Cardiovascular event risk prediction method, apparatus, device, and medium

Abstract

This disclosure relates to a method, apparatus, device, and medium for predicting cardiovascular event risks. The method includes: collecting clinical data from a target user; preprocessing the clinical data to obtain a standardized feature representation; then using a pre-trained risk prediction model to predict the risk of the feature representation, obtaining the probability of the target user experiencing a major adverse cardiovascular event; and finally generating a corresponding risk prediction result and a visual interpretation report based on the risk probability. Thus, by using the feature representation obtained from preprocessing clinical data, and performing risk prediction based on a pre-trained risk prediction model, the probability of the target user experiencing a major adverse cardiovascular event is obtained, and a risk prediction result and a visual interpretation report are generated, thereby achieving a balance between high-precision prediction and high-transparency interpretation, providing clinicians with powerful data-driven decision support.

Description

Technical Field

[0001] This disclosure relates to the field of medical information technology and artificial intelligence, and in particular to a method, device, equipment and medium for predicting cardiovascular event risks. Background Technology

[0002] Acute ST-segment elevation myocardial infarction (STEMI) is a critical cardiovascular event, and timely emergency percutaneous coronary intervention (PCI) is currently the most effective means of revascularization. However, even if the surgery is successful, some patients still experience major adverse cardiovascular events (MACE) during hospitalization, such as cardiac death, re-infarction, and heart failure, which seriously affect prognosis and consume a large amount of medical resources.

[0003] Currently, widely used clinical risk assessment tools, such as the GRACE score and TIMI score, are mostly based on traditional logistic regression or Cox proportional hazards models. While these models provide some risk stratification capabilities, their predictive performance is limited (AUC is typically between 0.7 and 0.8), and they struggle to capture complex, non-linear interactions of clinical characteristics. Furthermore, traditional models are often "black boxes" or lack interpretability, making it difficult for physicians to understand their decision-making rationale, thus limiting their application in personalized precision medicine.

[0004] In recent years, machine learning techniques, especially ensemble learning algorithms (such as XGBoost), have demonstrated outstanding performance in medical prediction tasks. However, their "black box" nature raises questions about credibility and interpretability, which has become a major obstacle to their clinical application. Summary of the Invention

[0005] To address the aforementioned technical problems, this disclosure provides a method, apparatus, device, and medium for predicting cardiovascular event risks.

[0006] Firstly, this disclosure provides a method for predicting the risk of cardiovascular events, including: Collect clinical data from the target users; The clinical data is preprocessed to obtain standardized feature representations; The risk probability of the target user experiencing a major adverse cardiovascular event is obtained by performing risk prediction on the feature representation using a pre-trained risk prediction model. Based on the risk probability, a corresponding risk prediction result and a visual explanation report are generated.

[0007] Secondly, this disclosure provides a cardiovascular event risk prediction device, comprising: The data acquisition module is used to collect clinical data from the target users; The first processing module is used to preprocess the clinical data to obtain standardized feature representations; The second processing module is used to perform risk prediction on the feature representation through a pre-trained risk prediction model to obtain the risk probability of the target user experiencing major adverse cardiovascular events. The third processing module is used to generate corresponding risk prediction results and visual explanation reports based on the risk probability.

[0008] Thirdly, this disclosure provides a cardiovascular event risk prediction device, including: processor; Memory, used to store executable instructions; The processor is used to read executable instructions from memory and execute the executable instructions to implement the cardiovascular event risk prediction method of the first aspect.

[0009] Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the cardiovascular event risk prediction method of the first aspect.

[0010] The technical solution provided in this disclosure has the following advantages compared with the prior art: The cardiovascular event risk prediction method of this disclosure collects clinical data from a target user, preprocesses the clinical data to obtain standardized feature representations, then uses a pre-trained risk prediction model to predict the risk probability of the target user experiencing a major adverse cardiovascular event, and finally generates a corresponding risk prediction result and a visual interpretation report based on the risk probability. Thus, by using feature representations obtained from preprocessing clinical data, performing risk prediction based on a pre-trained risk prediction model, obtaining the risk probability of a target user experiencing a major adverse cardiovascular event, and generating risk prediction results and a visual interpretation report, a unified approach of high-precision prediction and high-transparency interpretation is achieved, providing clinicians with powerful data-driven decision support. Attached Figure Description

[0011] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0012] Figure 1 A flowchart illustrating a cardiovascular event risk prediction method provided in this embodiment of the disclosure; Figure 2 A flowchart illustrating another cardiovascular event risk prediction method provided in this embodiment of the disclosure; Figure 3 This is a schematic diagram of the structure of a cardiovascular event risk prediction device provided in an embodiment of the present disclosure; Figure 4 This is a schematic diagram of the structure of a cardiovascular event risk prediction device provided in an embodiment of this disclosure. Detailed Implementation

[0013] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0014] It should be understood that the various steps described in the method implementation of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method implementation may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0015] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] To address the aforementioned problems, this disclosure provides a method, apparatus, device, and medium for predicting cardiovascular event risk. The following is a detailed description... Figure 1-2 The cardiovascular event risk prediction method provided in the embodiments of this disclosure will be described in detail.

[0020] Figure 1 A flowchart illustrating a cardiovascular event risk prediction method provided in an embodiment of this disclosure is shown.

[0021] In this embodiment of the disclosure, the cardiovascular event risk prediction method can be executed by an electronic device. The electronic device may include, but is not limited to, devices such as computer equipment, cloud servers, or cloud server clusters.

[0022] like Figure 1 As shown, the cardiovascular event risk prediction method may include the following steps.

[0023] S110. Collect clinical data from the target users.

[0024] In this embodiment of the disclosure, the electronic device can collect clinical data from the target user.

[0025] Optionally, the target users can be patients who have undergone emergency percutaneous coronary intervention (PCI) for acute ST-segment elevation myocardial infarction (STEMI).

[0026] Specifically, after patients with acute ST-segment elevation myocardial infarction (STEMI) undergo emergency percutaneous coronary intervention (PCI), electronic devices can acquire the target user's clinical data, which may include demographic information, vital signs, laboratory test results, cardiac function indicators, and derived indicators reflecting metabolic status.

[0027] S120. The clinical data is preprocessed to obtain a standardized feature representation.

[0028] In this embodiment of the disclosure, the electronic device can preprocess the clinical data to obtain a standardized feature representation.

[0029] Specifically, after collecting clinical data from the target user, the electronic device can preprocess the clinical data, that is, transform the raw and heterogeneous clinical data into a clean, unified, and standardized feature representation that is suitable for machine learning models to digest.

[0030] S130. The risk of the feature representation is predicted by a pre-trained risk prediction model to obtain the risk probability of the target user experiencing major adverse cardiovascular events.

[0031] In this embodiment of the disclosure, the electronic device can use a pre-trained risk prediction model to predict the risk of the feature representation and obtain the risk probability of the target user experiencing a major adverse cardiovascular event.

[0032] Alternatively, the pre-trained risk prediction model can be a model that learns patterns from massive amounts of historical data.

[0033] Specifically, electronic devices can input feature representations into a pre-trained risk prediction model, which can make risk predictions based on the feature representations and output the probability of a target user experiencing a major adverse cardiovascular event (MACE).

[0034] S140. Based on the risk probability, generate the corresponding risk prediction results and a visual explanation report.

[0035] In this embodiment of the disclosure, the electronic device can generate a corresponding risk prediction result and a visual explanation report based on the risk probability.

[0036] This approach enables the collection of clinical data from target users, followed by preprocessing to obtain standardized feature representations. A pre-trained risk prediction model then predicts the risk probability of the target user experiencing a major adverse cardiovascular event. Finally, based on these risk probabilities, a corresponding risk prediction result and a visual interpretation report are generated. Thus, by using feature representations obtained through preprocessing clinical data and performing risk prediction based on a pre-trained risk prediction model to obtain the risk probability of the target user experiencing a major adverse cardiovascular event, and generating risk prediction results and a visual interpretation report, a unified approach of high-precision prediction and high-transparency interpretation is achieved, providing clinicians with powerful data-driven decision support.

[0037] Optionally, clinical data may include demographic information, vital signs, laboratory test results, cardiac function indicators, and derived indicators reflecting metabolic status. Demographic information and vital signs constitute the patient's basic and immediate state; laboratory test results reflect pathophysiological changes at the cellular and molecular levels; cardiac function indicators directly assess the heart's pumping capacity and structural state; and derived indicators reflecting metabolic status are not raw measurements, but rather composite indicators calculated using specific formulas that comprehensively reflect deeper metabolic disorders such as insulin resistance. This data structure ensures that the model can comprehensively "perceive" the patient's health status from different levels and perspectives.

[0038] The demographic information may include age, sex, smoking history, etc.; vital signs may include heart rate, systolic blood pressure, diastolic blood pressure, etc.; laboratory test indicators may include brain natriuretic peptide (BNP), creatine kinase isoenzyme MB (CK-MB), myoglobin (MYO), high-sensitivity troponin I (hs-TnI), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), glycated hemoglobin (HbA1c), creatinine (Cr), fasting blood glucose (FBG), etc.; cardiac function indicators may include left ventricular ejection fraction (LVEF), Killip classification (cardiac function classification, Killip classification), etc.; and derived indicators reflecting metabolic status may include the triglyceride-glucose index (used to assess insulin resistance, TyG index), etc.

[0039] For example, in the cardiology department of a hospital, a 68-year-old male STEMI patient successfully underwent emergency PCI and returned to his ward. The resident physician clicked the "MACE Risk Assessment" button in the electronic medical record system. The system automatically retrieved the patient's latest blood test results (troponin I: 5.2 ng / mL, BNP: 480 pg / mL) from the laboratory information system, obtained vital signs (heart rate: 102 bpm, blood pressure: 100 / 65 mmHg) from the electrocardiogram monitoring system, and extracted information from the structured fields of the medical record (age: 68 years, Killip classification: II, smoking history: yes). The system then performed preprocessing and model calculations internally, and a moment later, an interactive report interface popped up on the physician's workstation, prominently displaying at the top: "In-hospital MACE predicted risk probability: 72.3%".

[0040] Therefore, by integrating the TyG index with traditional markers of cardiac function and myocardial injury, the model no longer focuses solely on myocardial injury (such as troponin) and hemodynamics (such as blood pressure), but can gain insights into deeper metabolic disorder backgrounds. This may help identify "hidden high-risk" patients whose traditional indicators appear normal but whose metabolic abnormalities lead to high risk, thus enabling earlier and more comprehensive risk warnings.

[0041] Optionally, S120 may specifically include: performing data quality processing, feature engineering processing, and data standardization processing on the clinical data to obtain a standardized feature representation.

[0042] In this embodiment of the disclosure, the electronic device can preprocess the clinical data, including data quality processing, feature engineering processing, and data standardization processing, thereby obtaining a standardized feature representation.

[0043] Specifically, electronic devices can first perform data quality processing on clinical data, such as intelligent imputation of missing values ​​(e.g., imputing with the mean for continuous variables following a normal distribution, with the median for skewed distributions, and with the mode for categorical variables), and detection and correction of outliers (e.g., using the Tukey method based on interquartile ranges or truncating or treating them as missing values ​​based on reasonable clinical knowledge). Next, feature engineering is performed on the clinical data, such as feature creation (e.g., calculating TyG index, body mass index, pulse pressure, etc.), feature transformation (e.g., taking the logarithm of skewed indicators to make them closer to a normal distribution), and feature extension (e.g., combining "diabetes status" with "blood glucose level" to form a composite feature reflecting blood glucose control). Finally, data standardization is performed on the clinical data, such as unifying features with different dimensions and value ranges to a standard scale (usually a standard normal distribution with a mean of 0 and a variance of 1). For continuous variables, Z-score standardization is commonly used; for ordered categorical variables, ordered encoding can be used; and for unordered categorical variables, one-hot encoding is used. This results in a standardized feature representation.

[0044] For example, in the data processing pipeline, the system discovered a missing record for the patient's low-density lipoprotein cholesterol (LDL-C). After assessment, determining that LDL-C is a continuous variable, the system used the median (2.6 mmol / L) of this indicator learned from the training data to impute it (data quality control). Subsequently, the system calculated non-high-density lipoprotein cholesterol (non-HDL-C = TC - HDL-C) as a new feature based on the existing data (feature engineering). Finally, all continuous variables, such as age, BNP, and troponin I, were subtracted from their corresponding means in the training dataset, and then divided by their standard deviation to complete Z-score standardization (data standardization), ultimately generating a fixed-length feature vector composed entirely of numbers, i.e., the feature representation.

[0045] Thus, data quality processing significantly reduces the interference of noise and erroneous data on the model; feature engineering creates more discriminative information through human prior knowledge; and data standardization eliminates feature weight bias caused by differences in units, ensuring that the model can fairly learn the contribution of each feature. The combination of these three processes enables the machine learning model to converge stably, achieve optimal performance, and provide highly reliable and repeatable prediction results.

[0046] Optionally, the cardiovascular event risk prediction method may further include: preprocessing historical clinical data to obtain a training set and a test set; and training and evaluating the model based on the training set and the test set to obtain the risk prediction model.

[0047] In this embodiment of the disclosure, the electronic device can preprocess historical clinical data to obtain a training set and a test set, and perform model training and performance evaluation based on the training set and the test set to obtain the risk prediction model.

[0048] Specifically, electronic devices can acquire historical clinical data, such as using the same data preprocessing procedures as the prediction time to prepare training data, ensuring consistency between the training and application environments, and dividing the data into a training set for learning patterns and a test set for objective evaluation. A model is built on the training set, its generalization ability is evaluated on the test set, and finally, the model with the best performance is selected as the deployment version.

[0049] For example, retrospectively, complete inpatient data and clearly defined in-hospital MACE outcome labels were collected from 5,000 STEMI patients undergoing emergency PCI who met the criteria. These 5,000 data points were randomly shuffled and divided into training, validation, and test sets at a ratio of 70%, 15%, and 15%, respectively. Using the training set data, multiple candidate models were trained based on the XGBoost algorithm framework, with hyperparameters adjusted using the validation set. Finally, all candidate models were evaluated using a completely unseen test set, and the model with the highest AUC-ROC and calibration curve closest to the diagonal on the test set was selected as the final deployable risk prediction model.

[0050] This establishes the scientific rigor and auditability of the model building process, greatly enhancing its credibility in clinical settings.

[0051] Optionally, the cardiovascular event risk prediction method may further include: during the training of the risk prediction model, using the minimum absolute contraction and selection operator regression method to select a subset of key prediction features from the historical clinical data for model training.

[0052] In this embodiment of the disclosure, during the training of the risk prediction model, the initial clinically collected indicators may number in the dozens or even hundreds, which inevitably include redundant, irrelevant, or weakly correlated features. The electronic device can employ a minimum absolute shrinkage and selection operator regression method (such as LASSO regression) to select a subset of key predictive features from the historical clinical data for model training. LASSO regression is a method for feature selection while building a generalized linear model. It adds a penalty term for the absolute value of the model coefficients to the loss function, automatically compressing the coefficients of many unimportant features to zero during the process of minimizing the prediction error, thereby achieving automatic feature selection.

[0053] For example, before training the XGBoost model, LASSO logistic regression analysis is performed on all 60 initial features of the training set. Through 10-fold cross-validation, the regularization strength coefficient λ that achieves the optimal balance between model bias and variance is determined. At this λ value, the LASSO model outputs coefficients, where the coefficients of 40 features are compressed to 0, and the coefficients of the remaining 14 features are non-zero, including BNP, MYO, LVEF, HDL-C, hs-TnI, TG, TyG index, LDL-C, HbA1c, CK-MB, smoking history, heart rate, Killip classification, and age. These 14 features are identified as the "key predictive feature subset." The subsequent XGBoost model will use only these 14 features for training and prediction.

[0054] This significantly reduces the number of features, lowering model complexity, enabling faster training and prediction, reducing the risk of overfitting, and improving the model's generalization ability on new data. Fewer features mean that interpretation methods like SHAP generate simpler charts, allowing for faster focus on core risk drivers and reducing the burden of understanding the model's decision-making logic. The fact that high-performance prediction can be achieved by collecting and inputting only a few dozen key metrics simplifies the front-end data collection process, reducing the time and economic costs of unnecessary checks, which is particularly beneficial for rapid deployment and application in resource-constrained scenarios such as emergency rooms.

[0055] Optionally, the risk prediction result includes a risk level determined by comparing the risk probability with a preset threshold, and the risk level includes at least one of low risk, medium risk, and high risk.

[0056] Specifically, based on our hospital's historical data and clinical practice consensus, the pre-set risk thresholds are set as follows: low risk (<30%), medium risk (30%-70%), and high risk (>70%). By comparing the risk probability with these thresholds, patients are categorized into different risk levels, such as low risk, medium risk, and high risk. Each level can be associated with different clinical treatment pathways and monitoring intensity recommendations. For example, for a patient with a predicted probability of 72.3%, the system automatically classifies them as high risk. In the generated report, this risk level will be marked with a prominent red label or icon. Simultaneously, based on the pre-set clinical pathway template, the system automatically lists the following recommendations in the report's recommendation section: 1. Transfer to the cardiac intensive care unit for intensive monitoring; 2. Repeat echocardiography within 24 hours to assess cardiac function; 3. Initiate intensive drug therapy and request consultation with a heart failure specialist, etc.

[0057] This enables the translation from complex probabilistic outputs to clear clinical instructions. It significantly improves the clinical operability and action guidance value of the predicted results, leading to more precise resource allocation and patient management.

[0058] Optionally, S140 may specifically include: performing interpretability analysis on the risk prediction results using the Shapley additivity interpretation framework, and generating the visualization interpretation report.

[0059] In this embodiment, the electronic device can employ the Shapley Additive Explanations (SHAP) framework to perform interpretability analysis on the risk prediction results and generate the visualized explanation report. SHAP, based on Shapley values ​​in game theory, assigns a numerical value to the contribution of each feature to a single prediction result. This value can be positive (increasing the predicted risk) or negative (decreasing the predicted risk). By calculating the SHAP value for each feature, the impact of each feature on the risk prediction result is clearly demonstrated.

[0060] For example, electronic devices can call the SHAP interpreter to analyze the XGBoost model's predictions. The results showed that high BNP (480 pg / mL) was the largest risk driver, contributing +0.15 to the SHAP value; low left ventricular ejection fraction (45%) contributed +0.12; and age (68 years) contributed +0.08. The absence of a history of diabetes was a protective factor, contributing -0.05. The system generated two types of visualizations: one for the patient, visually displaying the push and pull factors using arrow length and color; and another for the physician, summarizing the most important features globally.

[0061] Therefore, by providing visual and clinically intuitive explanations, the black-box decision-making process can be made transparent, enabling the understanding and verification of the model's reasoning logic.

[0062] Figure 2 A flowchart illustrating another cardiovascular event risk prediction method provided in an embodiment of this disclosure is shown.

[0063] like Figure 2As shown, electronic devices can collect clinical data from target users through a data acquisition module, such as demographic information, vital signs, laboratory test indicators, cardiac function indicators, and derived indicators reflecting metabolic status, obtained through electronic medical records and bedside devices. Next, the data processing module performs data quality processing on the clinical data, such as intelligent imputation of missing values ​​(e.g., imputing the mean for continuously distributed variables, the median for skewed distributions, and the mode for categorical variables), and detection and correction of outliers (e.g., using the Tukey method based on interquartile ranges or truncating or treating them as missing values ​​based on reasonable clinical knowledge). Then, feature engineering is performed on the clinical data, such as feature creation (e.g., calculating TyG index, body mass index, pulse pressure, etc.), feature transformation (e.g., taking the logarithm of skewed indicators to make them closer to a normal distribution), and feature extension (e.g., interacting "diabetes status" with "blood glucose level" to form a composite feature reflecting blood glucose control). Finally, the clinical data undergoes data standardization, such as unifying features with different dimensions and value ranges onto a standard scale (usually a standard normal distribution with a mean of 0 and a variance of 1). For continuous variables, Z-score standardization is commonly used; for ordered categorical variables, ordered coding can be used; and for unordered categorical variables, one-hot coding is employed. This results in a standardized feature representation.

[0064] Furthermore, the electronic device can use LASSO regression through the feature selection module to select a subset of key predictive features from the historical clinical data for model training, such as selecting 14 key predictive variables: BNP, MYO, LVEF, HDL-C, hs-TnI, TG, TyG index, LDL-C, HbA1c, CK-MB, smoking history, heart rate, Killip classification, and age.

[0065] Furthermore, electronic devices can use the model building and computation module to train multiple candidate models using training set data and the XGBoost algorithm as the basic framework, adjusting hyperparameters through a validation set. Finally, all candidate models are evaluated using a completely unseen test set, and the model with the highest AUC-ROC and calibration curve closest to the diagonal on the test set is selected as the final deployable risk prediction model.

[0066] Furthermore, the electronic device can use the model interpretation module and the Shapley Additive Interpretation Framework (SHAP) to perform interpretability analysis on the risk prediction results, generating the visualized interpretation report. It outputs the MACE risk probability and its corresponding risk level (e.g., low, medium, and high risk). A numerical value is assigned to the contribution of each feature to a single prediction result. This value can be positive (increasing the predicted risk) or negative (decreasing the predicted risk). By calculating the SHAP value for each feature, the impact of each feature on the risk prediction results is clearly demonstrated.

[0067] Figure 3 A schematic diagram of the structure of a cardiovascular event risk prediction device provided in an embodiment of this disclosure is shown.

[0068] like Figure 3 As shown, the cardiovascular event risk prediction device 300 may include a data acquisition module 310, a first processing module 320, a second processing module 330 and a third processing module 340.

[0069] The data acquisition module 310 can be used to collect clinical data from target users.

[0070] The first processing module 320 can be used to preprocess the clinical data to obtain standardized feature representations.

[0071] The second processing module 330 can be used to perform risk prediction on the feature representation through a pre-trained risk prediction model to obtain the risk probability of the target user experiencing major adverse cardiovascular events.

[0072] The third processing module 340 can be used to generate corresponding risk prediction results and visual explanation reports based on the risk probability.

[0073] Therefore, in this embodiment, clinical data of the target user can be collected, preprocessed to obtain standardized feature representations, and then a pre-trained risk prediction model can be used to predict the risk probability of the target user experiencing a major adverse cardiovascular event. Finally, based on the risk probability, a corresponding risk prediction result and a visual interpretation report are generated. Thus, by using feature representations obtained from preprocessing clinical data, and performing risk prediction based on a pre-trained risk prediction model, the risk probability of the target user experiencing a major adverse cardiovascular event is obtained, and a risk prediction result and a visual interpretation report are generated, thereby achieving a balance between high-precision prediction and high-transparency interpretation, providing clinicians with powerful data-driven decision support.

[0074] In some embodiments of this disclosure, the clinical data includes demographic information, vital signs, laboratory test results, cardiac function indicators, and derived indicators reflecting metabolic status.

[0075] In some embodiments of this disclosure, the first processing module 320 may be specifically used to perform data quality processing, feature engineering processing, and data standardization processing on the clinical data to obtain a standardized feature representation.

[0076] In some embodiments of this disclosure, the cardiovascular event risk prediction device 300 further includes: The fourth processing module can be used to preprocess historical clinical data to obtain training and test sets; The fifth processing module can be used to train the model and evaluate its performance based on the training set and the test set to obtain the risk prediction model.

[0077] In some embodiments of this disclosure, the cardiovascular event risk prediction device 300 further includes: The sixth processing module can be used to select a subset of key predictive features from the historical clinical data for model training by employing the minimum absolute shrinkage and selection operator regression method during the training process of the risk prediction model.

[0078] In some embodiments of this disclosure, the risk prediction result includes a risk level determined by comparing the risk probability with a preset threshold, and the risk level includes at least one of low risk, medium risk, and high risk.

[0079] In some embodiments of this disclosure, the third processing module can also be used to perform interpretability analysis on the risk prediction results using the Shapley additivity interpretation framework, and generate the visual interpretation report.

[0080] It should be noted that, Figure 3 The cardiovascular event risk prediction device 300 shown can perform... Figure 1-2 The various steps in the method embodiment shown are implemented. Figure 1-2 The processes and effects in the method embodiments shown are not described in detail here.

[0081] Figure 4 A schematic diagram of the structure of a cardiovascular event risk prediction device provided in an embodiment of this disclosure is shown.

[0082] In some embodiments of this disclosure, Figure 4 The cardiovascular event risk prediction device shown can be an electronic device. Specifically, the electronic device can include, but is not limited to, devices such as computer equipment, cloud servers, or cloud server clusters.

[0083] like Figure 4 As shown, the cardiovascular event risk prediction device may include a processor 401 and a memory 402 storing computer program instructions.

[0084] Specifically, the processor 401 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0085] Memory 402 may include a large-capacity storage for information or instructions. For example, and not limitingly, memory 402 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 402 may include removable or non-removable (or fixed) media. Where appropriate, memory 402 may be internal or external to the integrated gateway device. In a particular embodiment, memory 402 is a non-volatile solid-state memory. In a particular embodiment, memory 402 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable programmable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0086] The processor 401 reads and executes computer program instructions stored in the memory 402 to perform the steps of the cardiovascular event risk prediction method provided in this embodiment of the disclosure.

[0087] In one example, the cardiovascular event risk prediction device may also include a transceiver 403 and a bus 404. Wherein, as... Figure 4 As shown, the processor 401, memory 402 and transceiver 403 are connected via bus 404 and communicate with each other.

[0088] Bus 404 includes hardware, software, or both. For example, and not limitingly, a bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 404 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0089] This disclosure also provides a computer-readable storage medium that can store a computer program that, when executed by a processor, enables the processor to implement the cardiovascular event risk prediction method provided in this disclosure.

[0090] The aforementioned storage medium may, for example, include a memory 402 containing computer program instructions, which can be executed by the processor 401 of the cardiovascular event risk prediction device to complete the cardiovascular event risk prediction method provided in this embodiment. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), compact disc ROM (CD-ROM), magnetic tape, floppy disk, and optical data storage device.

[0091] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0092] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the risk of cardiovascular events, characterized in that, include: Collect clinical data from the target users; The clinical data is preprocessed to obtain standardized feature representations; The risk probability of the target user experiencing a major adverse cardiovascular event is obtained by performing risk prediction on the feature representation using a pre-trained risk prediction model. Based on the risk probability, a corresponding risk prediction result and a visual explanation report are generated.

2. The method according to claim 1, characterized in that, The clinical data includes demographic information, vital signs, laboratory test results, cardiac function indicators, and derived indicators reflecting metabolic status.

3. The method according to claim 1, characterized in that, The preprocessing of the clinical data to obtain standardized feature representations includes: The clinical data is subjected to data quality processing, feature engineering, and data standardization to obtain standardized feature representations.

4. The method according to claim 1, characterized in that, Before performing risk prediction on the feature representation using a pre-trained risk prediction model to obtain the risk probability of the target user experiencing a major adverse cardiovascular event, the method further includes: Historical clinical data were preprocessed to obtain training and testing sets; The risk prediction model is obtained by training the model and evaluating its performance based on the training set and the test set.

5. The method according to claim 4, characterized in that, The method further includes: During the training of the risk prediction model, the minimum absolute contraction and selection operator regression method is used to select a subset of key prediction features from the historical clinical data for model training.

6. The method according to claim 1, characterized in that, The risk prediction result includes a risk level determined by comparing the risk probability with a preset threshold, and the risk level includes at least one of low risk, medium risk and high risk.

7. The method according to claim 1, characterized in that, Generate a visual explanation report, including: The risk prediction results are analyzed for interpretability using the Shapley additivity interpretation framework, and a visual interpretation report is generated.

8. A cardiovascular event risk prediction device, characterized in that, include: The data acquisition module is used to collect clinical data from the target users; The first processing module is used to preprocess the clinical data to obtain standardized feature representations; The second processing module is used to perform risk prediction on the feature representation through a pre-trained risk prediction model to obtain the risk probability of the target user experiencing major adverse cardiovascular events. The third processing module is used to generate corresponding risk prediction results and visual explanation reports based on the risk probability.

9. A cardiovascular event risk prediction device, characterized in that, include: processor; Memory, used to store executable instructions; The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the cardiovascular event risk prediction method according to any one of claims 1-7.

10. A non-volatile computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, causes the processor to implement the cardiovascular event risk prediction method according to any one of claims 1-7.

Images