Prognosis of heart failure combined with atrial fibrillation and risk stratification method and system for catheter ablation

Abstract

The present invention provides a method and system for risk stratification of prognosis and catheter ablation of heart failure combined with atrial fibrillation, the method comprising: step S1: collecting and preprocessing target feature data; the feature data comprises basic clinical phenotype feature data of the patient; the preprocessing comprises deletion, screening and variable analysis; step S2: constructing a Cox-PH model using the preprocessed feature data, and calculating the corresponding feature weights; step S3: clustering the target based on the weight value to obtain the result of risk stratification. The present invention can calculate the risk ratio of catheter ablation to endpoint events for subgroups in the clustering, thereby obtaining a group suitable for catheter ablation; when calculating dissimilarity, the weights of different features can also be considered, which greatly improves the accuracy of the calculation.

Description

Technical Field

[0001] The present invention relates to the field of medical intelligent decision-making, and in particular, to a method and system for risk stratification of prognosis and catheter ablation of heart failure combined with atrial fibrillation. Background Art

[0002] Medical intelligent decision-making is an important tool in the medical field, which can help medical professionals develop treatment plans for patients more quickly and comprehensively. Atrial fibrillation (AF) and heart failure (HF) often coexist due to common risk factors and pathophysiological mechanisms, and the coexistence of the two often indicates a poor prognosis. At present, the combination of atrial fibrillation (AF) and heart failure (HF) poses a major challenge to patients' risk assessment and treatment decisions.

[0003] Existing methods only consider the weight of features when calculating data similarity, and ignore the weight of data when calculating dissimilarity.

[0004] In a Chinese patent document with publication number CN110957046B, a medical health case knowledge matching method and system are disclosed, which relates to the field of data processing technology. First, medical health case data and external annotation feature information are obtained, and the medical health case data and the external annotation feature information are preprocessed; feature selection processing is performed on the preprocessed medical health case data to obtain key features; then the corresponding feature weights are obtained based on a genetic algorithm; preliminary case similarity is obtained based on the feature weights and the Pearson parameter r method; finally, the final case similarity is obtained based on the preliminary case similarity and the preprocessed external annotation feature information, and similar medical health cases are obtained based on the final case similarity. However, this patent document mainly reduces the impact of poor quality and low-level information on retrieval results by introducing external annotation feature information, thereby improving the accuracy of retrieval results and health diagnosis and treatment decisions, which is essentially different from the method used in the present invention. Summary of the invention

[0005] In view of the defects in the prior art, the object of the present invention is to provide a method and system for the prognosis of heart failure combined with atrial fibrillation and the risk stratification of catheter ablation.

[0006] A method for prognosis of heart failure combined with atrial fibrillation and risk stratification of catheter ablation provided by the present invention comprises:

[0007] Step S1: Collect and pre-process target feature data;

[0008] The characteristic data include basic clinical phenotype characteristic data of the patient;

[0009] The preprocessing includes deletion, screening and variable analysis;

[0010] Step S2: constructing a Cox-PH model using the preprocessed feature data and calculating the corresponding feature weights;

[0011] Step S3: Cluster the targets based on the weight values ​​to obtain risk stratification results.

[0012] Preferably, step S1 includes the following sub-steps:

[0013] Step S1.1: Delete the abnormal case data in the basic clinical phenotype characteristic data of the patient to obtain the cleaned data;

[0014] Step S1.2: performing variable analysis based on the cleaned data; the variable analysis includes univariate analysis and multivariate analysis; arranging the basic clinical phenotype characteristic data of the patient to obtain screening characteristics;

[0015] Step S1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

[0016] Preferably, step S2 includes the following sub-steps:

[0017] Step S2.1: Establish a Cox-PH model by combining dummy variables with endpoint events to obtain the C-index value of the complete model;

[0018] Step S2.2: Obtain the C-index value of the Cox-PH model established by the remaining variables after deducting the single dummy variable, calculate the difference between the C-index value of the Cox-PH model and the complete model, and use the difference as the initial weight of the dummy variable;

[0019] Step S2.3: Take the square root of the initial weights to obtain the balanced feature weights.

[0020] Preferably, step S3 includes clustering the targets using balanced feature weights and the KModes method; and calculating sample distances using the weighted Manhattan distance to obtain risk stratified subgroups.

[0021] Preferably, the step S3 further comprises: further calculating the risk ratio of the two groups of patients in different subgroups in terms of whether or not to undergo catheter ablation, so as to obtain a group suitable for catheter ablation.

[0022] According to the present invention, a prognosis and catheter ablation risk stratification system for heart failure combined with atrial fibrillation includes:

[0023] Module M1: collect and pre-process target feature data;

[0024] The characteristic data include basic clinical phenotype characteristic data of the patient;

[0025] The preprocessing includes deletion, screening and variable analysis;

[0026] Module M2: Use the preprocessed feature data to build the Cox-PH model and calculate the corresponding feature weights;

[0027] Module M3: Cluster the targets based on weight values ​​to obtain risk stratification results.

[0028] Preferably, the module M1 includes the following submodules:

[0029] Module M1.1: Delete abnormal case data from the patient's basic clinical phenotype characteristic data to obtain cleaned data;

[0030] Module M1.2: Perform variable analysis based on the cleaned data; the variable analysis includes univariate analysis and multivariate analysis; sort the basic clinical phenotype characteristic data of the patient to obtain screening characteristics;

[0031] Module M1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

[0032] Preferably, the module M2 includes the following submodules:

[0033] Module M2.1: Use dummy variables to combine with endpoint events to establish the Cox-PH model and obtain the C-index value of the complete model;

[0034] Module M2.2: Get the C-index value of the Cox-PH model established by the remaining variables after deducting a single dummy variable, calculate the difference between it and the C-index value of the complete model, and use the difference as the initial weight of the dummy variable;

[0035] Module M2.3: Take the square root of the initial weights to obtain the balanced feature weights.

[0036] Preferably, the module M3 includes using balanced feature weights and the KModes method to cluster the targets; using the weighted Manhattan distance to calculate the sample distance to obtain risk stratified subgroups.

[0037] Preferably, the module M3 further comprises: further calculating the risk ratio of two groups of patients in different subgroups in terms of whether or not to undergo catheter ablation, to obtain a group suitable for catheter ablation.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention achieves the purpose of obtaining feature weights by taking the square root of the Cox-PH model C-index value based on all variables and deducting the Cox-PH model C-index value established by the remaining variables after deleting a single variable; by adopting the K-Modes clustering method suitable for categorical variables, patients are clustered and grouped considering different feature weights, thereby achieving good risk stratification results for patients with atrial fibrillation and heart failure, and having good practicality.

[0040] 2. The present invention can calculate the risk ratio of catheter ablation and endpoint events for subgroups in the clustering, thereby obtaining a group suitable for catheter ablation. When calculating dissimilarity, the weights of different features can also be considered, which greatly improves the accuracy of the calculation.

[0041] 3. The method provided by the present invention has the potential to promote joint decision-making for patients with atrial fibrillation and heart failure in routine practice. It can identify the group most likely to benefit from catheter ablation through risk stratification and can be widely used in the field of medical intelligent decision-making.

[0042] Other beneficial effects of the present invention will be explained in the specific implementation manner through the introduction of specific technical features and technical solutions. Through the introduction of these technical features and technical solutions, those skilled in the art should be able to understand the beneficial technical effects brought about by the technical features and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0044] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0046] The present invention performs survival analysis based on the basic clinical phenotype characteristics of the patients by associating them with the prognostic endpoint events, and establishes a Cox-PH model to obtain feature weights; further, a K-Modes clustering method that can take feature weights into account is used to cluster the patients to obtain subgroups with good prognostic distinction.

[0047] Reference Figure 1 As shown, a risk stratification approach for the prognosis and catheter ablation of atrial fibrillation and heart failure combined with atrial fibrillation includes:

[0048] Step S1: Collect and pre-process the target’s feature data:

[0049] Step S1.1: Delete the abnormal case data in the basic clinical phenotype characteristic data of the patient to obtain the cleaned data;

[0050] Step S1.2: Perform univariate analysis and multivariate analysis based on the cleaned data to sort out the basic clinical phenotype characteristic data of the patients and obtain screening characteristics;

[0051] Step S1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

[0052] Step S2: Modeling is performed based on the preprocessed feature data, and weights are calculated using the model:

[0053] Step S2.1: Establish a Cox-PH model based on dummy variables and endpoint events to obtain the C-index value of the complete model;

[0054] Step S2.2: Obtain the C-index value of the Cox-PH model established by the remaining variables after deducting the single dummy variable, calculate the difference between the C-index value of the Cox-PH model and the complete model, and use the difference as the initial weight of the dummy variable;

[0055] Step S2.3: Take the square root of the initial weights to obtain the balanced feature weights.

[0056] Step S3: Cluster the targets based on the weight values ​​to obtain the risk stratification results:

[0057] The targets were clustered using balanced feature weights and the K-Modes method. The weighted Manhattan distance was used to calculate sample distances to obtain subgroups differentiated by risk stratification. The risk ratios of catheter ablation and outcome events in different subgroups were further calculated to obtain groups suitable for catheter ablation.

[0058] The above are basic embodiments of the present invention. The solutions of the present invention are further described below through two preferred embodiments.

[0059] Example 1

[0060] Atrial fibrillation combined with heart failure poses a major challenge to patient risk assessment and treatment decision-making. The present invention develops a shared decision-making tool that can stratify the risk of patients with atrial fibrillation and heart failure and guide radiofrequency catheter ablation decisions in this group. The specific steps are as follows:

[0061] Step 1: Check the basic clinical phenotype characteristic data of the patient and delete the case data containing abnormal values ​​to obtain the cleaned data;

[0062] Step 2: Based on the cleaned data, the Cox-PH model in survival analysis was used to perform univariate analysis and multivariate analysis, and the basic clinical phenotype characteristics of the patients were excluded according to the P value and the doctor's guidance and suggestions to obtain the screening characteristics;

[0063] Step 3: The selected feature reference baseline table is converted into categorical variables, and the categorical variables are further converted into dummy variables;

[0064] Step 4: The dummy variables were combined with the endpoint event (all-cause death) to establish the Cox-PH model and obtain the C-index value of the complete model;

[0065] Step 5: Obtain the C-index value of the Cox-PH model established by the remaining variables after deducting the single dummy variable, and calculate the difference between it and the C-index value of the complete model as the initial weight of the dummy variable;

[0066] Step 6: Take the square root of the initial weights to obtain the balanced feature weights;

[0067] Step 7: Apply the balanced feature weights and corresponding data, and cluster the patients using the KModes method applicable to categorical variables. In particular, when calculating the sample distance, Manhattan (L1) distance is used and feature weights are considered to obtain subgroups with good risk stratification. The present invention can also consider the weights of different features when calculating dissimilarity.

[0068] Step 8: Further calculate the risk ratio of the two groups of patients in different subgroups and determine whether to undergo catheter ablation to obtain the group suitable for catheter ablation, and finally obtain a shared decision-making tool to support risk stratification and catheter ablation decision-making for patients with heart failure and atrial fibrillation.

[0069] Hazard ratio is a common reference indicator in survival analysis. Hazard ratio, formally known as Hazard Ratio in English, is the ratio of two hazard rates. Hazard rate is the percentage of events occurring per unit time in the total number of subjects.

[0070] The risk ratio calculation formula is:

[0071]

[0072] Hazard Ratio ( HR) = risk (event rate) in the intervention group / risk (event rate) in the control group

[0073] The hazard rate calculation formula is:

[0074]

[0075] In the above formula, is the benchmark risk function, is the characteristic variable, is the coefficient corresponding to the characteristic variable.

[0076] The present invention achieves the purpose of obtaining feature weights by taking the square root of the Cox-PH model C-index value established by deducting the remaining variables after deleting a single variable based on the Cox-PH model C-index value of all variables; by adopting the KModes clustering method suitable for categorical variables, patients are clustered and grouped considering different feature weights, thereby achieving good risk stratification results for patients with atrial fibrillation and heart failure, and having good practicality; the risk ratio of whether to use catheter ablation can be calculated for the subgroups in the grouping, and the group suitable for catheter ablation can be obtained. When calculating dissimilarity, the weights of different features can also be considered, which greatly improves the accuracy of the calculation.

[0077] Example 2

[0078] Patients with atrial fibrillation (AF) and heart failure (HF) face significant challenges in risk assessment and treatment decisions. There is a need to develop a shared decision-making tool to help risk stratify and guide decisions for radiofrequency catheter ablation (RFCA) of AF in patients with HF.

[0079] Unsupervised clustering methods combined with supervised learning models were used to screen patients who met the HF diagnostic criteria from the Chinese Atrial Fibrillation Registry, which has been continuously recruited since 2011, to derive a shared decision-making tool. External validation of the tool was performed in the CABANA trial and two prospective cohorts of heart failure. Data analysis was performed from November 2023 to May 2024.

[0080] In this example, the China Atrial Fibrillation Registry and the two heart failure cohorts were multicenter cohorts that recruited patients diagnosed with AF or HF. CABANA is a multicenter, randomized, open-label trial that recruited patients with AF; some of these patients also met the clinical diagnosis of HF. The main outcomes and measures were: composite events (all-cause death, non-fatal ischemic stroke / transient ischemic attack, and major bleeding), cardiovascular death, cardiovascular hospitalization, and AF recurrence.

[0081] Among 1774 patients in the derivation cohort, three clusters were identified using a shared decision tool based on 24 easily accessible clinical characteristics. The incidence rate of the composite event (per 100 person-years) was highest in cluster 3 (14.72), followed by cluster 1 (9.24), and lowest in cluster 2 (5.66) (Log-rank P<0.001). Compared with cluster 1, the adjusted hazard ratio (HR) for the composite event was 0.64 (95% CI, 0.53-0.78) in cluster 2 and 1.44 (95% CI, 1.16-1.80) in cluster 3. The risk stratification results for all-cause and cardiovascular mortality were consistent across clusters. The tool showed consistent risk stratification across different left ventricular ejection fraction (LVEF) ranges and in the external validation dataset. Compared with medical therapy, RFCA significantly improved outcomes in cluster 1 patients (HR for the composite event in the AF cohort = 0.34; 95% CI, 0.22-0.54; HR for the composite event in the CABANA dataset, which consisted mainly of European and American patients = 0.52; 95% CI, 0.29-0.94).

[0082] This example demonstrates the potential of machine learning-based tools to facilitate shared decision making in patients with AF and HF in routine practice, identifying the group most likely to benefit from RFCA through risk stratification.

[0083] The present invention also provides a risk stratification system for the prognosis and catheter ablation of heart failure combined with atrial fibrillation. The risk stratification system for the prognosis and catheter ablation of heart failure combined with atrial fibrillation can be realized by executing the process steps of the risk stratification method for the prognosis and catheter ablation of heart failure combined with atrial fibrillation, that is, a person skilled in the art can understand the risk stratification method for the prognosis and catheter ablation of heart failure combined with atrial fibrillation as a preferred implementation of the risk stratification system for the prognosis and catheter ablation of heart failure combined with atrial fibrillation.

[0084] Specifically, a risk stratification system for prognosis and catheter ablation of heart failure with atrial fibrillation includes:

[0085] Module M1: collect and pre-process target feature data;

[0086] The characteristic data include basic clinical phenotype characteristic data of the patient;

[0087] The preprocessing includes deletion, screening and variable analysis;

[0088] Module M2: Use the preprocessed feature data to build the Cox-PH model and calculate the corresponding feature weights;

[0089] Module M3: Cluster the targets based on weight values ​​to obtain risk stratification results.

[0090] The module M1 includes the following submodules:

[0091] Module M1.1: Delete abnormal case data from the patient's basic clinical phenotype characteristic data to obtain cleaned data;

[0092] Module M1.2: Perform variable analysis based on the cleaned data; the variable analysis includes univariate analysis and multivariate analysis; sort the basic clinical phenotype characteristic data of the patient to obtain screening characteristics;

[0093] Module M1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

[0094] The module M2 includes the following submodules:

[0095] Module M2.1: Use dummy variables to combine with endpoint events to establish the Cox-PH model and obtain the C-index value of the complete model;

[0096] Module M2.2: Get the C-index value of the Cox-PH model established by the remaining variables after deducting a single dummy variable, calculate the difference between it and the C-index value of the complete model, and use the difference as the initial weight of the dummy variable;

[0097] Module M2.3: Take the square root of the initial weights to obtain the balanced feature weights.

[0098] The module M3 includes using balanced feature weights and the KModes method to cluster the targets; using the weighted Manhattan distance to calculate the sample distance to obtain risk stratified subgroups.

[0099] The module M3 also includes: further calculating the risk ratio of the two groups of patients in different subgroups whether to use catheter ablation, so as to obtain a group suitable for catheter ablation.

[0100] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0101] In the description of the present application, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0102] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for the prognosis of heart failure combined with atrial fibrillation and risk stratification of catheter ablation, characterized in that: include: Step S1: collecting and preprocessing patient characteristic data; The patient characteristic data include basic clinical phenotype characteristic data of the patient; The preprocessing includes deletion, screening and variable analysis; Step S2: constructing a Cox-PH model using the preprocessed feature data and calculating the corresponding feature weights; The dummy variables were combined with the endpoint events to establish the Cox-PH model and obtain the C-index value of the complete model; Get the C-index value of the Cox-PH model established by the remaining variables after deducting a single dummy variable, calculate the difference between it and the C-index value of the complete model, and use the difference as the initial weight of the dummy variable; Take the square root of the initial weights to obtain the balanced feature weights; Step S3: Cluster the patients based on the weight values ​​to obtain risk stratification results.

2. A method for prognosis and catheter ablation risk stratification of heart failure combined with atrial fibrillation according to claim 1, characterized in that: The step S1 comprises the following sub-steps: Step S1.1: Delete the abnormal case data in the basic clinical phenotype characteristic data of the patient to obtain the cleaned data; Step S1.2: performing variable analysis based on the cleaned data; the variable analysis includes univariate analysis and multivariate analysis; arranging the basic clinical phenotype characteristic data of the patient to obtain screening characteristics; Step S1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

3. The method for prognosis and catheter ablation risk stratification of heart failure combined with atrial fibrillation according to claim 1, characterized in that: The step S3 includes using the balanced feature weights and KModes method to cluster the patients; using the weighted Manhattan distance to calculate the sample distance to obtain risk stratified subgroups.

4. A method for prognosis and catheter ablation risk stratification of heart failure combined with atrial fibrillation according to claim 3, characterized in that: The step S3 further includes: further calculating the risk ratio of the two groups of patients in different subgroups, whether to use catheter ablation or not, to obtain a group suitable for catheter ablation.

5. A prognosis and catheter ablation risk stratification system for heart failure combined with atrial fibrillation, characterized in that: include: Module M1: Collect and pre-process patient characteristic data; The patient characteristic data include basic clinical phenotype characteristic data of the patient; The preprocessing includes deletion, screening and variable analysis; Module M2: Use the preprocessed feature data to build the Cox-PH model and calculate the corresponding feature weights; The dummy variables were combined with the endpoint events to establish the Cox-PH model and obtain the C-index value of the complete model; Get the C-index value of the Cox-PH model established by the remaining variables after deducting a single dummy variable, calculate the difference between it and the C-index value of the complete model, and use the difference as the initial weight of the dummy variable; Take the square root of the initial weights to obtain the balanced feature weights; Module M3: Cluster patients based on weight values ​​to obtain risk stratification results.

6. A risk stratification system for prognosis and catheter ablation of heart failure combined with atrial fibrillation according to claim 5, characterized in that: The module M1 Includes the following submodules: Module M1.1: Delete abnormal case data from the patient's basic clinical phenotype characteristic data to obtain cleaned data; Module M1.2: Perform variable analysis based on the cleaned data; the variable analysis includes univariate analysis and multivariate analysis; sort the basic clinical phenotype characteristic data of the patient to obtain screening characteristics; Module M1.3: Convert the reference baseline table of screening characteristics into categorical variables, and further convert the categorical variables into dummy variables.

7. A risk stratification system for prognosis and catheter ablation of heart failure combined with atrial fibrillation according to claim 5, characterized in that: The module M3 includes using balanced feature weights and the KModes method to cluster patients; using the weighted Manhattan distance to calculate sample distances to obtain risk stratified subgroups.

8. A risk stratification system for prognosis and catheter ablation of heart failure combined with atrial fibrillation according to claim 7, characterized in that: The module M3 also includes: further calculating the risk ratio of the two groups of patients in different subgroups whether to use catheter ablation, so as to obtain a group suitable for catheter ablation.

Images