function is calculated, and the model parameters are optimized according to the loss function calculation results to achieve reverse propagation. In this way, when the preset model training requirements such as training time, training times or prediction accuracy are met, the model training can be completed, and a thrombolytic risk prediction model for myocardial infarction patients that can be put into practical use is obtained.

[0048] Among them, it can be understood that in actual applications, the model structure involved in the thrombolytic risk prediction model for myocardial infarction patients itself, as well as the training framework and loss function involved in the model training process, can adopt existing solutions, or solutions optimized from existing solutions, or novel self-developed solutions. These are all possible under actual circumstances and can be configured according to actual conditions.

[0049] In this way, when the thrombolytic risk prediction model for myocardial infarction patients completes the thrombolytic risk prediction processing for myocardial infarction patients under the current circumstances, it will output the corresponding thrombolytic risk prediction results for myocardial infarction patients, and at this time the thrombolytic risk prediction results for the myocardial infarction patients can be extracted.

[0050] Step S103, outputting corresponding risk reminders based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients.

[0051] When the thrombolysis risk prediction processing result of the myocardial infarction patient output by the thrombolysis risk prediction model for the myocardial infarction patient is extracted, the thrombolysis risk prediction processing result for the myocardial infarction patient can be processed according to the preset data usage strategy.

[0052] For example, it can be stored, forwarded, output a prompt of completed processing, output a pre-adapted myocardial infarction diagnosis / thrombolysis plan, or perform further analysis and processing. These are all possible, and the specific strategy content can be configured according to clinical work and actual conditions.

[0053] For example, the present application may specifically involve a risk reminder mechanism, which may output risk reminders of different contents or in different forms according to the predicted different degrees of thrombolytic risk, so as to prompt medical staff to decide on response measures suitable for the current situation within an appropriate time, such as suspending thrombolytic therapy, arranging thrombolytic therapy as soon as possible, changing the thrombolytic therapy plan, arranging corresponding observation / treatment plans after thrombolytic therapy, and clearly informing family members of the degree of risk, etc. The specific response measures involved can be flexibly configured, and the present application does not specifically limit them here.

[0054] Among them, corresponding to the risk reminder mechanism here, the processing equipment of this application can use its own equipment or external equipment to provide risk reminders on site in different specific reminder methods such as text, images, voice, lights or vibrations, or can use remote reminder methods such as voice calls, emails, system messages, etc. to provide risk reminders. These settings can obviously be more flexible in actual situations and can be adjusted according to actual needs.

[0055] Of course, specific risk reminders can also be done manually. The processing device only needs to provide the specific reminder content required by medical staff to remind the current user of the risk. This is also possible in actual situations.

[0056] In addition, the specific risk reminder content output can be general risk reminder content, risk reminder content that is adapted to the current specific risk situation, or personalized / customized risk reminder content, all of which can be configured according to actual needs.

[0057] So, from Figure 1 It can be seen from the illustrated embodiments that, in response to the thrombolytic risk assessment needs of myocardial infarction patients, this application has created a complete treatment plan for the thrombolytic risk prediction model for myocardial infarction patients, starting from the two stages before and after thrombolysis, and continued to optimize other details. In this way, the thrombolytic risk of myocardial infarction patients can achieve efficient and high-precision prediction goals, and provide accurate and reliable data references for the clinical work of myocardial infarction patients in hospital application scenarios, thereby providing better medical quality and safety guarantees for myocardial infarction patients.

[0058] Continue to the above Figure 1 Each step of the illustrated embodiment and possible implementation methods thereof in practical applications are described in detail.

[0059] Corresponding to the configuration of the characteristic variables of the current myocardial infarction patient to be obtained in this application, and also corresponding to the configuration of the characteristic variables targeted by the thrombolytic risk prediction model for myocardial infarction patients in this application, the following processing contents may be included in the early / preliminary scheme design:

[0060] For the configuration target of feature variables, this application can start from patient data with different feature variables. Taking the Electronic Health Record (EHR) data as an example, for the original EHR data, the original feature variables can be collected according to the patient's serial number (SN), including the patient's demographic information (gender, age, etc.), medical history (coronary heart disease, diabetes, hypertension, hyperlipidemia, cerebrovascular disease, etc.), vital signs (systolic blood pressure, diastolic blood pressure, respiratory rate, body temperature, etc.), chief complaints (chest pain, back pain, chest tightness, syncope, pain, fever, etc.), laboratory tests (troponin, total cholesterol, albumin, globulin, direct bilirubin, basophils, creatinine, etc.), ultrasound examination (regional ventricular wall motion abnormalities, ventricular aneurysm, atrial septal defect, pericardial effusion, mitral regurgitation, atrial and ventricular diameter, ejection fraction, etc.), electrocardiographic characteristics (ST-T changes, ST-segment elevation, ST-segment depression, ST-segment shortening, ST-segment prolongation, maximum heart rate, minimum heart rate, arrhythmias including atrial premature beats, atrial tachycardia, etc.), coronary angiography (location and severity of the left main coronary artery, preoperative and postoperative TIMI grade, etc.), drug treatment (antiplatelet therapy, anticoagulants, thrombolytics, ACEI, ARB, beta-blockers, etc.) and clinical evaluation (TIMI score, GRACE score, ACUITY score and CRUSADE score), etc.

[0061] For these original feature variables, natural language processing, machine learning technology and clear logical rules can be used to standardize and structure them, and finally 471 feature variables are determined.

[0062] Next, these 471 characteristic variables were divided into two categories: before and after thrombolysis.

[0063] At this time, characteristic variables with a missing rate greater than 35% were excluded, as well as binary variables with a consistency greater than 90% and continuous characteristic variables with a variance less than 0.1, leaving 141 characteristic variables.

[0064] Subsequently, the chi-square test was applied to the categorical characteristic variables among the 141 characteristic variables, and the T test was used for the continuous variables among the 141 characteristic variables to screen out characteristic variables with a p value less than 0.05.

[0065] Finally, continuous variables are filled with their means and discrete variables are filled with their medians.

[0066] Taking into account the redundancy between characteristic variables, characteristic variables with correlation higher than 0.3 were deleted according to clinical research significance, and finally 35 characteristic variables were selected.

[0067] Among them, 19 characteristic variables came from the pre-thrombolysis stage and 16 characteristic variables came from the post-thrombolysis stage.

[0068] The 19 first characteristic variables involved in the pre-thrombolysis stage may specifically include, as an exemplary embodiment:

[0069] Age (1), surgical history (2), smoking history (3), drinking history (4), pregnancy history (5), chest pain (6), weakness or fatigue (7), edema or swelling (8), hypertension (9), diabetes (10), anticoagulants (11), ST-T changes before thrombolysis (12), pericardial effusion before thrombolysis (13), diastolic blood pressure before thrombolysis (14), pulse before thrombolysis (15), body temperature before thrombolysis (16), in-hospital delay time (17), onset time (18), amino-terminal pro-brain natriuretic peptide before thrombolysis (19).

[0070] The 16 second characteristic variables involved in the post-thrombolysis stage may specifically include, as an exemplary embodiment:

[0071] Pericardial effusion after thrombolysis (1), diastolic blood pressure after thrombolysis (2), systolic blood pressure after thrombolysis (3), pulse after thrombolysis (4), body temperature after thrombolysis (5), hemoglobin content after thrombolysis (6), creatinine after thrombolysis (7), uric acid after thrombolysis (8), albumin after thrombolysis (9), total protein after thrombolysis (10), neutrophil percentage after thrombolysis (11), total bilirubin after thrombolysis (12), monocyte count after thrombolysis (13), red blood cell distribution width after thrombolysis (14), amino-terminal pro-brain natriuretic peptide after thrombolysis (15), cardiac troponin after thrombolysis (16).

[0072] It should be understood that the series of characteristic variables or indicators involved above are existing indicators in the hospital scenario, so the specific contents of these characteristic variables themselves will not be explained in detail here.

[0073] In addition, for the thrombolytic risk prediction model for myocardial infarction patients designed by the present application, the present application can also adopt a model fusion mechanism in the detailed operation. That is to say, corresponding to different aspects of thrombolytic risk prediction needs, the present application can be configured with different sub-prediction models, and these sub-prediction models are fused into an overall thrombolytic risk prediction model for myocardial infarction patients. In this way, a more refined and accurate prediction effect can be achieved for the thrombolytic risk of myocardial infarction patients in different situations.

[0074] In this regard, this application has designed two sets of model design solutions.

[0075] In the first model design, the thrombolytic risk of myocardial infarction patients may involve three risk types, which may specifically involve a composite complication risk type, a bleeding risk type, and a death risk type;

[0076] Corresponding to the model training work, the present application method may further include:

[0077] By annotating the characteristic variables of sample myocardial infarction patients, training was carried out based on pre-set different types of initial models for the two stages, namely the pre-thrombolysis stage and the post-thrombolysis stage, as well as for the three risk types. During the training process, the optimal model parameters were trained through grid search and 5-fold cross validation, and the models responsible for risk prediction corresponding to different stages and risk types were selected through preset evaluation indicators to form a thrombolysis risk prediction model for myocardial infarction patients obtained by integrating multiple models. Among them, the different types of initial models included six models: initial logistic regression (LR), initial random forest (RF), initial XGBoost, initial gradient boosting decision tree (GBDT), initial K-nearest neighbor (KNN) and initial support vector machine (SVM).

[0078] In this embodiment, it can be seen that the risk prediction target of thrombolysis for myocardial infarction patients performed in this application may involve two stages, and each stage may be further divided into three results, such as: combined complications before and after thrombolysis, bleeding, and death.

[0079] Among them, complex complications are determined according to the patient's discharge diagnosis report, including reinfarction, ischemic shock, cardiogenic shock and congestive heart failure; bleeding is divided into cerebral hemorrhage, hematoma, gastrointestinal bleeding, coronary artery bleeding, etc. according to the discharge diagnosis report and the head, heart and gastrointestinal endoscopy reports; death refers to the all-cause mortality rate, which is a clearly recorded death during hospitalization according to the discharge diagnosis report.

[0080] In this case, the present application can specifically adopt the six machine learning models mentioned above. For the pre-thrombolytic stage, 19 first characteristic variables are used to train and verify the six models. In the model configuration link of each group, the model prediction effect is measured based on evaluation indicators such as AUC, and the model that achieves the best prediction effect in the three risk types is selected.

[0081] Similarly, in the post-thrombolysis stage, the six models were trained and verified using 35 characteristic variables collected in the two stages before and after thrombolysis. In the model configuration link of each group, the model prediction effect was measured based on evaluation indicators such as AUC, and the model that achieved the best prediction effect in the three risk types was selected.

[0082] Thus, the overall thrombolysis risk prediction model for myocardial infarction patients is obtained by integrating six models, including six models corresponding to the two stages before and after thrombolysis and three risk types.

[0083] Among them, as an example of a training framework, the sample set (i.e., the characteristic variables of the labeled sample myocardial infarction patients) can be divided into a training set (80%) and a test set (20%); in order to select the six best models, the present application can use grid search and 5-fold cross validation to optimize the hyperparameters of the model to maximize the AUC; in order to solve the problem of data imbalance in the three results, SMOTE sampling was performed on the training set. During this process, multiple models were developed for different stages and different risk types. Subsequently, the test set was used to evaluate the performance indicators of the model and determine the best model.

[0084] Under the first model design scheme shown above, in actual application, by comparison, it can be found that:

[0085] (1) For the risk type of combined complications, the AUC values ​​of the models in the pre-thrombolytic stage were very similar, among which the random forest model showed the highest AUC value of 0.85 in the pre-thrombolytic stage. In the post-thrombolytic stage, the XGBoost model had the highest AUC value of 0.91, followed by the random forest and gradient boosting decision tree models with AUC values ​​of 0.8. Therefore, the random forest model was confirmed as the best model in the pre-thrombolytic stage, and the XGBoost model was confirmed as the best model in the post-thrombolytic stage.

[0086] (2) For bleeding risk type, in the pre-thrombolysis stage, the AUC value of random forest and logistic regression was the highest at 0.71, but the sensitivity and specificity of the logistic regression model were higher than those of the random forest, so the logistic regression model was selected as the best model; for the post-thrombolysis stage, the K nearest neighbor model achieved the highest AUC value of 0.78, followed by the random forest model with an AUC value of 0.71, so in the post-thrombolysis stage, the K nearest neighbor model was confirmed as the best model.

[0087] (3) For the mortality risk type, the random forest model performed best in the pre-thrombolysis stage, with the highest AUC value of 0.79. In comparison, the AUC value of the XGBoost model was close to it, which was 0.75. In the post-thrombolysis stage, the random forest model also showed the highest AUC value of 0.89, and the AUC value of the XGBoost model was 0.86, which was close to it. However, the sensitivity and F1 value of the XGBoost model were higher than those of the random forest model. Therefore, the random forest model was confirmed as the best model in the pre-thrombolysis stage, and the XGBoost model was confirmed as the best model in the post-thrombolysis stage.

[0088] Correspondingly, as an exemplary embodiment, the preset evaluation index uses the AUC index as the main evaluation index, and other indicators (such as accuracy, precision, positive predictive value, sensitivity and specificity, etc.) as secondary evaluation indicators. The thrombolytic risk prediction model for myocardial infarction patients may specifically include:

[0089] The first random forest model is responsible for risk prediction for composite complication risk types in the pre-thrombolytic stage;

[0090] The first XGBoost model responsible for risk prediction for composite complication risk types in the post-thrombolysis stage;

[0091] A logistic regression model responsible for risk prediction for bleeding risk types in the pre-thrombolytic phase;

[0092] The K-nearest neighbor model is responsible for risk prediction for bleeding risk types in the post-thrombolysis stage;

[0093] The second random forest model is responsible for risk prediction for death risk types in the pre-thrombolysis stage;

[0094] The second XGBoost model is responsible for risk prediction for mortality risk types in the post-thrombolysis stage.

[0095] In the second model design, the thrombolytic risk of myocardial infarction patients may involve two risk types, which may specifically involve bleeding risk type and death risk type;

[0096] Corresponding to the model training work, the present application method may further include:

[0097] By annotating the characteristic variables of sample myocardial infarction patients, training was carried out based on different types of pre-set initial models for bleeding risk type and death risk type, and the optimal model parameters were trained through grid search and 5-fold cross validation during the training process. The models responsible for risk prediction of bleeding risk type and death risk type were selected through preset evaluation indicators to form two models required for the thrombolytic risk prediction model for myocardial infarction patients. Among them, different types of initial models included initial logistic regression model, initial random forest model, initial XGBoost model, initial gradient boosting decision tree model and initial support vector machine model. The thrombolytic risk prediction model for myocardial infarction patients specifically outputs the model predicted risk situation through CRUSADE score, ACUITY score, GRACE score or TIMI score.

[0098] It can be noted that this application introduces the CRUSADE score, ACUITY score, GRACE score and TIMI score to better fit the hospital application scenarios, so as to better quantify the specific risk situations of the patient's death risk type and bleeding risk type.

[0099] Specifically, these four scores are existing indicators involved in quantifying thrombolytic risks under different evaluation systems in clinical work.

[0100] In addition, the CRUSADE score, ACUITY score, GRACE score and TIMI score can also be further divided into two categories: ST-segment elevation myocardial infarction (STEMI) and non-ST-segment elevation myocardial infarction (NSTEMI). Specifically, they are:

[0101] (1) In the CRUSADE score, the classification rules for STEMI are 0-30 points as low risk, 31-40 points as intermediate risk, and greater than 40 points as high risk; the classification rules for NSTEMI are 0-20 points as very low risk, 21-30 points as low risk, 31-40 points as intermediate risk, 41-50 points as high risk, and greater than 50 points as very high risk;

[0102] (2) In the ACUITY score, the classification rule for STEMI is 0-10 points for low risk, 10-14 points for intermediate risk, and greater than 14 points for high risk; the classification rule for NSTEMI is 0-9 points for low risk, 10-14 points for intermediate risk, 15-19 points for high risk, and greater than 19 points for extremely high risk. Finally, the high, medium and low scores are merged according to their respective classification rules.

[0103] (3) In the TIMI score, the classification rules for STEMI are 0-3 points for low risk, 4-6 points for intermediate risk, and 7-14 points for high risk; the classification rules for NSTEMI are 0-2 points for low risk, 3-4 points for intermediate risk, and 5-7 points for high risk;

[0104] (4) In the GRACE score, the classification rule for STEMI is 49-125 points as low risk, 126-154 points as intermediate risk, and 155-319 points as high risk; the classification rule for NSTEMI is 1-108 points as low risk, 109-140 points as intermediate risk, and 141-372 points as high risk.

[0105] It can be seen that the CRUSADE score, ACUITY score, GRACE score and TIMI score provide a clearer and more intuitive quantitative scheme for further quantifying the specific risk degree of thrombolysis risk in patients with myocardial infarction in this application.

[0106] Under the second model design scheme shown above, in actual application, by comparison, it can be found that:

[0107] (1) For the bleeding risk type under the CRUSADE score, this application first considers dividing the target variable group according to the score level. CRSADE>=3 means that extremely low risk and low risk are divided into the non-bleeding group, and medium risk, high risk and extremely high risk are divided into the bleeding group. Compared with the actual results, the prediction accuracy, specificity and sensitivity are calculated; CRUSADE>=4 can be deduced in the same way. Secondly, in the model established by this application based on the CRUSADE score, the AUC values ​​of the support vector machine model, random forest model, gradient boosting decision tree model and XGBoost model are all 0.81, and the accuracy, sensitivity and specificity of the model are the same. Therefore, after considering the ROC curve of the model, the random forest model is confirmed to be the optimal model. Although the accuracy of the random forest model is lower than CRSADE>=4, the specificity, sensitivity and F1 score of the model are higher, so the random forest model is better than simple level division.

[0108] (2) For the bleeding risk type under the ACUITY score, ACUITY>=3 means that low risk and medium risk are classified as the non-bleeding group, and high risk and very high risk are classified as the bleeding group; the same applies to ACUITY>=4. In the machine learning model established by ACUITY, all evaluation indicators are compared. Although the sensitivity of the random forest model is slightly lower than that of ACUITY>=3, the precision and accuracy of the model are higher than ACUITY>=3. Therefore, the simple ACUITY level grouping is not as effective as the random forest model in identifying bleeding risk, and the random forest model is confirmed to be the optimal model.

[0109] (3) For the types of death risk under the GRACE score, the GRACE score was divided into two different threshold groups according to the level, GRACE>=2 and GRACE>=3, and the risk of death was predicted respectively, and compared with the assessment of death risk by the machine learning model. The random forest model had the highest accuracy of 0.66 and a sensitivity of 0.50, which was lower than GRACE>=2 (0.71), but the specificity and F1 score were higher than GRACE>=2. Therefore, after comprehensive consideration, the random forest model was confirmed as the optimal model in the GRACE score.

[0110] (4) For the types of death risk under the TIMI score, the TIMI score was divided into two different threshold groups according to the level: TIMI>=2 and TIMI>=3 to predict the risk of death, and the risk of death was compared with the assessment of the machine learning model. Among the machine learning models, the logistic regression model had the highest AUC of 0.64, but its prediction accuracy was relatively lower than that of TIMI>=2 and TIMI>=3, and its sensitivity was higher than that of TIMI>=2 and TIMI>=3. Therefore, from the perspective of the identification of the risk of death, the logistic regression model was confirmed as the optimal model.

[0111] Correspondingly, as an exemplary embodiment, the preset evaluation index uses the AUC index as the main evaluation index, and other indicators (such as accuracy, precision, positive predictive value, sensitivity and specificity, etc.) as secondary evaluation indicators;

[0112] When the model predicts the risk situation through the CRUSADE score, the thrombolytic risk prediction model for myocardial infarction patients includes a first random forest model responsible for risk prediction for bleeding risk types;

[0113] When the ACUITY score is used to output the model's predicted risk, the thrombolytic risk prediction model for myocardial infarction patients includes a second random forest model responsible for risk prediction for bleeding risk types;

[0114] When the GRACE score is used to output the model's predicted risk, the thrombolysis risk prediction model for myocardial infarction patients includes a third random forest model responsible for risk prediction for the type of death;

[0115] When the model predicts risk using the TIMI score, the thrombolytic risk prediction model for myocardial infarction patients includes a logistic regression model responsible for risk prediction for the type of death.

[0116] In addition, for the fusion model obtained after completing the model training, namely the thrombolytic risk model for myocardial infarction patients, during the model training, it can also output the characteristic variables that are confirmed during the model training process to contribute to the model risk prediction effect and meet the preset conditions. For example, the top 10 characteristic variables with the largest contributions, namely the top ten predictive factors, can be confirmed.

[0117] For example, in specific operations, a radar chart can be used to more clearly and vividly display the characteristic variables that have made greater contributions.

[0118] Specifically, corresponding to the first model design scheme above, as an example, there are:

[0119] (1) For the bleeding risk type in the pre-thrombolytic stage, the most important factor affecting the outcome prediction is the patient's body temperature, followed by the patient's onset time, whether anticoagulants were used during hospitalization, whether there was a history of smoking, and age. Finally, the in-hospital delay time, pulse and diastolic blood pressure before thrombolysis, whether there was a history of drinking, and whether there were chest pain symptoms had an important impact on the outcome prediction.

[0120] (2) For the bleeding risk type in the post-thrombolytic stage, the patient's body temperature is still the most important, followed by the onset time, diastolic blood pressure after thrombolysis, whether anticoagulants are used during hospitalization, amino-terminal pro-brain natriuretic peptide after thrombolysis, whether there is a history of smoking, uric acid index after thrombolysis, body temperature after thrombolysis, and in-hospital delay time also have an important impact on the prediction of results.

[0121] (3) For the risk type of composite complications in the pre-thrombolytic stage, the patient's body temperature is the most important influencing factor, followed by onset time, in-hospital delay time and diastolic blood pressure before thrombolysis, which have a greater impact on the prediction of results. Finally, amino-terminal pro-brain natriuretic peptide, pulse before thrombolysis, history of pregnancy, age, history of hypertension and diabetes also have an important impact on the prediction of results.

[0122] (4) For the risk type of composite complications in the post-thrombolytic stage, diabetes, body temperature before thrombolysis, systolic blood pressure after thrombolysis, presence of pericardial effusion after thrombolysis, in-hospital delay time, and uric acid after thrombolysis are important. The second most important factors are whether there is a history of pregnancy, onset time, creatinine concentration after thrombolysis, and whether there are signs of edema or swelling.

[0123] (5) For the mortality risk type in the pre-thrombolytic stage, the patient's onset time is the most important influencing factor, followed by the patient's body temperature, in-hospital delay time and age, which all have an important impact on the outcome prediction.

[0124] (6) For the mortality risk type in the post-thrombolytic stage, the most important influencing factor after thrombolysis is the red blood cell distribution width after thrombolysis. Neutrophils, onset time, body temperature before thrombolysis, albumin after thrombolysis, the presence of hypertension, pulse and creatinine concentration after thrombolysis, as well as age and in-hospital delay time are also important influencing factors.

[0125] Under this setting, it is understandable that it can provide medical staff with patient indicators that require more attention and are more worthy of attention, providing an effective reference.

[0126] In addition, for those characteristic variables determined to have a large contribution, they can also be configured as specific model inputs of the corresponding models. Instead of keeping the inputs of each model in the fusion model, that is, the thrombolytic risk prediction model for myocardial infarction patients consistent, the screening of characteristic variables may also be involved in the middle, corresponding to the input requirements of specific models, so as to further improve the prediction efficiency and accuracy of thrombolytic risk for myocardial infarction patients in different aspects in terms of details.

[0127] The above is an introduction to the thrombolytic risk prediction method for myocardial infarction patients provided in this application. In order to facilitate better implementation of the thrombolytic risk prediction method for myocardial infarction patients provided in this application, this application also provides a thrombolytic risk prediction device for myocardial infarction patients from the perspective of functional modules.

[0128] See also Figure 2 , Figure 2 This is a schematic diagram of a structure of a device for predicting the risk of thrombolysis in patients with myocardial infarction in this application. In this application, the device 200 for predicting the risk of thrombolysis in patients with myocardial infarction may specifically include the following structure:

[0129] The acquisition unit 201 is used to acquire the target characteristic variables of the current thrombolysis stage of the myocardial infarction patient whose thrombolysis risk is to be predicted, wherein the current thrombolysis stage is the pre-thrombolysis stage or the post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables;

[0130] Prediction unit 202, used for inputting target characteristic variables into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used for predicting the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model;

[0131] The reminder unit 203 is used to output corresponding risk reminders based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients.

[0132] In an exemplary embodiment, the 19 first characteristic variables involved in the pre-thrombolysis stage specifically include:

[0133] Age, surgical history, smoking history, drinking history, pregnancy history, chest pain, weakness or fatigue, edema or swelling, hypertension, diabetes, anticoagulants, ST-T changes before thrombolysis, pericardial effusion before thrombolysis, diastolic blood pressure before thrombolysis, pulse before thrombolysis, body temperature before thrombolysis, in-hospital delay time, onset time, amino-terminal pro-brain natriuretic peptide before thrombolysis.

[0134] In another exemplary embodiment, the 16 second characteristic variables involved in the post-thrombolysis stage specifically include:

[0135] Pericardial effusion after thrombolysis, diastolic blood pressure after thrombolysis, systolic blood pressure after thrombolysis, pulse after thrombolysis, body temperature after thrombolysis, hemoglobin content after thrombolysis, creatinine after thrombolysis, uric acid after thrombolysis, albumin after thrombolysis, total protein after thrombolysis, neutrophil percentage after thrombolysis, total bilirubin after thrombolysis, monocyte count after thrombolysis, red blood cell distribution width after thrombolysis, amino-terminal pro-brain natriuretic peptide after thrombolysis, and cardiac troponin after thrombolysis.

[0136] In another exemplary embodiment, in the first set of model design solutions, the thrombolytic risk of myocardial infarction patients involves three risk types, and the three risk types specifically involve a composite complication risk type, a bleeding risk type, and a death risk type;

[0137] The apparatus further comprises a training unit 204, configured to:

[0138] By annotating the characteristic variables of sample myocardial infarction patients, training was carried out based on different types of pre-set initial models for the two stages, namely the pre-thrombolysis stage and the post-thrombolysis stage, as well as for the three risk types. During the training process, the optimal model parameters were trained through grid search and 5-fold cross validation. The models responsible for risk prediction corresponding to different stages and risk types were selected through preset evaluation indicators to form a thrombolysis risk prediction model for myocardial infarction patients obtained by integrating multiple models. Among them, different types of initial models included initial logistic regression model, initial random forest model, initial XGBoost model, initial gradient boosting decision tree model, initial K nearest neighbor model and initial support vector machine model.

[0139] In another exemplary embodiment, the preset evaluation index uses the AUC index as the main evaluation index and other indexes as secondary evaluation indexes. The thrombolytic risk prediction model for myocardial infarction patients specifically includes:

[0140] The first random forest model is responsible for risk prediction for composite complication risk types in the pre-thrombolytic stage;

[0141] The first XGBoost model responsible for risk prediction for composite complication risk types in the post-thrombolysis stage;

[0142] A logistic regression model responsible for risk prediction for bleeding risk types in the pre-thrombolytic phase;

[0143] The K-nearest neighbor model is responsible for risk prediction for bleeding risk types in the post-thrombolysis stage;

[0144] The second random forest model is responsible for risk prediction for death risk types in the pre-thrombolysis stage;

[0145] The second XGBoost model is responsible for risk prediction for mortality risk types in the post-thrombolysis stage.

[0146] In another exemplary embodiment, in the second set of model design solutions, the thrombolytic risk of myocardial infarction patients involves two risk types, and the two risk types specifically involve a bleeding risk type and a death risk type;

[0147] The apparatus further comprises a training unit 204, configured to:

[0148] By annotating the characteristic variables of sample myocardial infarction patients, training was carried out based on different types of pre-set initial models for bleeding risk type and death risk type, and the optimal model parameters were trained through grid search and 5-fold cross validation during the training process. The models responsible for risk prediction of bleeding risk type and death risk type were selected through preset evaluation indicators to form two models required for the thrombolytic risk prediction model for myocardial infarction patients. Among them, different types of initial models included initial logistic regression model, initial random forest model, initial XGBoost model, initial gradient boosting decision tree model and initial support vector machine model. The thrombolytic risk prediction model for myocardial infarction patients specifically outputs the model predicted risk situation through CRUSADE score, ACUITY score, GRACE score or TIMI score.

[0149] In another exemplary embodiment, the preset evaluation index uses the AUC index as the main evaluation index and other indexes as secondary evaluation indexes;

[0150] When the Crusade score is used to output the model's predicted risk situation, the thrombolytic risk prediction model for myocardial infarction patients includes a first random forest model responsible for risk prediction for bleeding risk types;

[0151] When the model predicts risk using the Acuity score, the thrombolysis risk prediction model for myocardial infarction patients includes a second random forest model responsible for risk prediction for bleeding risk types;

[0152] When the Grace score is used to output the model prediction risk, the thrombolysis risk prediction model for myocardial infarction patients includes a third random forest model responsible for risk prediction for death type;

[0153] When the Crusade score is used to output the model's predicted risk, the thrombolytic risk prediction model for myocardial infarction patients includes a logistic regression model responsible for risk prediction for the type of death.

[0154] This application also provides a processing device from the perspective of hardware structure, see Figure 3 , Figure 3 301, a memory 302, and an input / output device 303. The processor 301 is used to execute the computer program stored in the memory 302 to implement the following Figure 1 The steps of the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment; or, the processor 301 is used to execute the computer program stored in the memory 302 to implement the following Figure 2 Corresponding to the functions of each unit in the embodiment, the memory 302 is used to store the processor 301 executing the above Figure 1The computer program required for the method for predicting thrombolytic risk in patients with myocardial infarction in the corresponding embodiment.

[0155] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the memory 302, and executed by the processor 301 to complete the present application. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0156] The processing device may include, but is not limited to, a processor 301, a memory 302, and an input / output device 303. Those skilled in the art will appreciate that the illustration is merely an example of a processing device and does not constitute a limitation on the processing device, and may include more or fewer components than shown in the illustration, or a combination of certain components, or different components. For example, the processing device may also include a network access device, a bus, etc., and the processor 301, the memory 302, the input / output device 303, etc. are connected via a bus.

[0157] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the processing device, and uses various interfaces and lines to connect various parts of the entire device.

[0158] The memory 302 can be used to store computer programs and / or modules. The processor 301 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 302 and calling the data stored in the memory 302. The memory 302 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the processing device, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0159] When the processor 301 is used to execute the computer program stored in the memory 302, the following functions can be implemented:

[0160] Obtain the target characteristic variables of the current thrombolysis stage of the myocardial infarction patient whose thrombolysis risk is to be predicted, wherein the current thrombolysis stage is the pre-thrombolysis stage or the post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables;

[0161] Inputting the target characteristic variable into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by the characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used to predict the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model;

[0162] Based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients, corresponding risk reminders are output.

[0163] A person skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described myocardial infarction patient thrombolysis risk prediction device, processing equipment and corresponding units can refer to the following. Figure 1 The description of the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment will not be repeated here.

[0164] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0165] To this end, the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the present application as follows: Figure 1 The steps of the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment, the specific operation can be referred to as follows Figure 1 The description of the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment will not be repeated here.

[0166] The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0167] Due to the instructions stored in the computer-readable storage medium, the present application can be executed. Figure 1 The steps of the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment, therefore, the present application can be implemented as follows Figure 1 The beneficial effects that can be achieved by the method for predicting the risk of thrombolysis in patients with myocardial infarction in the corresponding embodiment are detailed in the previous description and will not be repeated here.

[0168] The above is a detailed introduction to the thrombolytic risk prediction method, device, processing equipment and computer-readable storage medium for myocardial infarction patients provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting thrombolytic risk in patients with myocardial infarction, characterized in that: The method comprises: Obtaining target characteristic variables of the current thrombolysis stage of the myocardial infarction patient whose thrombolysis risk is to be predicted, wherein the current thrombolysis stage is the pre-thrombolysis stage or the post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables; Inputting the target characteristic variable into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used to predict the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model; Based on the thrombolysis risk prediction result of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients, a corresponding risk reminder is output.

2. The method according to claim 1, characterized in that The 19 first characteristic variables involved in the pre-thrombolysis stage specifically include: Age, surgical history, smoking history, drinking history, pregnancy history, chest pain, weakness or fatigue, edema or swelling, hypertension, diabetes, anticoagulants, ST-T changes before thrombolysis, pericardial effusion before thrombolysis, diastolic blood pressure before thrombolysis, pulse before thrombolysis, body temperature before thrombolysis, in-hospital delay time, onset time, amino-terminal pro-brain natriuretic peptide before thrombolysis.

3. The method according to claim 1, characterized in that: The 16 second characteristic variables involved in the post-thrombolysis stage specifically include: Pericardial effusion after thrombolysis, diastolic blood pressure after thrombolysis, systolic blood pressure after thrombolysis, pulse after thrombolysis, body temperature after thrombolysis, hemoglobin content after thrombolysis, creatinine after thrombolysis, uric acid after thrombolysis, albumin after thrombolysis, total protein after thrombolysis, neutrophil percentage after thrombolysis, total bilirubin after thrombolysis, monocyte count after thrombolysis, red blood cell distribution width after thrombolysis, amino-terminal pro-brain natriuretic peptide after thrombolysis, and cardiac troponin after thrombolysis.

4. The method according to claim 1, characterized in that: In the first set of model design schemes, the thrombolytic risk of myocardial infarction patients involves three risk types, and the three risk types specifically involve a composite complication risk type, a bleeding risk type, and a death risk type; The method further comprises: Through the characteristic variables of the labeled sample myocardial infarction patients, training is carried out based on pre-set different types of initial models for the two stages, the pre-thrombolysis stage and the post-thrombolysis stage, and for the three risk types, and the optimal model parameters are trained through grid search and 5-fold cross validation during the training process, and the models responsible for risk prediction corresponding to different stages and risk types are selected through preset evaluation indicators to form the myocardial infarction patient thrombolysis risk prediction model obtained by fusing multiple models, wherein the different types of initial models include initial logistic regression model, initial random forest model, initial XGBoost model, initial gradient boosting decision tree model, initial K nearest neighbor model and initial support vector machine model.

5. The method according to claim 4, characterized in that The preset evaluation index uses the AUC index as the main evaluation index and other indicators as secondary evaluation indicators. The thrombolytic risk prediction model for myocardial infarction patients specifically includes: A first random forest model responsible for risk prediction for the composite complication risk type in the pre-thrombolytic stage; A first XGBoost model responsible for risk prediction for the composite complication risk type in the post-thrombolysis stage; A logistic regression model responsible for risk prediction for the bleeding risk type during the pre-thrombolytic stage; A K-nearest neighbor model responsible for risk prediction for the bleeding risk type in the post-thrombolysis stage; A second random forest model responsible for risk prediction for the death risk type in the pre-thrombolysis stage; A second XGBoost model is responsible for risk prediction for the death risk type in the post-thrombolysis stage.

6. The method according to claim 1, characterized in that In the second model design scheme, the thrombolytic risk of the myocardial infarction patient involves two risk types, and the two risk types specifically involve a bleeding risk type and a death risk type; The method further comprises: Through the characteristic variables of the labeled sample myocardial infarction patients, training is carried out based on different types of pre-set initial models for the bleeding risk type and the death risk type, and the optimal model parameters are trained through grid search and 5-fold cross validation during the training process. The models responsible for the risk prediction work corresponding to the bleeding risk type and the death risk type are selected through preset evaluation indicators to form a fusion of two models required for the thrombolytic risk prediction model for myocardial infarction patients, wherein the different types of initial models include an initial logistic regression model, an initial random forest model, an initial XGBoost model, an initial gradient boosting decision tree model and an initial support vector machine model. The thrombolytic risk prediction model for myocardial infarction patients specifically outputs the model predicted risk situation through CRUSADE score, ACUITY score, GRACE score or TIMI score.

7. The method according to claim 6, characterized in that The preset evaluation index uses the AUC index as the main evaluation index and other indicators as secondary evaluation indicators; When the model prediction risk situation is outputted through the Crusade score, the thrombolysis risk prediction model for myocardial infarction patients includes a first random forest model responsible for risk prediction for the bleeding risk type; When the model prediction risk situation is outputted through the Acuity score, the thrombolysis risk prediction model for myocardial infarction patients includes a second random forest model responsible for risk prediction for the bleeding risk type; When the model prediction risk situation is outputted through the Grace score, the thrombolysis risk prediction model for myocardial infarction patients includes a third random forest model responsible for risk prediction for the death type; When the model predicts the risk situation through the Crusade score, the thrombolysis risk prediction model for myocardial infarction patients includes a logistic regression model responsible for risk prediction for the death type.

8. A device for predicting the risk of thrombolysis in patients with myocardial infarction, characterized in that: The device comprises: an acquisition unit, used for acquiring a target characteristic variable of a myocardial infarction patient whose thrombolysis risk is to be predicted at a current thrombolysis stage, wherein the current thrombolysis stage is a pre-thrombolysis stage or a post-thrombolysis stage, the pre-thrombolysis stage involves 19 first characteristic variables, and the post-thrombolysis stage involves 16 second characteristic variables in addition to the 19 first characteristic variables; A prediction unit, used for inputting the target characteristic variable into a pre-configured thrombolysis risk prediction model for myocardial infarction patients, wherein the thrombolysis risk prediction model for myocardial infarction patients is a model trained by characteristic variables of labeled sample myocardial infarction patients, and the thrombolysis risk prediction model for myocardial infarction patients is used for predicting the thrombolysis risk of myocardial infarction patients based on the characteristic variables input into the model; The reminder unit is used to output corresponding risk reminders based on the thrombolysis risk prediction results of myocardial infarction patients output by the thrombolysis risk prediction model for myocardial infarction patients.

9. A processing device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the method according to any one of claims 1 to 7 when calling the computer program in the memory.

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

Images