A method and system for risk stratification of chronic coronary syndrome based on coronary microcirculatory resistance index

Abstract

This invention discloses a risk stratification method for chronic coronary syndrome based on the coronary microcirculation resistance index, belonging to the field of medical data processing technology. The method includes: acquiring clinical characteristic data of patients to be assessed; inputting the clinical characteristic data into a preset risk stratification model; and obtaining the risk group of the patients to be assessed according to preset splitting conditions. The preset splitting conditions determine the splitting priority by analyzing the nonlinear correlation and discriminative power of multidimensional characteristic data, and combine this with a survival tree algorithm to train and generate a multi-level classification model. This invention overcomes the limitations of single-indicator assessment, effectively captures the nonlinear interactions between clinical indicators, significantly improves the accuracy of risk stratification, accurately identifies very low-risk and high-risk subgroups, avoids overtreatment or undertreatment, and the model is highly interpretable and easy to clinically apply.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a method and system for risk stratification of chronic coronary syndrome based on the coronary microcirculation resistance index. Background Technology

[0002] Chronic coronary syndrome (CCS) is one of the main types of coronary heart disease, and its risk stratification is crucial for clinical treatment decisions. Coronary microcirculatory dysfunction is an important pathological mechanism of CCS. Angiography-derived microcirculatory resistance index (AMR), as a novel microcirculatory functional indicator, has advantages such as being guidewire-free, easy to operate, and highly repeatable.

[0003] However, current technologies only use AMR as a single indicator for prognostic assessment, without fully considering its nonlinear interaction with other clinical indicators (such as NT-proBNP, age, and degree of coronary artery stenosis). This results in insufficient accuracy in risk stratification, making it impossible to identify low-risk subgroups with elevated AMR but good actual prognosis, and also making it difficult to accurately identify high-risk individuals who truly require intensive treatment.

[0004] Therefore, there is an urgent need for an intelligent risk stratification method that can integrate multi-dimensional clinical information and capture nonlinear interactions to improve the accuracy of individualized prognostic assessment for CCS patients. Summary of the Invention

[0005] To address the above problems, this invention provides a method and system for risk stratification of chronic coronary syndrome based on the coronary microcirculation resistance index.

[0006] A first aspect of the present invention provides a method for risk stratification of chronic coronary syndrome based on the coronary microcirculation resistance index, comprising the following steps: Obtain clinical characteristic data of the patient to be evaluated; The clinical characteristic data is input into a preset risk stratification model, and the risk group of the patient to be evaluated is obtained according to preset splitting conditions.

[0007] As a preferred method, the steps for determining the preset splitting conditions are as follows: Obtain the data type from the clinical feature data; Determine the data groups among the various feature data participating in the evaluation, and analyze the nonlinear correlations among the data groups; Determine the splitting threshold for each of the data groups; Based on the nonlinear correlation and the splitting threshold, the splitting priority of each data group in the risk stratification model is determined, and the splitting priority includes a first splitting condition, a second splitting condition, and an Nth splitting condition, where N is the number of data groups set.

[0008] As a preferred approach, before the risk stratification model is preset, the following steps are performed: Acquire the clinical characteristic data of the historical patient cohort and the historical follow-up survival data of the corresponding historical patients; The model is trained using the survival tree algorithm, with the first to Nth splitting conditions used as recursive splitting conditions. Iterative training determines the splitting threshold boundary for each splitting condition to maximize the survival difference between child nodes in the survival tree algorithm until the preset tree depth limit or node sample lower limit is met. Complete the training and output the preset risk stratification model.

[0009] As a preferred approach, each group in the data group is set as one or more data types from the clinical feature data.

[0010] As a preferred embodiment, the preset splitting conditions include: The first splitting condition is to conduct initial risk classification based on the microcirculation resistance index derived from angiography; The second splitting condition is the high-value branch that meets the first splitting condition, and the secondary prognostic subdivision is performed based on the level of N-terminal brain natriuretic peptide precursor. The third splitting condition involves combining age parameters with the degree of coronary artery stenosis for branches that do not meet the second splitting condition, resulting in a multi-dimensional collaborative determination.

[0011] As a preferred method, the clinical characteristic data is input into a preset risk stratification model, and the risk group of the patient to be assessed is obtained according to preset splitting conditions. This specifically includes the following steps: The first splitting condition is executed: if the microcirculation resistance index is ≥2.5, the patient is divided into the high microcirculation resistance group; otherwise, the patient is divided into the low microcirculation resistance group. The second splitting condition is applied: for the high microcirculation resistance group, if the N-terminal pro-brain natriuretic peptide level is <31.5 pg / mL, the patient is identified as a very low-risk subgroup. The third splitting condition is as follows: For patients with an N-terminal pro-brain natriuretic peptide level ≥31.5 pg / mL and belonging to the high microcirculation resistance group, if they meet the following criteria: age <78.5 years and coronary artery stenosis ≤70%, or coronary artery stenosis >70% and N-terminal pro-brain natriuretic peptide level <332 pg / mL, then the patients are identified as the intermediate-risk subgroup. Patients who do not meet the second and third splitting criteria are identified as a high-risk subgroup.

[0012] A second aspect of the present invention provides a risk stratification system for chronic coronary syndrome based on the coronary microcirculation resistance index, comprising: The data acquisition module is configured to acquire clinical characteristic data of the patient to be evaluated; The model grouping module is configured to input the clinical feature data into a preset risk stratification model, execute recursive judgment logic according to the preset splitting conditions, and output the risk group to which the patient to be evaluated belongs.

[0013] Compared with the prior art, the present invention has the following advantages: This invention integrates multidimensional clinical features such as AMR, NT-proBNP, age, and degree of coronary artery stenosis using a survival tree algorithm, effectively capturing the nonlinear interactions between variables and significantly improving the accuracy of prognostic assessment for patients with chronic coronary syndrome.

[0014] Meanwhile, this invention overcomes the shortcomings of traditional methods in that they cannot segment the population, effectively avoiding clinical overtreatment; at the same time, it accurately identifies high-risk groups with advanced age, severe stenosis, or significantly elevated NT-proBNP, providing clear quantitative evidence for enhanced intervention. Attached Figure Description

[0015] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0016] Figure 1 This is a structural block diagram of the system provided in the embodiments of the present invention.

[0017] Figure 2 This is a schematic diagram of the model structure provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This disclosure first provides a risk stratification system for chronic coronary syndrome based on the coronary microcirculation resistance index, such as... Figure 1-2 As shown, it includes a data acquisition module and a model grouping module. The data acquisition module is configured to input clinical characteristic data into a preset risk stratification model, perform recursive judgments based on preset splitting conditions, and output the risk group to which the patient to be assessed belongs.

[0020] Specifically, in this embodiment of the disclosure, the system is implemented via the provided method steps, which specifically include the following steps: Obtain clinical characteristic data of the patient to be evaluated; Clinical characteristic data are input into a pre-set risk stratification model, and the risk group of the patient to be evaluated is obtained according to the pre-set splitting conditions.

[0021] In this embodiment, the steps for determining the preset splitting conditions are as follows: Obtain the data type from the clinical feature data; Determine the data groups among the various feature data participating in the evaluation, and analyze the nonlinear correlations among the data groups; Determine the splitting threshold for each of the data groups; Based on the nonlinear correlation and the splitting threshold, the splitting priority of each data group in the risk stratification model is determined, and the splitting priority includes a first splitting condition, a second splitting condition, and an Nth splitting condition, where N is the number of data groups set.

[0022] In this embodiment, when acquiring the data types in the clinical feature data, the process includes performing structured parsing on the raw data to distinguish between discrete medical history identifier values, continuous physiological parameter values, and time-series follow-up record values; determining the data groups among the various feature data participating in the assessment; dividing the parsed features into a first feature data group, a second feature data group, and an Mth feature data group; constructing a multidimensional feature correlation matrix; and calculating the nonlinear correlation between each data group; the nonlinear correlation is quantified by a feature dependency metric based on information entropy decay or an interaction gain value based on mutual information.

[0023] The technical effectiveness of analyzing this nonlinear correlation lies in the fact that prognostic indicators in chronic pathological states often exhibit threshold saturation effects or high-order synergistic interaction characteristics. Linear regression models are susceptible to collinearity interference and cannot characterize nonlinear transitions. However, by quantifying the nonlinear correlation to remove redundant features, we can identify variable combinations with strong discriminative power for the occurrence of endpoint events, thus providing an input basis for subsequent survival tree splitting. Based on the strength of the nonlinear correlation and the outcome distribution characteristics of the historical cohort, an algorithm combining greedy search and grid optimization is used to traverse each data group and calculate the difference in the incidence rate of events between groups at each candidate split point to determine the data groups. The optimal splitting threshold boundary is determined; based on the strength of the nonlinear correlation and the discrimination effectiveness score corresponding to the optimal splitting threshold, the splitting priority sequence of each data group in the risk stratification model is determined; the splitting priority is sequentially mapped to the first splitting condition, the second splitting condition, up to the Nth splitting condition, where N is the number of data groups actually participating in the recursive splitting; the determination of the splitting priority is based on the log-rank statistic of the survival difference contributed by each feature at the initial node. The higher the statistic value, the stronger the significance of achieving population risk separation under the current splitting condition, thereby ensuring that the model can achieve the initial screening of at-risk populations at shallow nodes.

[0024] It should be noted that, in this embodiment, based on the strength of the nonlinear correlation and the outcome distribution characteristics of the historical queue, an algorithm combining greedy search and grid optimization is used to traverse each of the data groups and calculate the difference in the occurrence rate of events between groups at each candidate split point. The traversal algorithm used is not limited to the algorithm combining greedy search and grid optimization provided in this embodiment. Any algorithm that can be implemented to obtain the difference in the occurrence rate of events between groups can be used to implement this embodiment.

[0025] As a preferred approach, before the risk stratification model is preset, the following steps are performed: Acquire the clinical characteristic data of the historical patient cohort and the historical follow-up survival data of the corresponding historical patients; The model is trained using the survival tree algorithm, with the first to Nth splitting conditions used as recursive splitting conditions. Iterative training determines the splitting threshold boundary for each splitting condition to maximize the survival difference between child nodes in the survival tree algorithm until the preset tree depth limit or node sample lower limit is met. Complete the training and output the preset risk stratification model.

[0026] Specifically, in this embodiment, the clinical feature data of the historical patient cohort and the historical follow-up survival data of the corresponding historical patients are obtained. The follow-up survival data includes survival time values, endpoint event occurrence identifier values, and truncation status marker values. The dataset is then subjected to multiple imputation of missing values, outlier reduction, and variable standardization to construct a structured model training sample pool. Model training is performed using the survival tree algorithm as the underlying execution framework, and the first to the Nth splitting conditions are configured as a set of candidate recursive splitting conditions. During the training iteration process, the algorithm takes the set of candidate conditions as input and uses node splitting criteria based on log-rank test or condition inference tree to evaluate the candidate splitting performance of each feature group under the current parent node layer by layer. Through recursive iterative optimization, the precise splitting threshold boundary of each splitting condition is dynamically determined. The optimization process takes maximizing the separation degree of the survival curve between each child node as the core objective function and introduces a complexity penalty value to constrain the growth of the tree structure.

[0027] The effectiveness of this training mechanism lies in the fact that the survival tree algorithm does not require pre-assuming a global proportional hazards hypothesis, can adaptively identify nonlinear threshold jumps in clinical indicators, and recursively separates highly heterogeneous populations layer by layer through recursive segmentation, so that the populations within each terminal leaf node have highly consistent prognostic risk characteristics, significantly reducing model variance and improving clinical interpretability. Recursive segmentation is continuously executed until a preset convergence termination condition is met. This convergence termination condition includes at least reaching a preset maximum tree depth limit, the current node sample size reaching a preset minimum node sample size lower limit, or the statistical significance of survival differences between child nodes being lower than a preset threshold. After training convergence, a cost complexity pruning operation is performed on the generated initial decision tree to remove redundant branches that do not significantly contribute to generalization performance, extract the optimal subtree structure, complete model training, and output the preset risk stratification model containing node mapping rules, threshold parameter sets, and survival probability estimates for each level.

[0028] As a preferred approach, each group in the data group is set as one or more data types from the clinical feature data.

[0029] As a preferred embodiment, the preset splitting conditions include: The first splitting condition is to conduct initial risk classification based on the microcirculation resistance index derived from angiography; The second splitting condition is the high-value branch that meets the first splitting condition, and the secondary prognostic subdivision is performed based on the level of N-terminal brain natriuretic peptide precursor. The third splitting condition involves combining age parameters with the degree of coronary artery stenosis for branches that do not meet the second splitting condition, resulting in a multi-dimensional collaborative determination.

[0030] As a preferred method, the clinical characteristic data is input into a preset risk stratification model, and the risk group of the patient to be assessed is obtained according to preset splitting conditions. This specifically includes the following steps: The first splitting condition is executed: if the microcirculation resistance index is ≥2.5, the patient is divided into the high microcirculation resistance group; otherwise, the patient is divided into the low microcirculation resistance group. The second splitting condition is applied: for the high microcirculation resistance group, if the N-terminal pro-brain natriuretic peptide level is <31.5 pg / mL, the patient is identified as a very low-risk subgroup. The third splitting condition is as follows: For patients with an N-terminal pro-brain natriuretic peptide level ≥31.5 pg / mL and belonging to the high microcirculation resistance group, if they meet the following criteria: age <78.5 years and coronary artery stenosis ≤70%, or coronary artery stenosis >70% and N-terminal pro-brain natriuretic peptide level <332 pg / mL, then the patients are identified as the intermediate-risk subgroup. Patients who do not meet the second and third splitting criteria are identified as a high-risk subgroup.

[0031] This embodiment also details the clinical cohort construction and data processing procedures used to train and validate the risk stratification model.

[0032] The data collection subjects were patients who underwent coronary angiography due to suspected ischemic heart disease. To exclude confounding factors, strict exclusion criteria were set, including: acute coronary syndrome, severe liver, kidney or other organ disease, history of malignant tumors, hemodynamic instability, myocarditis, cardiomyopathy, Takotsubo syndrome, severe vascular overlap or artifacts on coronary angiography images, and patients lacking objective evidence of cardiac discomfort and myocardial ischemia.

[0033] After screening, 1415 patients were ultimately included in the analysis. Clinical characteristic data of the patients were extracted through the hospital's data recording system. This data included at least: demographic characteristics (sex, age); cardiovascular risk factors (smoking status, history of hypertension, history of type II diabetes); laboratory test indicators (glycated hemoglobin, low-density lipoprotein, estimated glomerular filtration rate eGFR, N-terminal pro-brain natriuretic peptide (NT-proBNP) level); coronary artery anatomy (degree of coronary artery stenosis, derived from surgical records); and medication use (statins, SGLT2 inhibitors). For the core functional indicator, the angiography-derived microcirculation resistance index (AMR), automated, guidewire-free calculation was performed using the AngioPlus software system (Version 2.0) from Shanghai Bodong Medical Technology Co., Ltd. AMR was measured in all three coronary arteries of each patient, and the arithmetic mean of the AMR of the three vessels was used as the patient's coronary microcirculation function assessment value. Based on clinical consensus, an AMR ≥ 2.5 was defined as high AMR, and an AMR < 2.5 was defined as low AMR. In this embodiment, there were 1026 patients with high AMR values ​​(accounting for 72.5%) and 389 patients with low AMR values ​​(accounting for 27.5%).

[0034] Major adverse cardiovascular events (MACE) were defined as the composite endpoint of all-cause mortality, rehospitalization due to heart failure or angina, non-fatal stroke, or ischemia-driven percutaneous coronary intervention (PCI). Follow-up was conducted by blinded researchers via telephone, with automatic reminder SMS messages sent one week prior to follow-up to improve response rates; the final cohort response rate was 85%. The median follow-up time was 843 days, during which 76 MACE events occurred, representing an overall incidence of 5.37%. Standardized procedures were followed during data preprocessing: variables with missing values ​​>50% were directly removed; the remaining variables were processed using multiple imputation (using the `mice` package in R), and the imputed dataset 1 was used for subsequent statistical analysis. Normally distributed data are expressed as mean ± standard deviation, skewed data as median [interquartiles], and categorical variables as frequency (proportion). Intergroup comparisons were performed automatically using the `compareGroups` package in R to avoid bias introduced by manual intervention.

[0035] This embodiment details the logic for determining the preset splitting conditions and the configuration of model training parameters. To avoid model overfitting and variable collinearity interference, prognostic features are screened in a subset of patients with high AMR values. Age, gender, hypertension, type II diabetes, eGFR, NT-proBNP level, and coronary artery stenosis are selected as candidate covariates input into the survival tree algorithm framework (implemented using the rpart package in R). The specific configuration of model training parameters is as follows: the minimum number of samples for node splitting (minsplit) is set to 20, the minimum number of events in a leaf node (minbucket) is set to 15, and the maximum tree depth (maxdepth) is limited to 3. These parameter settings aim to prevent the decision tree from becoming overly complex and ensure the model has clinical interpretability. During training, the algorithm automatically analyzes the correlation between various feature data involved in the evaluation and the discriminative power of the splitting threshold settings, and determines the splitting priority of each feature variable based on the discriminative power. Combining the priority order and multidimensional co-mapping relationship between different variables, the first splitting condition, the second splitting condition, and the third splitting condition are derived sequentially. After acquiring clinical characteristic data and corresponding long-term follow-up survival data of historical patient cohorts, iterative optimization is performed using the survival tree algorithm as the training framework. With the objective function of maximizing the survival difference between child nodes and the significance of the log-rank test, the threshold boundaries of each condition are determined through recursive segmentation until the preset tree depth limit or the lower limit of the node sample size is met, completing the training and outputting the preset risk stratification model.

[0036] This embodiment details the specific grouping path and threshold mapping after the patient data to be evaluated is input into a preset risk stratification model. The multi-level splitting condition execution steps after model training and solidification are as follows: The first split condition was implemented: initial risk classification was performed based on the microcirculation resistance index derived from angiography.

[0037] If AMR < 2.5, it is directly classified as a low microcirculation resistance group; If AMR ≥ 2.5, then proceed to the high AMR group branch for further judgment.

[0038] Second division conditions were implemented: for the high AMR group, a second prognostic subdivision was performed with an NT-proBNP level of 31.5 pg / mL as the cutoff.

[0039] If NT-proBNP < 31.5 pg / mL, these patients are identified as a very low-risk subgroup. This node includes 223 patients, with an actual event rate of 0%.

[0040] The third split criteria were applied to patients with NT-proBNP levels ≥31.5 pg / mL who belonged to the high AMR group, and a multidimensional collaborative assessment was performed with a coronary artery stenosis degree of 70% as the cutoff.

[0041] In branches with coronary artery stenosis ≤70%, a third division is performed with an age of 78.5 years as the boundary; Patients aged <78.5 years were classified into the first part of the intermediate-risk subgroup, comprising 574 patients with an event rate of 2.6%. In branches with coronary artery stenosis >70%, a third split was performed using an NT-proBNP level of 332 pg / mL as the cutoff. Patients with NT-proBNP <332 pg / mL were classified into the second part of the intermediate-risk subgroup, comprising 130 patients with an event rate of 10%. Patients aged ≥78.5 years had an event rate of 15%, and those with NT-proBNP ≥332 pg / mL had an event rate of 25.6%. Patients not meeting the criteria for the second or third split were uniformly identified as high-risk. Based on clinical operability and event rate gradients, the nodes generated from the survival tree were integrated with the low AMR population to form the final risk stratification output.

[0042] This example illustrates the specific logic for outputting 2-year MACE predicted risk values ​​based on risk groups and the model performance verification process.

[0043] The specific configuration of the risk output is as follows: if the risk group is the low microcirculatory resistance group, the output 2-year MACE predicted risk value is 7.5%; if it is the very low risk subgroup, the output risk value is 0%; if it is the intermediate risk subgroup, the output risk value is 4.0%; and if it is the high risk subgroup, the output risk value is 19.2%. To verify the model efficacy, Kaplan-Meier survival curves were plotted for each risk group and a log-rank test was performed (using the R software survminer package). The results showed that the survival differences between the stratified groups were statistically significant. Cox multivariate regression analysis was performed, and after adjusting for confounding factors, the hazard ratio (HR) of the high-risk subgroup relative to the reference group was 2.41, P=0.005. Sensitivity analysis was performed: in the high-AMR patient population, 100 resamplings of prognostic clinical characteristics were conducted to test the stability of the survival tree feature selection. The results showed that the core splitting variables remained highly consistent.

[0044] Three prognostic prediction models were constructed and compared: Model 1 uses AMR ≥ 2.5 as the boundary only; Model 2 uses AMR and 100 resampling features to construct a linear model; Model 3 is the multidimensional interaction model based on the survival tree of this invention. Evaluation was performed by calculating the consistency index (C-index) and net reclassification index (NRI). The results show that Model 3's NRI is 0.309 higher than the single-indicator model (P=0.001), and its C-index is significantly better than the comparative model. This confirms that this invention significantly improves the accuracy of risk stratification and the net clinical benefit by capturing the nonlinear interactions between multidimensional clinical indicators.

[0045] This embodiment provides a systematic deployment scheme and clinical workflow for risk stratification methods. The system architecture includes a data acquisition module, a model building module, a risk grouping module, a risk calculation module, and a clinical decision support module.

[0046] In terms of engineering implementation, this embodiment develops an online risk calculation platform based on the Shiny framework of the R language. The platform deploys program algorithms based on the method steps provided in this embodiment to realize the platform's functions and is deployed on a publicly accessible server.

[0047] In this embodiment, a clinical decision support module is also provided with the following workflow: based on the output risk group and the corresponding MACE predicted risk value, a preset clinical decision rule base is invoked to automatically generate structured treatment suggestion text. The suggestion content includes treatment intensity grading, drug intervention guidelines, and follow-up period settings. For example, for the very low-risk subgroup, routine follow-up and avoidance of overtreatment are suggested; for the low- and intermediate-risk subgroups, optimization of basic drug treatment and regular check-ups are suggested; for the high-risk subgroup, initiation of intensive treatment assessment and shortening of the follow-up period are suggested. The system pushes the risk group, predicted risk value, and structured treatment suggestion text to a web-based online computing platform or a mobile access interface for visualization, realizing rapid bedside assessment and individualized treatment guidance. This invention, by integrating multidimensional features and using the survival tree algorithm to capture nonlinear interactions, identifies for the first time a low-risk subgroup with elevated AMR but good prognosis, effectively avoiding overtreatment; at the same time, it accurately identifies high-risk populations, providing quantitative basis for intensive intervention. The model output of this embodiment is intuitive, has high clinical operability, and is suitable for individualized management of chronic coronary syndrome in medical institutions at all levels.

[0048] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent variations only. Individual components and functions are optional unless explicitly required, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed combinations. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0049] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods for each specific application to achieve the described functions, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the described devices, apparatuses, and units can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0050] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, function, and operation of implementations of apparatus, methods, and computer program products according to embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description; sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based device that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A risk stratification method for chronic coronary syndrome based on the coronary microcirculation resistance index, characterized in that, Includes the following steps: Obtain clinical characteristic data of the patient to be evaluated; The clinical characteristic data is input into a preset risk stratification model, and the risk group of the patient to be assessed is obtained according to preset splitting conditions; The preset splitting conditions include: The first splitting condition is to conduct initial risk classification based on the microcirculation resistance index derived from angiography; The second splitting condition is the high-value branch that meets the first splitting condition, and the secondary prognostic subdivision is performed based on the level of N-terminal brain natriuretic peptide precursor. The third splitting condition involves a multi-dimensional collaborative determination based on age parameters and the degree of coronary artery stenosis for branches that do not meet the second splitting condition. The first splitting condition is executed: if the microcirculation resistance index is ≥2.5, the patient is divided into the high microcirculation resistance group; otherwise, the patient is divided into the low microcirculation resistance group. The second splitting condition is applied: for the high microcirculation resistance group, if the N-terminal pro-brain natriuretic peptide level is <31.5 pg / mL, the patient is identified as a very low-risk subgroup. The third splitting condition is as follows: For patients with an N-terminal pro-brain natriuretic peptide level ≥31.5 pg / mL and belonging to the high microcirculation resistance group, if they meet the following criteria: age <78.5 years and coronary artery stenosis ≤70%, or coronary artery stenosis >70% and N-terminal pro-brain natriuretic peptide level <332 pg / mL, then the patients are identified as the intermediate-risk subgroup. Patients who do not meet the second and third splitting criteria are identified as a high-risk subgroup.

2. The method according to claim 1, characterized in that, The steps for determining the preset splitting conditions are as follows: Obtain the data type from the clinical feature data; Determine the data groups among the various feature data participating in the evaluation, and analyze the nonlinear correlations among the data groups; Determine the splitting threshold for each of the data groups; Based on the nonlinear correlation and the splitting threshold, the splitting priority of each data group in the risk stratification model is determined, and the splitting priority includes a first splitting condition, a second splitting condition, and an Nth splitting condition, where N is the number of data groups set.

3. The method according to claim 2, characterized in that, Before the risk stratification model is preset, the following steps are performed: Acquire the clinical characteristic data of the historical patient cohort and the historical follow-up survival data of the corresponding historical patients; The model is trained using the survival tree algorithm, with the first to Nth splitting conditions used as recursive splitting conditions. Iterative training determines the splitting threshold boundary for each splitting condition to maximize the survival difference between child nodes in the survival tree algorithm until the preset tree depth limit or node sample lower limit is met. Complete the training and output the preset risk stratification model.

4. The method according to claim 2, characterized in that, Each group in the data group is set as one or more data types from the clinical feature data.

5. A risk stratification system for chronic coronary syndrome based on the coronary microcirculation resistance index, suitable for implementing the method of claim 1, characterized in that, include: The data acquisition module is configured to acquire clinical characteristic data of the patient to be evaluated; The model grouping module is configured to input the clinical feature data into a preset risk stratification model, execute recursive judgment logic according to the preset splitting conditions, and output the risk group to which the patient to be evaluated belongs.

Images