Cardiovascular and cerebrovascular disease risk prediction method and system

By constructing a risk factor database and optimizing risk level judgment model, the problem of insufficient ability to analyze real-time health data in traditional methods is solved, more accurate risk assessment and personalized prevention measures are achieved, and the prevention effect of cardiovascular and cerebrovascular diseases is improved.

CN118711822BActive Publication Date: 2024-11-08BEIJING ANCHI ZHONGKAI TECH CO LTD

Patent Information

Application Number
CN202411195076.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-08
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

Traditional cardiovascular and cerebrovascular disease risk prediction methods rely on static models, lack the ability to analyze patients' real-time health data, and cannot effectively process missing data, resulting in limited accuracy of evaluation results, affecting the timeliness and effectiveness of preventive measures.

Method used

Using text mining and key information extraction based on clinical research and medical guidelines, a risk factor database is constructed, and weight adjustment is made through recursive feature elimination and adaptive enhancement learning. Combined with long and short-term memory networks to analyze time series data, optimize risk level judgment model, and provide Personalized preventive measures.

Benefits of technology

More accurate risk assessment and personalized preventive measures have been achieved, the early identification and prevention effects of high-risk populations have been improved, and the incidence and mortality of cardiovascular and cerebrovascular diseases have been reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118711822B_ABST
    Figure CN118711822B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of risk prediction technology, specifically to a method and system for predicting the risk of cardiovascular and cerebrovascular diseases, comprising the following steps: based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed to obtain a risk factor database. The present invention, by using recursive feature elimination and adaptive reinforcement learning, performs weighted calculation of risk factors in a weight adjustment model, and divides the risk level in combination with a gradient boosting tree, thereby achieving a more accurate risk assessment, and by analyzing time series data in combination with a long short-term memory network, the risk level judgment model is optimized, so that the risk prediction is closer to the actual health status of the patient, so as to accurately identify high-risk groups at an early stage, provide targeted preventive measures, and effectively reduce the incidence and mortality of cardiovascular and cerebrovascular diseases.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of risk prediction, and in particular to a method and system for predicting cardiovascular and cerebrovascular disease risks. Background Art

[0002] The field of risk prediction technology focuses on developing algorithms and models to assess the probability of an individual developing a specific health condition in the future. Including in the field of cardiovascular and cerebrovascular diseases, this type of technology aims to predict an individual's risk of future cardiovascular and cerebrovascular events by analyzing a variety of data such as a patient's medical history, lifestyle factors, and biomarkers. The application of this technology field is of great significance for early diagnosis, the formulation of preventive measures, and the rational allocation of medical resources.

[0003] Among them, the cardiovascular and cerebrovascular disease risk prediction method is a method that uses statistics, data analysis, machine learning and other technologies to predict the probability of an individual developing cardiovascular and cerebrovascular diseases. Its purpose is to identify high-risk groups so that early intervention and preventive measures can be taken to reduce the incidence and mortality of cardiovascular and cerebrovascular diseases. Through accurate prediction, medical service providers can develop targeted health management plans for patients, including diet, exercise and drug treatment, to reduce the burden of disease and improve the quality of life.

[0004] Traditional methods rely on static risk assessment models and lack the ability to dynamically analyze patients' real-time health data, especially when dealing with missing data, which limits the accuracy of the assessment results. Failure to fully utilize time series data to analyze the changing trends of patients' health status leads to the inability to adjust patients' risk levels in real time, affecting the timeliness and effectiveness of preventive measures. The above shortcomings limit the application effect of traditional methods in individualized health management and early intervention, and fail to provide patients with the best health management measures. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for predicting the risk of cardiovascular and cerebrovascular diseases.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for predicting cardiovascular and cerebrovascular disease risk, comprising the following steps:

[0007] S1: Based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed to obtain a risk factor database;

[0008] S2: Based on the risk factor database, conduct regular data analysis and update the weights of risk factors, including evaluating the impact of clinical research on the existing model, and adjusting the weights according to the evaluation results to generate a weight adjustment model;

[0009] S3: Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to perform weighted calculations on the patient's risk factors, and a risk score for each patient is calculated based on the patient's historical health data to obtain individual risk score data;

[0010] S4: Based on the individual risk score data, collect the time series data of the patient's biomarkers, select key time points for analysis, identify potential risk change trends, and obtain risk development trend information;

[0011] S5: Based on the risk development trend information, use the long short-term memory network to conduct comparative analysis of data at multiple time points, combine clinical research results, determine the patient's risk level, optimize the prediction model, and establish a risk level judgment model;

[0012] S6: Based on the risk level judgment model, analyze the patient's health status and living background, formulate targeted measures, including adjustments to living habits and proposals for medical measures, and generate risk prevention proposals;

[0013] S7: Follow up and evaluate the effectiveness of the proposed risk prevention measures, record the changing trend of the patient's health status after the implementation of the measures, and evaluate the expected effects of the preventive measures to obtain a prevention effect evaluation record.

[0014] The present invention has improvements in that the risk factor database includes the patient's age, gender, genetic factors, living habits, historical medical history and recent medical interventions; the weight adjustment model includes the risk factor weights associated with the disease, the update cycle and the weight adjustment mechanism; the risk development trend information includes the time series analysis results, the changing trends of key risk indicators, and the early warning signals of potential risk events; the risk level judgment model includes the risk level classification standards, the prediction model optimization strategy, and the risk level assessment method; the prevention effect evaluation record includes the comparative analysis of the patient's health indicators before and after the implementation of the measures, and the long-term tracking monitoring information; the biomarker specifically refers to a measurable indicator in the body, which can reflect the health status, physiological process, pathological state or drug treatment, such as the measurement of glucose, cholesterol, protein, hormone levels, blood pressure, heart rate or other body functions in the blood, etc.

[0015] The present invention improves that, based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed. The steps of obtaining a risk factor database are specifically as follows:

[0016] S101: Based on existing clinical research and medical guidelines, extract risk factor information associated with cardiovascular and cerebrovascular diseases, and preliminarily organize the information to obtain a preliminary list of risk factors;

[0017] S102: Based on the preliminary risk factor list, data cleaning is performed, including removing duplicate items, verifying the accuracy of information, and supplementing missing risk factor descriptions to obtain an adjusted risk factor list;

[0018] S103: constructing a mapping table of the correlation between risk factors and cardiovascular and cerebrovascular diseases according to the adjusted risk factor list, and obtaining a risk factor database by evaluating the strength of the correlation research information of the risk factors.

[0019] The present invention is improved in that, based on the risk factor database, regular data analysis is performed to update the weights of risk factors, including evaluating the impact of clinical research on the existing model and adjusting the weights according to the evaluation results. The steps of generating the weight adjustment model are specifically as follows:

[0020] S201: Based on the risk factor database, perform preliminary weight assignment of risk factors, perform initial weight allocation according to mapping strength information of the relationship between risk factors and cardiovascular and cerebrovascular diseases, and generate an initial weight model;

[0021] S202: performing periodic evaluation of weights based on the initial weight model, including collecting clinical research results, adjusting existing risk factor weights to reflect the criticality of new medical information, and obtaining a weight update model;

[0022] S203: Evaluate the weight update model and determine the risk factor weights, and generate a weight adjustment model by comparing the impact of new and old research data on the risk factor weights.

[0023] The present invention is improved in that, based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to perform weighted calculation on the patient's risk factors, and the risk score of each patient is calculated according to the patient's historical health data. The steps of obtaining individual risk score data are specifically as follows:

[0024] S301: Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to traverse the patient's historical health data, and weighted calculation is performed on each risk factor. If risk factor information is missing in the patient data, the default value in the model is used to fill in and obtain a weighted risk factor score;

[0025] The recursive feature elimination is performed according to the formula:

[0026]

[0027] Calculate feature weights and generate feature importance rankings;

[0028] in, Indicates The weight of the feature, It is The feature in The value of the iteration, is the coefficient of the feature in the model, is the historical health status classification, is the lifestyle stability index, Represents the total number of features in the feature set;

[0029] The adaptive enhancement algorithm is based on the formula:

[0030]

[0031] Optimize weight adjustment to obtain weighted risk factor scores;

[0032] in, It is Round The weight of the samples, It is Round The weight of the samples, It is The weight of the round classifier, is the true label of the sample, It is The prediction result of the wheel classifier, is the environmental adaptation coefficient, is the access to medical resources, is the normalization factor, exp represents the exponential power of the natural exponential function e;

[0033] S302: Summarize the weighted risk factor scores, apply the gradient boosting tree, calculate the total score by accumulating the scores of each risk factor, and divide the risk level according to the total score to obtain the risk level allocation result;

[0034] S303: Based on the risk level allocation result, a risk score is allocated to each patient to reflect the patient's risk of cardiovascular and cerebrovascular diseases, and obtain individual risk score data.

[0035] The present invention has been improved in that, based on the individual risk score data, the time series data of the patient's biomarkers are collected, key time points are selected for analysis, potential risk change trends are identified, and the steps of obtaining risk development trend information are specifically as follows:

[0036] S401: Based on the individual risk score data, collect the time series data of the patient's biomarkers, including blood pressure, blood sugar and cholesterol indicators, and verify whether each data is recorded in time sequence to generate a biomarker data record;

[0037] S402: Analyze the biomarker data records, select key time points, including before and after disease diagnosis, treatment start and end time nodes, analyze the data at the time points, identify data changes during the differential health state conversion period, and obtain key time point analysis results;

[0038] S403: Based on the analysis results of the key time points, identify the change trend in the time series, including continuous rise, fall or fluctuation patterns, determine the development trend of the patient's cardiovascular and cerebrovascular disease risk, and obtain risk development trend information.

[0039] The present invention has been improved in that, based on the risk development trend information, a long short-term memory network is used to perform comparative analysis of data at multiple time points, and the patient's risk level is determined in combination with clinical research results, and the prediction model is optimized. The steps of establishing a risk level judgment model are specifically as follows:

[0040] S501: Based on the risk development trend information, using a long short-term memory network, select data at multiple key time points, including the time points of patient biomarker detection, loop through the time points, analyze the data changes at each time point, and determine the change trend of the data at the different time points to obtain the data trend at the key time points;

[0041] The long short-term memory network is based on the formula:

[0042]

[0043] Calculate the changing trend of data at different time points to obtain the data trend at key time points;

[0044] in, For time point The hidden state of For time point Input, is the weight matrix input to the hidden layer, is the weight matrix from hidden layer to hidden layer, is the bias term of the hidden layer, is the activation function, is the time window size, is the dynamically adjusted learning rate, Represents the number of memory cells, is the application of the activation function;

[0045] S502: Comparative analysis is performed on the data trends at the key time points to determine the commonalities and differences in the data changes at the time points, and combined with new clinical research results, the impact of the changes on the patient's risk level is evaluated to generate a risk impact analysis record;

[0046] S503: Based on the risk impact analysis record, the risk level is refined through the random forest algorithm, and according to the information in the clinical research results, the existing risk scoring model is optimized to establish a risk level judgment model.

[0047] The present invention is improved in that, based on the risk level judgment model, the patient's health status and life background are analyzed, and targeted measures are formulated, including the adjustment of living habits and the proposal of medical measures. The steps of generating risk prevention proposal measures are specifically as follows:

[0048] S601: Based on the risk level judgment model, analyzing the patient's historical health data and life background, including eating habits, exercise frequency, and historical medical history, identifying adjustable risk factors, and generating an adjustable risk factor list;

[0049] S602: Formulate lifestyle adjustment measures and medical intervention proposals for the list of adjustable risk factors, including increasing physical activity and adjusting diet structure, matching corresponding measures for each risk factor, and obtaining targeted intervention measures;

[0050] S603: Based on the targeted intervention measures, combined with the patient's acceptance capacity and situation, the intervention measures are optimized and adjusted, and the feasibility and targetedness of the measures are verified to generate risk prevention proposals.

[0051] The present invention is improved in that the risk prevention proposed measures are tracked and the effect is evaluated, the trend of changes in the patient's health status after the measures are implemented is recorded, and the expected effect of the preventive measures is evaluated. The steps of obtaining the preventive effect evaluation record are specifically as follows:

[0052] S701: Track and evaluate the effectiveness of the proposed risk prevention measures, collect data on changes in health status, including biomarkers and changes in daily living habits, record data after the measures are implemented, and generate a tracking record of the implementation of the measures;

[0053] S702: Based on the data in the tracking record of the implementation of the measure, analyze the changes in the health data before and after the measure, identify positive changes or directions that need to be adjusted, and obtain health status change analysis results;

[0054] S703: Based on the health status change analysis results, the effects of the preventive measures are evaluated, and the effectiveness and improvement points of the measures are determined to obtain a preventive effect evaluation record.

[0055] A cardiovascular and cerebrovascular disease risk prediction system, the system comprising:

[0056] The risk factor identification module identifies cardiovascular and cerebrovascular risk factors based on existing clinical research and medical guidelines, including age, gender, genetic factors and lifestyle habits, removes data duplication, verifies information accuracy, and supplements missing risk factor descriptions to generate a risk factor library;

[0057] The weight adjustment module performs preliminary weight assignment of risk factors based on the risk factor library, collects clinical research results according to the initial weight model, periodically evaluates and adjusts the existing risk factor weights, and constructs a weight adjustment model by comparing the impact of new and old research data on risk factor weights;

[0058] The individual risk scoring module uses recursive feature elimination and adaptive reinforcement learning algorithms based on the weight adjustment model to traverse the patient's historical health data, perform weighted calculations on each risk factor, summarize the scores, and divide the risk levels to obtain individual risk score data;

[0059] The risk trend analysis module collects time series data, including biomarkers, based on the individual risk score data, selects key time points for analysis, identifies risk change trends, and constructs trend change analysis;

[0060] The risk level judgment module is based on the trend change analysis, uses the long short-term memory network, selects data from multiple key time points, conducts comparative analysis, and optimizes the existing risk scoring model in combination with clinical research results to generate a risk level model;

[0061] The risk prevention and assessment module analyzes the patient's health status and living background based on the risk level model, formulates targeted measures, including lifestyle adjustments and medical measures, and conducts tracking and effect evaluation, records the changing trends of the patient's health status after the implementation of the measures, and establishes prevention effect evaluation records.

[0062] Compared with the prior art, the advantages and positive effects of the present invention are:

[0063] In the present invention, by using recursive feature elimination and adaptive reinforcement learning, weighted calculation of risk factors is performed in the weight adjustment model, and the risk level is divided in combination with the gradient boosting tree, so as to achieve more accurate risk assessment. When processing missing data, a default value filling strategy is adopted to ensure the integrity and continuity of the scoring system. By analyzing the time series data in combination with the long short-term memory network, the risk level judgment model is optimized, so that the risk prediction is closer to the actual health status of the patient, so as to accurately identify high-risk groups in the early stage and provide targeted preventive measures, thereby effectively reducing the incidence and mortality of cardiovascular and cerebrovascular diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A flow chart of a method for predicting cardiovascular and cerebrovascular disease risks proposed by the present invention;

[0065] Figure 2 This is a schematic diagram of the detailed flow of step S1 in the cardiovascular and cerebrovascular disease risk prediction method proposed in the present invention;

[0066] Figure 3 The present invention provides a schematic diagram of the detailed process of step S2 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0067] Figure 4 The present invention provides a schematic diagram of the detailed process of step S3 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0068] Figure 5 The present invention provides a schematic diagram of the detailed process of step S4 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0069] Figure 6 The present invention provides a schematic diagram of the detailed process of step S5 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0070] Figure 7 The present invention provides a detailed flowchart of step S6 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0071] Figure 8 The present invention provides a schematic diagram of the detailed process of step S7 in the method for predicting cardiovascular and cerebrovascular disease risks;

[0072] Fig. 9 A module diagram of a cardiovascular and cerebrovascular disease risk prediction system is proposed for the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0074] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0075] Example

[0076] See also Figure 1 The present invention provides a technical solution: a method for predicting cardiovascular and cerebrovascular disease risks, comprising the following steps:

[0077] S1: Based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed to obtain a risk factor database;

[0078] S2: Based on the risk factor database, conduct regular data analysis and update the weights of risk factors, including evaluating the impact of clinical research on the existing model, and adjust the weights according to the evaluation results to generate a weight adjustment model;

[0079] S3: Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to perform weighted calculations on the patient's risk factors, and the risk score of each patient is calculated based on the patient's historical health data to obtain individual risk score data;

[0080] S4: Based on individual risk score data, collect time series data of patient biomarkers, select key time points for analysis, identify potential risk change trends, and obtain risk development trend information;

[0081] S5: Based on the risk development trend information, the long short-term memory network is used to conduct comparative analysis of data at multiple time points, combined with clinical research results, to determine the patient's risk level, optimize the prediction model, and establish a risk level judgment model;

[0082] S6: Based on the risk level judgment model, analyze the patient's health status and living background, formulate targeted measures, including adjustments to lifestyle habits and medical measures, and generate risk prevention proposals;

[0083] S7: Follow up and evaluate the effectiveness of proposed risk prevention measures, record the changing trend of the patient's health status after the implementation of the measures, and evaluate the expected effects of the preventive measures to obtain a prevention effect evaluation record.

[0084] The risk factor database includes the patient's age, gender, genetic factors, lifestyle habits, medical history and recent medical interventions. The weight adjustment model includes the disease-associated risk factor weights, update cycle and weight adjustment mechanism. The risk development trend information includes time series analysis results, changing trends of key risk indicators, and early warning signals of potential risk events. The risk level judgment model includes risk level classification standards, prediction model optimization strategies, and risk level assessment methods. The prevention effect evaluation records include comparative analysis of patient health indicators before and after the implementation of measures, and long-term follow-up monitoring information.

[0085] See also Figure 2Based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed. The specific steps to obtain the risk factor database are as follows:

[0086] S101: Based on existing clinical research and medical guidelines, extract risk factor information associated with cardiovascular and cerebrovascular diseases, and preliminarily organize the information to obtain a preliminary list of risk factors;

[0087] S102: Based on the preliminary risk factor list, data cleaning is performed, including removing duplicates, verifying the accuracy of information, and supplementing missing risk factor descriptions to obtain an adjusted risk factor list;

[0088] S103: Based on the adjusted risk factor list, a mapping table of the correlation between risk factors and cardiovascular and cerebrovascular diseases is constructed, and a risk factor database is obtained by evaluating the strength of the correlation research information of the risk factors.

[0089] In S101, based on existing clinical research and medical guidelines, natural language processing (NLP) algorithms were used, and Python language was combined with NLTK and spaCy libraries to perform text preprocessing operations, including using spaCy's nlp object for text segmentation, using NLTK's stopwords set to remove stop words, and using spaCy's lemmatization method for word form restoration. Information on risk factors associated with cardiovascular and cerebrovascular diseases was extracted, and specific risk factor entities in the text were identified by executing spaCy's EntityRecognition function to generate a preliminary risk factor list.

[0090] In S102, based on the preliminary risk factor list, data cleaning technology is used, specifically the Pandas library in Python, and the DataFrame structure is used to store risk factor data. The drop_duplicates() method is executed to remove duplicates in the list. The isnull() and fillna() methods are used to check and fill in missing risk factor descriptions to ensure the completeness and accuracy of each risk factor information and generate an adjusted risk factor list.

[0091] In S103, according to the adjusted risk factor list, graph database technology, specifically Neo4j platform, is used to define nodes of risk factors and cardiovascular and cerebrovascular diseases using Cypher query language, and risk factor nodes are constructed through CREATE statements. Attributes including risk factor name and description are set, and MATCH statements and CREATE statements are used to establish association edges between risk factors and cardiovascular and cerebrovascular diseases. Association attributes are set for the edges to indicate the strength of the association between risk factors and diseases, thereby generating a risk factor database.

[0092] See also Figure 3 Based on the risk factor database, regular data analysis is performed to update the weights of risk factors, including evaluating the impact of clinical research on the existing model and adjusting the weights based on the evaluation results. The specific steps for generating the weight adjustment model are as follows:

[0093] S201: Based on the risk factor database, perform preliminary weight assignment of risk factors, perform initial weight allocation according to the mapping strength information of the relationship between risk factors and cardiovascular and cerebrovascular diseases, and generate an initial weight model;

[0094] S202: Based on the initial weight model, periodic evaluation of the weights is performed, including collecting clinical research results, adjusting the weights of existing risk factors to reflect the criticality of new medical information, and obtaining a weight update model;

[0095] S203: Evaluate the weight update model and determine the risk factor weights. Generate a weight adjustment model by comparing the impact of new and old research data on the risk factor weights.

[0096] S201: Based on the risk factor database, linear regression analysis method was used and data processing was performed using R language. First, risk factor data was imported using the read.csv() function of R language. The lm() function was used to perform linear modeling on the strength of the association between risk factors and cardiovascular and cerebrovascular diseases. Model parameters were set, including formula parameters that set risk factors as independent variables and the incidence of cardiovascular and cerebrovascular diseases as dependent variables. Model fitting was performed, and the model fit was evaluated using the summary() function. The regression coefficient of each risk factor was calculated as its initial weight to generate an initial weight model.

[0097] S202: Based on the initial weight model, the time series analysis method is used, and the R language is continued to be used to periodically evaluate the weights. First, the newly collected clinical research result data are converted into time series data through the ts() function, and the weights are adjusted based on the autoregressive integrated moving average model using the Arima() function. The model parameters are set, including the difference order, the number of autoregressive terms, and the number of moving average terms, to reflect the trend and seasonal changes in the time series data. The model is fitted, and the future weight change trend is predicted through the forecast() function to obtain the weight update model.

[0098] S203: Evaluate the weight update model and determine the risk factor weights. Use the ggplot2 package of R language to visualize the data. First, use the ggplot() function to create a graphic object, set the data source to the weight update model, use the geom_line() function to draw a weight change trend chart to intuitively show how the risk factor weights change over time, add data points through the geom_point() function, and highlight the differences before and after the weight adjustment. Finally, use the ggsave() function to save the graph and generate the weight adjustment model.

[0099] See also Figure 4 Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to weight the patient's risk factors, and the risk score of each patient is calculated based on the patient's historical health data. The specific steps to obtain the individual risk score data are as follows:

[0100] S301: Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to traverse the patient's historical health data and perform weighted calculations on each risk factor. If risk factor information is missing in the patient data, the default value in the model is used to fill in the information and obtain a weighted risk factor score.

[0101] S302: Summarize the weighted risk factor scores, apply the gradient boosting tree, calculate the total score by accumulating the scores of each risk factor, and divide the risk level according to the total score to obtain the risk level allocation result;

[0102] S303: Based on the risk level allocation result, a risk score is allocated to each patient to reflect the patient's risk of cardiovascular and cerebrovascular diseases, and obtain individual risk score data.

[0103] S301: Based on the weight adjustment model, the recursive feature elimination (RFE) algorithm and the adaptive boosting (AdaBoost) algorithm are used. The Python scikit-learn library is used to traverse the patient's historical health data, and each risk factor is weighted. Feature selection is performed through the RFE algorithm to determine the most important risk factors for predicting cardiovascular and cerebrovascular diseases. If risk factor information is missing in the patient data, the default values ​​preset by the AdaBoost algorithm are used to fill it in. The fit() method is used to apply the RFE algorithm to the dataset, and the selector attribute ranking_ is used to indicate the importance ranking of the risk factors. At the same time, AdaBoost uses the fit() method to train the model to adapt to the uncertainty or missing values ​​in the data to generate weighted risk factor scores.

[0104] Recursive feature elimination, according to the formula:

[0105]

[0106] Calculate feature weights and generate feature importance rankings;

[0107] in, Indicates The weight of the feature, It is The feature in The value of the iteration, is the coefficient of the feature in the model, is the historical health status classification, is the lifestyle stability index, Represents the total number of features in the feature set.

[0108] Execution process:

[0109] Select a machine learning model as the base model, set the feature set, and apply the formula to calculate the weight for each feature in the feature set. The features are ranked, the features with the lowest weight are removed, and this process is repeated in the next iteration until the predetermined number of features is reached or the stopping condition is met, and the final prediction model is built using the filtered feature set.

[0110] Adaptive enhancement algorithm, according to the formula:

[0111]

[0112] Optimize weight adjustment to obtain weighted risk factor scores;

[0113] in, It is Round The weight of the samples, It is Round The weight of the samples, It is The weight of the round classifier, is the true label of the sample, It is The prediction result of the wheel classifier, is the environmental adaptation coefficient, is the access to medical resources, is the normalization factor, and exp represents the exponential power of the natural exponential function e.

[0114] Execution process:

[0115] Assign the same initial weight to each sample in the data set, use the formula for multiple rounds of iterations, update the sample weights based on the predicted results and actual results of each round, give priority to samples that were previously predicted incorrectly, and combine all weak learners generated during the training process into a strong learner. Each weak learner is ranked according to its performance (through different weights.

[0116] S302: Summarize the weighted risk factor scores and use the gradient boosting tree (GBT) algorithm to calculate the patient's total risk score by accumulating the scores of each risk factor. Use the GradientBoostingClassifier class in Python's scikit-learn library to train the model through the fit() method and the decision_function() method to calculate the total score of each instance. The risk level is divided according to the total score to generate the risk level allocation result.

[0117] S303: According to the risk level allocation result, a risk score is allocated to each patient, and the calculated total risk score is converted into a risk level using the predict() method of the gradient boosting tree algorithm to reflect the patient's risk of cardiovascular and cerebrovascular diseases and obtain individual risk score data.

[0118] See also Figure 5 Based on individual risk score data, the time series data of patient biomarkers are collected, key time points are selected for analysis, and potential risk change trends are identified. The specific steps for obtaining risk development trend information are as follows:

[0119] S401: Based on the individual risk score data, collect the time series data of the patient's biomarkers, including blood pressure, blood sugar and cholesterol indicators, and verify whether each data is recorded in chronological order to generate biomarker data records;

[0120] S402: Analyze biomarker data records, select key time points, including before and after disease diagnosis, treatment start and end time nodes, analyze the data at the time points, identify data changes during the transition period of differential health states, and obtain key time point analysis results;

[0121] S403: Based on the analysis results of key time points, identify the changing trends in the time series, including continuous rise, fall or fluctuation patterns, determine the development trend of the patient's cardiovascular and cerebrovascular disease risk, and obtain risk development trend information.

[0122] S401: Based on individual risk score data, a time series analysis method is used to collect the time series data of patient biomarkers, including blood pressure, blood sugar and cholesterol indicators, using Python's Pandas library. The to_datetime() function of the Pandas library is used to ensure that each data is recorded in chronological order. The sort_values() function is used to sort by time to ensure that the data order is correct and generate biomarker data records.

[0123] S402: Analyze biomarker data records, select key time points, and use time series analysis techniques, including Python's DateTimeIndex function, to calibrate the time nodes before and after disease diagnosis, and the start and end of treatment. Use rolling() and mean() functions to perform moving average analysis on the data at the specified time points to smooth short-term fluctuations and identify data changes during the transition period of differential health states, and obtain key time point analysis results.

[0124] S403: Based on the analysis results of key time points, a time series trend analysis method is used, including linear regression analysis using Python's NumPy library to identify changing trends in time series, such as continuous rise, fall, or fluctuation patterns, and polyfit() function is used to fit time series data to identify and evaluate the development trend of patients' cardiovascular and cerebrovascular disease risks and obtain risk development trend information.

[0125] See also Figure 6 Based on the risk development trend information, the long short-term memory network is used to conduct comparative analysis of data at multiple time points. Combined with clinical research results, the patient's risk level is determined and the prediction model is optimized. The specific steps for establishing the risk level judgment model are as follows:

[0126] S501: Based on the risk development trend information, using the long short-term memory network, select data at multiple key time points, including the time points of patient biomarker detection, loop through the time points, analyze the data changes at each time point, and determine the change trend of the data at the different time points to obtain the data trend at the key time points;

[0127] S502: Compare and analyze data trends at key time points to determine the commonalities and differences in data changes at time points, combine new clinical research results, evaluate the impact of changes on patient risk levels, and generate risk impact analysis records;

[0128] S503: Based on the risk impact analysis records, the risk level is refined through the random forest algorithm, and the existing risk scoring model is optimized according to the information in the clinical research results to establish a risk level judgment model.

[0129] S501: Based on the risk development trend information, a long short-term memory (LSTM) network model is used to select data from multiple key time points, including the time points of patient biomarker detection. The time points are looped through using Python's TensorFlow and Keras libraries to analyze the data changes at each time point. The Sequential model construction process of the LSTM network is used, which includes configuring the units parameter of the LSTM layer to determine the number of memory units of the network, and selecting relu as the activation function to enhance the nonlinear processing capability. Finally, the training process is configured through the compile() method, in which the optimizer selects Adam and the loss function uses mean_squared_error to determine the change trend of the data at different time points and obtain the data trend at key time points.

[0130] The long short-term memory network is based on the formula:

[0131]

[0132] Calculate the changing trend of data at different time points to obtain the data trend at key time points;

[0133] in, For time point The hidden state represents the output state of the model at a specific time point. For time point The input represents the data observed at that point in time. is the weight matrix input to the hidden layer, which determines how the input data affects the hidden state. is the weight matrix from hidden layer to hidden layer, which controls how the hidden state at the previous time point affects the hidden state at the current time point. is the bias term of the hidden layer, which is used to add additional adjustments to optimize the model output. is the activation function, which enhances the model's ability to handle nonlinear problems. is the time window size, which adjusts the time series length of the model input to better capture the temporal characteristics of the data. is a dynamically adjusted learning rate that is automatically adjusted according to the model training progress to optimize the training process. Represents the number of adaptive memory units, which dynamically adjusts the number of units in the LSTM layer according to the complexity of the data to improve the learning efficiency and accuracy of the model. It is the application of the activation function, which represents the selection of an optimized nonlinear function to improve the nonlinear fitting ability of the model.

[0134] Execution process:

[0135] S1: According to the time window size ( ) Split the time series data and prepare the sequence for model input;

[0136] S2: Build the LSTM model and set the number of adaptive memory units ( ), applying advanced activation functions ( ) to enhance nonlinear processing capabilities, configure the Adam optimizer, and set a dynamic learning rate ( );

[0137] S3: Train the model and adjust the learning rate according to performance ( ), monitor the training progress and adjust the number of memory cells if necessary ( );

[0138] S4: Evaluate model performance and apply it to trend prediction on new data.

[0139] S502: Compare and analyze the data trends at key time points, use vectorization methods to compare the changes in data at each time point, use the Numpy library to calculate the data differences between time points, combine new clinical research results, and use the PCA (principal component analysis) method in the scikit-learn library to evaluate the impact of data changes on patient risk levels. PCA analysis helps identify the main components in data changes, and transforms the data through the fit_transform() method to generate risk impact analysis records.

[0140] S503: Based on the risk impact analysis records, the risk level is refined and the prediction model is optimized through the random forest algorithm. The RandomForestClassifier class in the scikit-learn library of Python is used. The n_estimators parameter is configured to define the number of decision trees. The max_depth parameter limits the maximum depth of the tree to prevent overfitting. The model is trained through the fit() method, and the predict() method is used to predict the patient's risk level based on the information in the clinical research results. Finally, a risk level judgment model is established.

[0141] See also Figure 7 Based on the risk level judgment model, analyze the patient's health status and living background, formulate targeted measures, including adjustments to lifestyle habits and proposals for medical measures, and generate risk prevention proposals as follows:

[0142] S601: Based on the risk level judgment model, analyze the patient's historical health data and life background, including eating habits, exercise frequency and historical medical history, identify adjustable risk factors, and generate an adjustable risk factor list;

[0143] S602: Based on the list of adjustable risk factors, formulate lifestyle adjustment measures and medical intervention proposals, including increasing physical activity and adjusting diet structure, matching corresponding measures for each risk factor, and obtaining targeted intervention measures;

[0144] S603: Based on targeted intervention measures, combined with the patient's acceptance capacity and situation, optimize and adjust the intervention measures, verify the feasibility and pertinence of the measures, and generate risk prevention proposals.

[0145] S601: Based on the risk level judgment model, data mining technology is used to use Python's Pandas library to analyze the patient's historical health data and life background, including eating habits, exercise frequency and historical medical history. By applying the association rule learning algorithm in data mining, such as the Apriori algorithm, the correlation between adjustable risk factors is identified. The Apriori algorithm is executed to identify frequent item sets and strong association rules by setting the minimum support and confidence parameters, and a list of adjustable risk factors is generated.

[0146] S602: Based on the list of adjustable risk factors, formulate lifestyle adjustment measures and medical intervention proposals, use decision tree analysis methods, and use the DecisionTreeClassifier class in Python's scikit-learn library to match corresponding measures for each risk factor. Use the fit() method to build a decision tree model, which splits nodes based on the attributes of risk factors, determines the optimal intervention measures, and obtains targeted intervention measures.

[0147] S603: Based on targeted intervention measures, combined with the patient's acceptance capacity and situation, optimization algorithms, such as genetic algorithms, are used to optimize and adjust the intervention measures. The genetic algorithm is implemented using Python's DEAP library. The intervention measures are represented by defining individual encoding methods, and the fitness function is set to evaluate the effectiveness and targeting of the measures. Selection, crossover and mutation operations are performed to search for the optimal solution, the feasibility and targeting of the measures are verified, and risk prevention proposals are generated.

[0148] See also Figure 8 , track and evaluate the effectiveness of the proposed risk prevention measures, record the changing trend of the patient's health status after the measures are implemented, and evaluate the expected effects of the preventive measures. The specific steps to obtain the prevention effect evaluation record are:

[0149] S701: Follow up and evaluate the effectiveness of proposed risk prevention measures, collect data on changes in health status, including biomarkers and changes in daily living habits, record data after implementing measures, and generate measures implementation tracking records;

[0150] S702: Based on the data in the tracking record of the implementation of the measure, analyze the changes in the health data before and after the measure, identify the positive changes or the direction that needs to be adjusted, and obtain the health status change analysis results;

[0151] S703: Based on the analysis results of health status changes, the effects of preventive measures are evaluated, and the effectiveness and improvement points of the measures are determined to obtain a preventive effect evaluation record.

[0152] S701: Track and evaluate the effectiveness of proposed risk prevention measures, using data collection and analysis techniques, using Python's Pandas library to collect health status change data, including biomarkers and changes in daily living habits, and record the data after the implementation of measures through the construction and data processing functions of DataFrame. Apply time series analysis methods, such as the rolling() function to perform moving average to smooth the data and generate a tracking record of the implementation of measures.

[0153] S702: Based on the data in the implementation tracking records of the measures, statistical analysis techniques, including t-test and ANOVA analysis in Python's scikit-learn library, are used to analyze the changes in health data before and after the measures. The t-test is used to compare the mean differences of data at two time points, while the ANOVA analysis is suitable for comparing data at multiple time points, identifying positive changes or directions that need to be adjusted, and generating health status change analysis results.

[0154] S703: Based on the results of the health status change analysis, an effect evaluation model, such as logistic regression analysis, is used. It is implemented using Python's statsmodels library. The logistic regression model is used to evaluate the deviation between the effect of the preventive measures and the expected goal. The configuration model includes independent variables such as health status change data and control variables such as patient acceptance. The dependent variable is the binary result of the preventive effect (effective / ineffective). The model parameters are estimated through the fit() method, and the evaluation results are extracted using the summary() method to obtain the preventive effect evaluation record.

[0155] See also Fig. 9 , Cardiovascular and cerebrovascular disease risk prediction system, the system includes:

[0156] The risk factor identification module identifies cardiovascular and cerebrovascular risk factors based on existing clinical research and medical guidelines, including age, gender, genetic factors and lifestyle habits, removes data duplication, verifies information accuracy, and supplements missing risk factor descriptions to generate a risk factor library;

[0157] The weight adjustment module performs preliminary weight assignment of risk factors based on the risk factor library. Based on the initial weight model, it collects clinical research results, periodically evaluates and adjusts the existing risk factor weights, and constructs a weight adjustment model by comparing the impact of new and old research data on risk factor weights.

[0158] The individual risk scoring module is based on a weight adjustment model, uses recursive feature elimination and adaptive reinforcement learning algorithms, traverses the patient's historical health data, performs weighted calculations on each risk factor, summarizes the scores, divides the risk levels, and obtains individual risk score data;

[0159] The risk trend analysis module collects time series data, including biomarkers, based on individual risk score data, selects key time points for analysis, identifies risk change trends, and constructs trend change analysis;

[0160] The risk level judgment module is based on trend change analysis, using long-short term memory networks, selecting data from multiple key time points, conducting comparative analysis, and combining clinical research results to optimize the existing risk scoring model and generate a risk level model;

[0161] The risk prevention and assessment module analyzes the patient's health status and living background based on the risk level model, formulates targeted measures, including lifestyle adjustments and medical measures proposals, and conducts tracking and effectiveness evaluation, records the changing trends of the patient's health status after the implementation of the measures, and establishes prevention effect evaluation records.

[0162] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for predicting cardiovascular and cerebrovascular disease risk, characterized in that: The following steps are involved: Based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed to obtain a risk factor database; Based on the risk factor database, perform regular data analysis and update the weights of risk factors, including evaluating the impact of clinical research on existing models, and adjusting the weights according to the evaluation results to generate a weight adjustment model; Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to perform weighted calculations on the patient's risk factors, and the risk score of each patient is calculated based on the patient's historical health data to obtain individual risk score data; Based on the individual risk score data, collect the time series data of the patient's biomarkers, select key time points for analysis, identify potential risk change trends, and obtain risk development trend information; Based on the risk development trend information, a long short-term memory network is used to conduct comparative analysis of data at multiple time points, and combined with clinical research results, the patient's risk level is determined, and the prediction model is optimized to establish a risk level judgment model; Based on the risk level judgment model, analyze the patient's health status and living background, formulate targeted measures, including adjustments to living habits and proposals for medical measures, and generate risk prevention proposals; Track and evaluate the effectiveness of the proposed risk prevention measures, record the changing trend of the patient's health status after the measures are implemented, and evaluate the expected effects of the preventive measures to obtain a prevention effect evaluation record; Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to perform weighted calculations on the patient's risk factors, and the risk score of each patient is calculated based on the patient's historical health data. The specific steps for obtaining individual risk score data are as follows: Based on the weight adjustment model, recursive feature elimination and adaptive reinforcement learning are used to traverse the patient's historical health data and perform weighted calculations on each risk factor. If risk factor information is missing in the patient data, the default value in the model is used to fill in the information and obtain a weighted risk factor score. The recursive feature elimination is performed according to the formula: ; Calculate feature weights and generate feature importance rankings; in, Indicates The weight of the feature, It is The feature in The value of the iteration, is the coefficient of the feature in the model, is the historical health status classification, is the lifestyle stability index, Represents the total number of features in the feature set; The adaptive enhancement algorithm is based on the formula: ; Optimize weight adjustment to obtain weighted risk factor scores; in, It is Round The weight of the samples, It is Round The weight of the samples, It is The weight of the round classifier, is the true label of the sample, It is The prediction result of the wheel classifier, is the environmental adaptation coefficient, is the access to medical resources, is the normalization factor, exp represents the exponential power of the natural exponential function e; Summarizing the weighted risk factor scores, applying a gradient boosting tree, calculating a total score by accumulating the scores of each risk factor, and performing risk level division according to the total score to obtain a risk level allocation result; According to the risk level allocation result, a risk score is allocated to each patient to reflect the patient's risk of cardiovascular and cerebrovascular diseases, and individual risk score data is obtained.

2. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: The risk factor database includes the patient's age, gender, genetic factors, living habits, historical medical history and recent medical interventions; the weight adjustment model includes the disease-associated risk factor weights, update cycle and weight adjustment mechanism; the risk development trend information includes time series analysis results, changing trends of key risk indicators, and early warning signals of potential risk events; the risk level judgment model includes risk level classification standards, prediction model optimization strategies, and risk level assessment methods; the prevention effect evaluation records include comparative analysis of patient health indicators before and after the implementation of measures, and long-term follow-up monitoring information.

3. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: Based on existing clinical research and medical guidelines, text mining and key information extraction are performed to identify the risk factors of cardiovascular and cerebrovascular diseases, and a relationship mapping table is constructed. The specific steps to obtain the risk factor database are as follows: Based on existing clinical research and medical guidelines, we extracted information on risk factors associated with cardiovascular and cerebrovascular diseases, and preliminarily sorted out the information to obtain a preliminary list of risk factors; Based on the preliminary risk factor list, data cleaning is performed, including removing duplicates, verifying the accuracy of information, and supplementing missing risk factor descriptions to obtain an adjusted risk factor list; According to the adjusted risk factor list, a mapping table of the correlation between risk factors and cardiovascular and cerebrovascular diseases is constructed, and a risk factor database is obtained by evaluating the strength of the correlation research information of the risk factors.

4. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: Based on the risk factor database, regular data analysis is performed to update the weights of risk factors, including evaluating the impact of clinical research on the existing model and adjusting the weights according to the evaluation results. The specific steps for generating the weight adjustment model are: Based on the risk factor database, perform preliminary weight assignment of risk factors, perform initial weight allocation according to mapping strength information of the relationship between risk factors and cardiovascular and cerebrovascular diseases, and generate an initial weight model; Based on the initial weight model, periodic evaluation of weights is performed, including collecting clinical research results, adjusting existing risk factor weights to reflect the criticality of new medical information, and obtaining a weight update model; The weight updating model is evaluated and the risk factor weights are determined. By comparing the impact of new and old research data on the risk factor weights, a weight adjustment model is generated.

5. The method for predicting cardiovascular and cerebrovascular disease risks according to claim 1, characterized in that: Based on the individual risk score data, the time series data of patient biomarkers are collected, key time points are selected for analysis, potential risk change trends are identified, and the steps for obtaining risk development trend information are as follows: Based on the individual risk score data, collecting time series data of biomarkers of the patient, including blood pressure, blood sugar and cholesterol indicators, and verifying whether each data is recorded in time sequence to generate biomarker data records; Analyze the biomarker data records, select key time points, including before and after disease diagnosis, treatment start and end time nodes, analyze the data at the time points, identify data changes during the differential health state transition period, and obtain key time point analysis results; Based on the analysis results of the key time points, the changing trends in the time series are identified, including continuous rising, falling or fluctuating patterns, the development trend of the patient's cardiovascular and cerebrovascular disease risk is determined, and risk development trend information is obtained.

6. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: Based on the risk development trend information, a long short-term memory network is used to conduct comparative analysis of data at multiple time points, and combined with clinical research results, the patient's risk level is determined, and the prediction model is optimized. The specific steps for establishing a risk level judgment model are as follows: Based on the risk development trend information, using a long short-term memory network, select data at multiple key time points, including the time points of patient biomarker detection, loop through the time points, analyze the data changes at each time point, and determine the change trend of the data at the different time points to obtain the data trend at the key time points; The long short-term memory network is based on the formula: ; Calculate the changing trend of data at different time points to obtain the data trend at key time points; in, For time point The hidden state of For time point Input, is the weight matrix input to the hidden layer, is the weight matrix from hidden layer to hidden layer, is the bias term of the hidden layer, is the activation function, is the time window size, is the dynamically adjusted learning rate, Represents the number of memory cells, is the application of the activation function; Compare and analyze the data trends at the key time points to determine the commonalities and differences in the data changes at the time points, combine the new clinical research results, evaluate the impact of the changes on the patient's risk level, and generate a risk impact analysis record; Based on the risk impact analysis records, the risk level is refined through the random forest algorithm, and according to the information in the clinical research results, the existing risk scoring model is optimized to establish a risk level judgment model.

7. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: Based on the risk level judgment model, the patient's health status and living background are analyzed, and targeted measures are formulated, including adjustments to living habits and proposals for medical measures. The specific steps for generating risk prevention proposal measures are as follows: Based on the risk level judgment model, analyzing the patient's historical health data and life background, including eating habits, exercise frequency, and historical medical history, identifying adjustable risk factors, and generating an adjustable risk factor list; Based on the list of adjustable risk factors, formulate lifestyle adjustment measures and medical intervention proposals, including increasing physical activity and adjusting diet structure, matching corresponding measures for each risk factor, and obtaining targeted intervention measures; Based on the targeted intervention measures, combined with the patient's acceptance capacity and situation, the intervention measures are optimized and adjusted, and the feasibility and targetedness of the measures are verified to generate risk prevention proposals.

8. The method for predicting cardiovascular and cerebrovascular disease risk according to claim 1, characterized in that: Track and evaluate the effectiveness of the proposed risk prevention measures, record the changing trend of the patient's health status after the measures are implemented, and evaluate the expected effects of the preventive measures. The specific steps for obtaining the prevention effect evaluation record are: Track and evaluate the effectiveness of the proposed risk prevention measures, collect data on changes in health status, including biomarkers and changes in daily living habits, record data after the implementation of the measures, and generate a tracking record of the implementation of the measures; Based on the data in the tracking record of the implementation of the measures, analyze the changes in health data before and after the measures, identify positive changes or directions that need to be adjusted, and obtain health status change analysis results; Based on the health status change analysis results, the effects of preventive measures are evaluated, and the effectiveness and improvement points of the measures are judged to obtain a preventive effect evaluation record.

9. Cardiovascular and cerebrovascular disease risk prediction system, characterized in that: According to any one of claims 1 to 8, the cardiovascular and cerebrovascular disease risk prediction method is implemented, and the system comprises: The risk factor identification module identifies cardiovascular and cerebrovascular risk factors based on existing clinical research and medical guidelines, including age, gender, genetic factors and lifestyle habits, removes data duplication, verifies information accuracy, and supplements missing risk factor descriptions to generate a risk factor library; The weight adjustment module performs preliminary weight assignment of risk factors based on the risk factor library, collects clinical research results according to the initial weight model, periodically evaluates and adjusts the existing risk factor weights, and constructs a weight adjustment model by comparing the impact of new and old research data on risk factor weights; The individual risk scoring module uses recursive feature elimination and adaptive reinforcement learning algorithms based on the weight adjustment model to traverse the patient's historical health data, perform weighted calculations on each risk factor, summarize the scores, and divide the risk levels to obtain individual risk score data; The risk trend analysis module collects time series data, including biomarkers, based on the individual risk score data, selects key time points for analysis, identifies risk change trends, and constructs trend change analysis; The risk level judgment module is based on the trend change analysis, uses the long short-term memory network, selects data from multiple key time points, conducts comparative analysis, and optimizes the existing risk scoring model in combination with clinical research results to generate a risk level model; The risk prevention and assessment module analyzes the patient's health status and living background based on the risk level model, formulates targeted measures, including lifestyle adjustments and medical measures, and conducts tracking and effect evaluation, records the changing trends of the patient's health status after the implementation of the measures, and establishes prevention effect evaluation records.

Citation Information

Patent Citations

  • Grain crop disease symptom effective feature recognition method based on LSELMRFE

    CN112232387A

  • Diabetes risk early warning method based on big data analysis

    CN117253614A

  • Intelligent medical equipment integrated health diagnosis system

    CN118230936A

  • MEDICAL DECISION SUPPORT SYSTEM

    EA202390296A1

  • System and methods for detect autonomic dysreflexia detection using machine learning classification

    US20240164697A1

Cited By

  • Plateau environment cerebrovascular disease risk prediction system and method based on digital twinning

    CN120496819A

  • Platelet aggregation inhibitor

    CN120496819B