A method, apparatus, medium, and product for predicting stroke recurrence risk stratification

Abstract

The application discloses a method, device, medium and product for predicting stroke recurrence risk stratification, relates to the technical field of risk prediction, and comprises the following steps: acquiring brain magnetic resonance imaging data and clinical data of a patient; processing the brain magnetic resonance imaging data of the patient to obtain a background brain age; calculating the difference between the background brain age and the actual age to obtain a brain age difference; constructing a risk pool based on the brain age difference and the clinical data; extracting a feature subset by using a random non-replacement sampling method, and performing feature screening and optimization to obtain an optimal feature combination; optimizing the hyperparameters of a risk prediction model; based on the optimized risk prediction model, using a Pareto curve analysis method, combining a target function and a constraint condition, and solving a risk stratification threshold; obtaining a patient risk stratification result based on the risk stratification threshold; and combining an anti-platelet therapy clinical practice guideline and clinical medication data to generate individualized treatment recommendations. The application can realize accurate stratification of stroke recurrence risk.

Description

Technical Field

[0001] This application relates to the field of risk prediction technology, and in particular to a method, device, medium and product for predicting the risk stratification of stroke recurrence. Background Technology

[0002] Ischemic stroke is characterized by high recurrence and disability rates. Timely and accurate identification of high-risk groups for stroke recurrence is crucial for developing individualized secondary prevention and antiplatelet therapy plans. Currently, mainstream clinical tools for assessing stroke recurrence risk include Stroke Prognosis Instrument-I (SPI-I), Stroke Prognosis Instrument-II (SPI-II), and the Essen Stroke Risk Score (ESRS). These traditional assessment systems rely solely on basic clinical variables to build linear predictive models, resulting in limited model discriminatory power and significant issues of misdiagnosis and underdiagnosis of high-risk patients. They fail to meet the current needs of precision medicine for personalized intervention.

[0003] Brain age gap (BAG) is an imaging biomarker that quantifies the degree to which physiological and pathological aging of the brain deviates from the actual age. Existing clinical studies have confirmed that BAG is highly correlated with the short-term and long-term risk of stroke recurrence. Data shows that for every year increase in BAG, the risk of stroke recurrence increases by 9% at 3 months and by 7% at 5 years, demonstrating its core potential for predicting stroke recurrence risk. However, there is currently a lack of dedicated predictive models that integrate BAG with clinical characteristics, and there is an urgent need for tools that can accurately stratify risk to match individualized treatment needs. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, medium, and product for predicting the risk stratification of stroke recurrence, which can construct a machine learning model based on brain age difference to achieve accurate stratification of stroke recurrence risk and provide technical support for optimizing secondary prevention strategies.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for predicting stroke recurrence risk stratification, the method comprising: Obtain the patient's brain magnetic resonance imaging data and clinical information; the brain magnetic resonance imaging data includes: T2-weighted fluid-attenuated inversion recovery (T2-FLAIR) sequence and diffusion-weighted imaging.

[0006] The patient's brain magnetic resonance imaging data were processed using a pre-trained brain age prediction model to obtain the contextual brain age (CBA).

[0007] The difference between the background brain age and the actual age is calculated to obtain the brain age difference.

[0008] Based on the brain age difference and the clinical data, a risk pool was constructed.

[0009] Based on the risk pool, a subset of features is extracted using random sampling without replacement to obtain several feature combinations.

[0010] The optimal feature combination is obtained by filtering and optimizing the aforementioned feature combinations.

[0011] Based on the optimal feature combination, the risk prediction model is optimized by using 10-fold cross-validation to obtain the optimized risk prediction model.

[0012] Based on the optimized risk prediction model, the Pareto curve analysis method is used, combined with the objective function and constraints, to solve for the risk stratification threshold.

[0013] Based on the risk stratification threshold, the patient risk stratification results are obtained; the risk stratification results include three categories of stroke recurrence risk: high, medium and low.

[0014] Based on the risk stratification results, combined with clinical practice guidelines for antiplatelet therapy and clinical drug use data, individualized treatment recommendations are generated.

[0015] Optionally, the risk pool includes 19 traditional clinical factors and 4 imaging-related factors.

[0016] The 19 traditional clinical factors include: patient age, history of diabetes, final diagnosis upon admission, first right systolic blood pressure upon admission, first right diastolic blood pressure upon admission, first left systolic blood pressure upon admission, first left diastolic blood pressure upon admission, hypertension classification, history of coronary artery disease, history of heart failure, history of stroke, history of transient ischemic attack, history of hypertension, history of heart disease, specific type of coronary artery disease, history of atrial fibrillation, history of other cardiovascular diseases, history of peripheral artery disease, and current smoking status.

[0017] The four imaging-related factors include: brain age difference, infarct volume, number of infarcts, and TOAST classification (Trial of Org 10172 in Acute Stroke Treatment) results.

[0018] Optionally, feature filtering and optimization are performed on the aforementioned feature combinations to obtain the optimal feature combination, specifically including: A logistic regression model is constructed, and its performance is evaluated using nested 5-fold cross-validation (including inner and outer loops) to obtain the evaluated model.

[0019] Based on the evaluated model, the area under the receiver operating characteristic (AUC) for each feature combination is calculated.

[0020] The feature combination with the largest area under the subject's operating characteristic curve is taken as the optimal feature combination.

[0021] Optionally, feature selection and optimization can also be performed using Least Absolute Shrinkage and Selection Operator (LASSO), Recursive Feature Elimination (RFE), or feature importance ranking based on a tree model to obtain the optimal feature combination.

[0022] It is also possible to determine the risk stratification threshold based on the optimized prediction model using the Youden index method, the maximum selection method, or the expert consensus method.

[0023] Optionally, the risk prediction model includes: Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Artificial Neural Network (ANN), Support Vector Machine (SVM), Gradient Boosting Decision Tree (GBDT), Light Gradient Boosting Machine (LightGBM), and Categorical Boosting (CatBoost).

[0024] Optionally, the expression for the objective function is: ; ; in, Let this be the first objective function; The second objective function; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; This refers to the proportion of the low-risk group in the total population.

[0025] Optionally, the expression for the constraint condition is: ; ; ; in, Thresholds for stratifying high and low risks; The threshold for low-risk stratification; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; The proportion of the low-risk group in the total population In a second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting stroke recurrence risk stratification as described in any of the above-mentioned methods.

[0026] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting stroke recurrence risk stratification as described above.

[0027] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for predicting stroke recurrence risk stratification as described above.

[0028] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method, device, medium, and product for predicting stroke recurrence risk stratification. By simultaneously acquiring multimodal brain magnetic resonance imaging data including T2-weighted fluid attenuation inversion recovery sequences and diffusion-weighted imaging, along with clinical data, it can simultaneously gather all the original data sources required for imaging aging indicators, lesion imaging indicators, and traditional clinical prognostic indicators. This provides a comprehensive and complete data foundation for the subsequent construction of a complete risk pool, avoiding the loss of predictive information caused by a single data dimension. A pre-trained brain age prediction model is used to calculate background brain age, eliminating the need for subjective human interpretation of brain aging, significantly reducing human error and labor costs, and ensuring a unified standard for brain age calculation for different patients. The age calculation results can be compared horizontally; by subtracting the patient's actual age from the background brain age, a quantitative index of brain age difference is obtained, which transforms the degree to which brain aging deviates from the normal physiological age into a numerical biomarker that can be input into the model, realizing the digitization of the pathological characteristics of brain aging and filling the gap in traditional assessments that lack quantitative factors for aging imaging; a risk pool is constructed by integrating brain age difference and clinical data, simultaneously covering variables of the traditional stroke risk scoring system with brain lesions and brain aging imaging features, enriching the dimensions of predictive features, and providing more risk discrimination information compared to traditional methods that rely solely on clinical data; random sampling without replacement is used to extract feature subsets of different sizes multiple times, which can traverse various feature combinations and fully explore the potential of features. By identifying potential correlations between features, high-value feature combinations are avoided from being overlooked with fixed feature combinations, thus improving the comprehensiveness of feature selection. Feature selection and optimization of several feature combinations yields the optimal combination, effectively mitigating feature selection bias caused by the randomness of a single experiment and ensuring that the selected feature combinations possess stable discriminative ability and good generalization performance. Based on the optimal feature combination, multiple machine learning algorithms are combined with 10-fold cross-validation to optimize hyperparameters, effectively alleviating model overfitting and fully leveraging the advantages of nonlinear fitting in machine learning. This significantly improves the accuracy of stroke recurrence probability prediction compared to traditional linear prediction models. Furthermore, Pareto curve analysis is employed, combined with a bi-objective function... The method solves for risk stratification thresholds using numbers and constraints, moving beyond simply pursuing the optimal model discriminant index. It takes into account the differences in recurrence rate gradients among groups and the proportion of population distribution, ensuring that the stratification thresholds fully align with real-world clinical needs. Based on the standardized thresholds obtained, it divides recurrence risk into high, medium, and low levels, outputting clear and directly interpretable grading results instead of single probability values, facilitating clinicians to quickly differentiate patients at different risk levels. Based on the risk stratification results, combined with clinical practice guidelines for antiplatelet therapy and clinical medication data, it automatically generates individualized treatment recommendations, bridging the gap between risk prediction and clinical intervention, enabling the risk stratification results to directly guide the development of secondary prevention plans.This application utilizes multimodal MRI imaging to construct biomarkers for quantifying brain age differences. Combined with standardized sampling screening, cross-validation optimization of machine learning models, and a Pareto stratification threshold solution process aligned with clinical constraints, it achieves high-precision, clinically relevant three-tiered stratification of stroke recurrence risk. Compared to traditional risk assessment tools, this significantly improves the accuracy of high-risk patient identification and reduces false positives and false negatives. It can directly output individualized antiplatelet therapy plans, providing complete, stable, and feasible automated technical support for precise secondary prevention of ischemic stroke. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is an application environment diagram of a method for predicting stroke recurrence risk stratification in one embodiment of this application; Figure 2 A flowchart illustrating a method for predicting stroke recurrence risk stratification according to an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] This application uses computer equipment to process brain magnetic resonance imaging data and clinical data to construct an accurate risk prediction model, realize dynamic stratification of stroke recurrence risk, and provide a scientific basis for individualized antiplatelet therapy decisions.

[0033] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] The method for predicting stroke recurrence risk stratification provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send patient brain MRI data and clinical data to server 104. The brain MRI data includes: T2-weighted liquid attenuation inversion recovery sequence and diffusion-weighted imaging. After receiving the patient's brain MRI data and clinical data, server 104 processes the patient's brain MRI data using a pre-trained brain age prediction model to obtain the background brain age; calculates the difference between the background brain age and the actual age to obtain the brain age difference; constructs a risk pool based on the brain age difference and the clinical data; and performs feature subset sampling using random sampling without replacement based on the risk pool. Several feature combinations are obtained; these combinations are then screened and optimized to obtain the optimal feature combination; based on this optimal feature combination, a 10-fold cross-validation method is used to optimize the hyperparameters of the risk prediction model, resulting in an optimized risk prediction model; based on this optimized model, Pareto curve analysis is used, combined with the objective function and constraints, to solve for the risk stratification threshold; based on this threshold, patient risk stratification results are obtained; these results include high, medium, and low stroke recurrence risks; based on these results, combined with clinical practice guidelines for antiplatelet therapy and clinical medication data, individualized treatment recommendations are generated. Server 104 can then feed back these individualized treatment recommendations to terminal 102. In addition, in some embodiments, the method for predicting stroke recurrence risk stratification can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly predict stroke recurrence risk stratification based on the patient's brain magnetic resonance imaging data and clinical data, or the server 104 can obtain the patient's brain magnetic resonance imaging data and clinical data from the data storage system and predict stroke recurrence risk stratification based on the patient's brain magnetic resonance imaging data and clinical data.

[0035] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0036] In one exemplary embodiment, such as Figure 2As shown, a method for predicting stroke recurrence risk stratification is provided. This method is executed by a computer device, specifically a terminal or server, or both. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included: S1: Acquire the patient's brain magnetic resonance imaging data and clinical information; the brain magnetic resonance imaging data includes: T2-weighted liquid attenuation inversion recovery sequence and diffusion-weighted imaging.

[0037] S2: The patient's brain magnetic resonance imaging data is processed using a pre-trained brain age prediction model (such as a mask-based brain age estimation network, MBA Net) to obtain the background brain age.

[0038] S3: Calculate the difference between the background brain age and the actual age to obtain the brain age difference. The brain age difference, as a core indicator reflecting the degree to which the brain aging trajectory deviates from an individual's actual age, can provide a key imaging biomarker for subsequent risk stratification.

[0039] S4: Construct a risk pool based on the brain age difference and the clinical data.

[0040] S5: Based on the risk pool, a subset of features is extracted using random sampling without replacement to obtain several feature combinations.

[0041] S6: Perform feature filtering and optimization on the aforementioned feature combinations to obtain the optimal feature combination.

[0042] S7: Based on the optimal feature combination, the risk prediction model is optimized by using 10-fold cross-validation to obtain the optimized risk prediction model.

[0043] S8: Based on the optimized risk prediction model, the Pareto curve analysis method is used to solve the risk stratification threshold by combining the objective function and constraints.

[0044] S9: Based on the risk stratification threshold, the patient risk stratification results are obtained; the risk stratification results include three categories of stroke recurrence risk: high, medium, and low.

[0045] S10: Based on the risk stratification results, combined with clinical practice guidelines for antiplatelet therapy and clinical drug use data, generate individualized treatment recommendations.

[0046] By implementing steps S1 to S10 above, this application incorporates brain age difference into the stroke recurrence prediction framework for the first time. By combining multimodal features and machine learning algorithms, it fills the gap in existing technology. Through Pareto curves and constraints, it achieves significant differences in recurrence rates among risk groups at each level, which can accurately identify high-risk patients and significantly reduce the misjudgment rate compared to ESRS.

[0047] It should be noted that the method for calculating brain age difference can be replaced by other brain age prediction algorithms or whole-brain age estimation methods. However, this application has the following advantages: First, its technical irreplaceability. This application is based on the MBANet model, which solves the problem that previous whole-brain algorithms are easily affected by infarct foci and ischemic penumbra fluctuations in the acute phase; second, its clinical value. This model aligns with the guideline recommendation to "initiate antiplatelet therapy in the acute phase." By selecting acute-phase clinical characteristics to assist in screening target patients, it helps to achieve precise linkage between risk prediction and clinical medication.

[0048] As an optional implementation, the risk pool includes: 19 traditional clinical factors and 4 imaging-related factors; As shown in Table 1, the 19 traditional clinical factors include: patient age, history of diabetes, final diagnosis upon admission, first right systolic blood pressure upon admission, first right diastolic blood pressure upon admission, first left systolic blood pressure upon admission, first left diastolic blood pressure upon admission, hypertension classification, history of coronary artery disease, history of heart failure, history of stroke, history of transient ischemic attack, history of hypertension, history of heart disease, specific type of coronary artery disease, history of atrial fibrillation, history of other cardiovascular diseases, history of peripheral artery disease, and current smoking status; covering key variables in stroke prognosis and risk assessment systems such as SPI-I, SPI-II, and ESRS.

[0049] Table 1 19 Traditional Clinical Factors

[0050] The four imaging-related factors include: brain age difference, infarct volume, number of infarcts, and TOAST classification results, which are used for etiological classification.

[0051] As an optional implementation, in step S6, feature filtering and optimization are performed on the plurality of feature combinations to obtain the optimal feature combination, specifically including: S61: An initial prediction model is constructed using logistic regression, and the performance of the initial prediction model is evaluated using nested five-fold cross-validation to obtain the evaluated model.

[0052] S62: Based on the evaluated model, calculate the area under the receiver operating characteristic curve for each feature combination.

[0053] S63: The feature combination with the largest area under the subject's operating characteristic curve is taken as the optimal feature combination.

[0054] Specifically, a subset of features (5-10 features each time) is extracted from the risk pool using random sampling without replacement. To reduce computational overhead, logistic regression is used to construct the initial prediction model, and nested five-fold cross-validation is used to evaluate the model performance. The area under the receiver operating characteristic (AUC) curve is used as the core evaluation metric, and the experiment is independently repeated 500 times to select the feature combination with the optimal AUC, ensuring the robustness and generalization ability of the selected feature set.

[0055] This step belongs to the feature engineering stage. In this step, logistic regression is only used as an evaluator for a subset of features. The core purpose of calculating AUC through nested cross-validation is to select the most predictive feature combinations from a massive risk pool, addressing the question of 'what data to use' rather than building the final model.

[0056] It should be noted that in this embodiment, feature selection and optimization can also be performed using minimum absolute shrinkage and selection operator regression, recursive feature elimination, or feature importance ranking based on a tree model to obtain the optimal feature combination.

[0057] In addition to the area under the receiver operating characteristic curve, other indicators such as accuracy, recall, F1 score, and calibration curve can be introduced as supplementary evaluation metrics.

[0058] As an optional implementation, in step S7, the model is trained using 10-fold cross-validation. A risk prediction model is constructed using four types of algorithms: logistic regression, random forest, extreme gradient boosting, and artificial neural networks (such as multilayer perceptron). The model hyperparameters are optimized to improve generalization ability, with new stroke events occurring within 90 days as the endpoint indicator.

[0059] This step belongs to the model training and optimization stage. In this step, based on the specific feature subset selected in step S6, four algorithms, including logistic regression and random forest, are used to perform hyperparameter tuning and final model training, which solves the problem of 'how to establish the optimal prediction relationship'.

[0060] It should be noted that, in addition to these four types of algorithms, other machine learning algorithms such as support vector machines, gradient boosting decision trees, lightweight gradient boosting machines, and categorical feature gradient boosting can also be used as alternative solutions for building risk prediction models.

[0061] As an optional implementation, in step S8, the risk stratification threshold is determined based on Pareto curve analysis, and an optimization objective function is defined: ; ; in, Let this be the first objective function; The second objective function; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; This refers to the proportion of the low-risk group in the total population.

[0062] Constraints: C1: (High threshold ≥ low threshold); C2: (The recurrence rate increases exponentially with each iteration); C3: (Population ratio constraint); in, Thresholds for stratifying high and low risks; The threshold for low-risk stratification; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; This refers to the proportion of the low-risk group in the total population.

[0063] By maximizing and And meet the constraints, determine the thresholds for high, medium and low risk stratification, and ensure that the recurrence rate of each group is significantly different and the population distribution meets clinical needs.

[0064] It should be noted that, based on the optimized prediction model, the risk stratification threshold can also be determined using the Youden index method, the maximum selection method, or the expert consensus method.

[0065] As an optional implementation method, the method for predicting stroke recurrence risk stratification further includes: comparing with traditional models (SPI-I, SPI-II, ESRS), verifying the discrimination and calibration through C-statistics and Brier score, and calculating the net reclassification improvement index (NRI) and integrated discrimination improvement index (IDI) to quantify the predicted value-added benefits.

[0066] Finally, based on the risk stratification results, combined with clinical practice guidelines for antiplatelet therapy and clinical drug use data, individualized treatment recommendations are provided to clinicians.

[0067] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 3 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores patient brain MRI images and clinical data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for predicting stroke recurrence risk stratification.

[0068] Figure 3 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0069] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0070] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0071] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.

[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0073] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0074] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for stratifying the risk of stroke recurrence, characterized in that, The method includes: Obtain the patient's brain magnetic resonance imaging data and clinical information; the brain magnetic resonance imaging data includes: T2-weighted fluid attenuation inversion recovery sequence and diffusion-weighted imaging; The patient's brain magnetic resonance imaging data were processed using a pre-trained brain age prediction model to obtain the background brain age. The difference between the background brain age and the actual age is calculated to obtain the brain age difference; Based on the brain age difference and the clinical data, a risk pool is constructed; Based on the risk pool, a subset of features is extracted using random sampling without replacement to obtain several feature combinations. The optimal feature combination is obtained by performing feature filtering and optimization on the aforementioned feature combinations. Based on the optimal feature combination, the risk prediction model is optimized by using 10-fold cross-validation to obtain the optimized risk prediction model. Based on the optimized risk prediction model, the Pareto curve analysis method is used to solve the risk stratification threshold by combining the objective function and constraints. Based on the risk stratification threshold, the patient risk stratification results are obtained; the risk stratification results include three categories of stroke recurrence risk: high, medium, and low. Based on the risk stratification results, combined with clinical practice guidelines for antiplatelet therapy and clinical drug use data, individualized treatment recommendations are generated.

2. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, The risk pool includes 19 traditional clinical factors and 4 imaging-related factors; The 19 traditional clinical factors include: patient age, history of diabetes, final diagnosis upon admission, first right systolic blood pressure upon admission, first right diastolic blood pressure upon admission, first left systolic blood pressure upon admission, first left diastolic blood pressure upon admission, hypertension classification, history of coronary artery disease, history of heart failure, history of stroke, history of transient ischemic attack, history of hypertension, history of heart disease, specific type of coronary artery disease, history of atrial fibrillation, history of other cardiovascular diseases, history of peripheral artery disease, and current smoking status. The four imaging-related factors include: brain age difference, infarct volume, number of infarcts, and TOAST classification results.

3. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, Feature filtering and optimization are performed on the aforementioned feature combinations to obtain the optimal feature combination, specifically including: An initial prediction model was constructed using logistic regression, and the performance of the initial prediction model was evaluated using nested five-fold cross-validation to obtain the evaluated model. Based on the evaluated model, the area under the receiver operating characteristic curve for each feature combination is calculated; The feature combination with the largest area under the subject's operating characteristic curve is taken as the optimal feature combination.

4. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, It can also use minimum absolute shrinkage and selection operator regression, recursive feature elimination, or tree-based feature importance ranking to perform feature screening and optimization, and obtain the optimal feature combination; It is also possible to determine the risk stratification threshold based on the optimized prediction model using the Youden index method, the maximum selection method, or the expert consensus method.

5. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, The risk prediction models include: logistic regression, random forest, extreme gradient boosting, artificial neural network, support vector machine, gradient boosting decision tree, lightweight gradient boosting machine, and categorical feature gradient boosting.

6. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, The expression for the objective function is: ； ； in, Let this be the first objective function; The second objective function; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; This refers to the proportion of the low-risk group in the total population.

7. The method for predicting stroke recurrence risk stratification according to claim 1, characterized in that, The expression for the constraint condition is: ； ； ； in, Thresholds for stratifying high and low risks; The threshold for low-risk stratification; The recurrence rate for the high-risk group; The recurrence rate for the medium-risk group; The recurrence rate for the low-risk group; The proportion of the high-risk group in the total population; This refers to the proportion of the low-risk group in the total population.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the method for predicting stroke recurrence risk stratification as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for predicting stroke recurrence risk stratification as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for predicting stroke recurrence risk stratification as described in any one of claims 1-7.

Images