Medical decision support method and system based on big data

Abstract

The invention provides a medical decision support method and system based on big data. The method comprises the following steps: integrating and fusing patient information from different medical institutions, personal health monitoring equipment and a public health database to obtain a dynamic update library; a graph neural network algorithm is applied to analyze the relation between the patients and the relation between the patients and the environmental factors in the dynamic update library so as to predict the individualized disease development trend, and dynamic risk early warning signals are generated; based on the dynamic risk early warning signal, through feature importance analysis and a local interpretable model, revealing key factors and an interaction mechanism influencing disease progression, and applying a health economics evaluation method to generate a medical suggestion report; and in combination with the medical suggestion report, establishing a dynamic learning and feedback mechanism, optimizing personalized suggestions in the medical suggestion report, and constructing a medical decision support system. According to the technical scheme provided by the invention, the credibility of medical decision making is remarkably improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of big data decision technology, and in particular, to a medical decision support method and system based on big data. Background Art

[0002] In the modern medical environment, it is increasingly important to integrate patient information from different medical institutions, personal health monitoring devices, and public health databases. With the rapid growth of medical data, a technology that can integrate and fuse these multi-source heterogeneous data is needed to create a dynamically updated library containing the patient's health management records throughout the life cycle. In addition, in order to provide personalized medical services, the system must be able to analyze newly acquired patient information in real time and generate dynamic risk warning signals to support early intervention and prevention.

[0003] Currently, many medical information systems can already collect and store large amounts of patient data. However, most systems are limited to static data analysis and lack effective modeling of complex interactions between patient relationships and environmental factors. Although some platforms attempt to apply machine learning algorithms for disease prediction, they often ignore the importance of features and model interpretability, making it difficult for doctors to understand the reasons behind the prediction results. At the same time, when formulating treatment plans, few systems can fully consider cost-effectiveness and personalized needs, which limits their practical application value.

[0004] The main defects of existing medical decision support systems are: insufficient integration capabilities for multi-source data, unable to form a comprehensive view of patient health; lack of effective model interpretation mechanism, making it difficult for prediction results to gain clinical trust; and the choice of treatment options usually fails to fully consider cost-effectiveness and personalized needs, thus affecting the effective allocation of resources and patient treatment outcomes. Summary of the invention

[0005] The embodiments of the present application provide a medical decision support method and system based on big data to solve the problems in the prior art that affect the effective allocation of resources and the treatment effect of patients.

[0006] In a first aspect, the present application provides a medical decision support method based on big data, including:

[0007] Integrate and merge patient information from different medical institutions, personal health monitoring devices, and public health databases to obtain a dynamic update library containing health management records of patients throughout their life cycle;

[0008] Applying a graph neural network algorithm to analyze the relationships between patients and between patients and environmental factors in the dynamically updated library to predict individualized disease development trends, and combining real-time streaming data analysis technology to instantly evaluate newly acquired patient information to generate dynamic risk warning signals;

[0009] Based on the dynamic risk warning signal, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm. Through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed. The health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library, and a medical advice report is generated in combination with the individual needs of patients and socioeconomic impacts.

[0010] In combination with the medical advice report, a dynamic learning and feedback mechanism is established to optimize the personalized suggestions in the medical advice report and build a medical decision support system.

[0011] Optionally, the application of a graph neural network algorithm analyzes the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and combines real-time streaming data analysis technology to instantly evaluate newly acquired patient information to generate dynamic risk warning signals, including:

[0012] Applying a graph neural network algorithm to model the relationships between patients and between patients and environmental factors in the dynamically updated library, to obtain simulation results of disease progression patterns;

[0013] Based on the simulation results of the disease progression pattern, real-time streaming data analysis technology is applied to instantly evaluate and process newly acquired patient information and generate an instant evaluation report;

[0014] Utilizing the instant assessment report and combining it with the simulation results of the disease progression pattern, the patient's future health risk is quantified to obtain a potential health risk score;

[0015] According to the potential health risk score, patients at different risk levels are classified through comprehensive evaluation and setting risk thresholds to generate dynamic risk warning signals.

[0016] Optionally, the simulation result based on the disease progression pattern uses real-time streaming data analysis technology to perform instant evaluation processing on newly acquired patient information and generate an instant evaluation report, including:

[0017] Based on the simulation result of the disease progression pattern, preprocessing the newly acquired patient information to obtain preprocessed patient information;

[0018] Applying real-time streaming data analysis technology to instantly analyze and process the pre-processed patient information, identify abnormalities or key change points related to disease progression, and obtain instant analysis results;

[0019] Based on the instant analysis results and the simulation results of the disease progression model, a comprehensive evaluation is performed on the patient's current health status to obtain an instant health status evaluation;

[0020] Based on the instant health status assessment, an instant assessment report is generated, and the instant assessment report includes the patient's current health status, risk factors and possible development trends.

[0021] Optionally, the instant assessment report is used in combination with the simulation result of the disease progression pattern to quantify the patient's future health risk to obtain a potential health risk score, including:

[0022] Utilizing the instant assessment report, the newly acquired patient information is analyzed and processed in detail to identify key indicators and abnormal changes in the current health status and generate a current health status assessment;

[0023] According to the current health status assessment, combined with the simulation results of the disease progression pattern, the patient's future health trend is predicted and processed to generate a future health trend prediction;

[0024] Based on the future health trend prediction, the patient's long-term health risk factors are quantitatively analyzed and processed using statistical models and machine learning algorithms to obtain a preliminary health risk score;

[0025] Using the preliminary health risk score, combined with the individual characteristics of the patient, to perform personalized adjustment processing to generate a personalized health risk score;

[0026] According to the personalized health risk score, a risk assessment system is introduced to set different risk thresholds, and patients at different risk levels are classified and processed to obtain a potential health risk score.

[0027] Optionally, based on the dynamic risk warning signal, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm, and through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed, and the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and combined with the individual needs of patients and socioeconomic impacts, a medical advice report is generated, including:

[0028] Based on the dynamic risk warning signal, the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm are deeply analyzed and processed to obtain preliminary analysis results;

[0029] Using the preliminary analysis results, feature importance analysis in the explanatory artificial intelligence algorithm is introduced to reveal the key factors affecting disease progression and generate a list of key factors;

[0030] According to the list of key factors, a local interpretable model is applied to analyze the causal relationship and interaction mechanism of the key factors, and a detailed analysis report is generated;

[0031] Based on the detailed analysis report, applying health economics evaluation methods, quantifying the cost-effectiveness ratio of different treatment plans integrated from the dynamically updated library, and generating a cost-effectiveness analysis report;

[0032] Combined with the cost-effectiveness analysis report, taking into account the individual needs of patients and socioeconomic impacts, personalized adjustments are made to generate personalized medical recommendations;

[0033] Based on the personalized medical advice, the detailed analysis report and the cost-effectiveness analysis report are combined to generate a medical advice report.

[0034] Optionally, based on the detailed analysis report, a health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library to generate a cost-effectiveness analysis report, including:

[0035] Based on the detailed analysis report, applying health economics evaluation methods, quantifying the cost-effectiveness ratio of different treatment plans integrated from the dynamically updated library, and generating a preliminary cost-effectiveness evaluation;

[0036] Using the preliminary cost-effectiveness assessment, combined with the key factors and their interaction mechanisms in the detailed analytical report, the costs and expected effects of each treatment option are refined to generate a refined cost-effectiveness assessment;

[0037] According to the detailed cost-effectiveness evaluation, socioeconomic factors are introduced to comprehensively adjust different treatment plans and generate a comprehensive cost-effectiveness evaluation;

[0038] Based on the comprehensive cost-effectiveness evaluation, cost-effectiveness thresholds of different treatment options are set, and each option is classified to obtain a cost-effectiveness classification result;

[0039] In combination with the cost-effectiveness classification results, the individual needs of patients are taken into consideration, and optimization recommendations are made on the selection of treatment options and resource allocation to generate a cost-effectiveness analysis report.

[0040] Optionally, the medical advice report is combined to establish a dynamic learning and feedback mechanism, optimize the personalized advice in the medical advice report, and build a medical decision support system, including:

[0041] Using the medical advice report, data feedback on the treatment plan adopted by the doctor and the actual implementation effect is collected to generate a feedback data set;

[0042] According to the feedback data set, combined with the individual needs and socioeconomic impact of the patient, the personalized recommendations in the medical recommendation report are evaluated to generate an evaluation result;

[0043] Based on the evaluation results, applying a machine learning algorithm to optimize the personalized recommendations in the medical recommendation report to generate optimized personalized recommendations;

[0044] Using the optimized personalized suggestions, updating the knowledge base of the medical decision support system to generate an updated knowledge base;

[0045] According to the updated knowledge base, adjusting the recommendation strategy of the system so that the recommendation strategy can be automatically updated according to new data and feedback to generate an updated recommendation strategy;

[0046] In combination with the updated recommendation strategy, a dynamic learning and feedback mechanism is established to improve the personalized recommendations in the medical advice report and build a medical decision support system.

[0047] In a second aspect, the present application embodiment provides a medical decision support system based on big data, including:

[0048] An integration module is used to integrate and fuse patient information from different medical institutions, personal health monitoring devices, and public health databases to obtain a dynamic update library containing health management records of patients throughout their life cycle;

[0049] A prediction module, which is used to apply a graph neural network algorithm to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and to instantly evaluate newly acquired patient information in combination with real-time streaming data analysis technology to generate dynamic risk warning signals;

[0050] A quantification module is used to introduce an explanatory artificial intelligence algorithm to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm based on the dynamic risk warning signal, reveal the key factors and interaction mechanisms that affect disease progression through feature importance analysis and local interpretable models in the artificial intelligence algorithm, apply health economics evaluation methods to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and generate a medical advice report in combination with the individual needs of patients and socioeconomic impacts;

[0051] The optimization module is used to combine the medical advice report, establish a dynamic learning and feedback mechanism, optimize the personalized suggestions in the medical advice report, and build a medical decision support system.

[0052] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a medical decision support method based on big data as described in the first aspect.

[0053] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a medical decision support method based on big data as described in the first aspect.

[0054] In an embodiment of the present application, patient information from different medical institutions, personal health monitoring devices, and public health databases is integrated and fused to obtain a dynamic update library, which contains health management records of patients throughout their life cycle; a graph neural network algorithm is used to analyze the relationship between patients and between patients and environmental factors in the dynamic update library to predict the development trend of individualized diseases, and real-time streaming data analysis technology is combined to instantly evaluate newly acquired patient information to generate dynamic risk warning signals; based on the dynamic risk warning signals, an explanatory artificial intelligence algorithm is introduced to deeply analyze the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm, and through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed, and the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and a medical advice report is generated in combination with the individualized needs of patients and socioeconomic impacts; combined with the medical advice report, a dynamic learning and feedback mechanism is established to optimize the personalized advice in the medical advice report and build a medical decision support system.

[0055] The technical solution of this application has the following beneficial effects:

[0056] By integrating and fusing patient information from different sources, including data from medical institutions, personal health monitoring devices, and public health databases, dynamic updates of patient health management records throughout their life cycle are achieved. This enables personalized disease development trend predictions for each patient, thereby optimizing treatment plans; the application of graph neural network algorithms combined with real-time streaming data analysis technology can instantly evaluate newly acquired patient information and generate dynamic risk warning signals. This approach helps to detect potential health problems early, take preventive measures in time, and reduce the possibility of disease deterioration; introduce explanatory artificial intelligence algorithms to parse model prediction results and reveal key factors affecting disease progression and their interaction mechanisms. This not only increases the transparency of model output results, but also increases the trust of doctors and other medical professionals in recommended plans; health economics evaluation methods are used to quantify the cost-effectiveness ratio of different treatment plans, combined with patients' individual needs and socioeconomic impacts, to ensure the effective allocation of resources. This approach can help medical institutions achieve the best health benefits within a limited budget; establish a dynamic learning and feedback mechanism, and continuously optimize personalized recommendations in medical advice reports by collecting data feedback on treatment plans adopted by doctors and actual implementation effects. This closed-loop system promotes the accumulation of knowledge and the advancement of technology, and improves the overall quality of medical services. When generating medical advice reports, in addition to considering the optimal medical solution, it also takes into account the socioeconomic impact. This helps to formulate treatment plans that are more in line with reality and improve patient satisfaction and treatment compliance.

[0057] Furthermore, by applying the graph neural network algorithm to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library, this method can accurately predict the development trend of individualized diseases, and combine real-time streaming data analysis technology to instantly evaluate newly acquired patient information and generate dynamic risk warning signals. This method not only simulates the disease progression pattern, but also responds quickly to the latest changes in health data and provides instant assessment reports. Furthermore, by quantifying future health risks and calculating potential health risk scores, the system can effectively identify high-risk patients, classify and manage patients at different risk levels by setting risk thresholds, and take intervention measures in advance. Ultimately, this method improves the accuracy and timeliness of early warnings, enhances support for personalized medical decision-making, helps optimize resource allocation, improves overall medical service quality and efficiency, and reduces patients' health risks and medical costs.

[0058] Furthermore, by introducing an explanatory artificial intelligence algorithm based on dynamic risk warning signals, the disease progression probability and risk assessment results generated by the graph neural network are deeply analyzed. This method can not only reveal the key factors affecting disease progression and their interaction mechanisms, but also combine health economics evaluation to quantify the cost-effectiveness ratio of different treatment options, and consider the individual needs and socioeconomic impacts of patients, thereby generating a detailed medical advice report. This method significantly enhances the transparency and interpretability of the model prediction results, making the medical decision-making process more scientific and reasonable; at the same time, through the application of feature importance analysis and local interpretable models, key influencing factors can be accurately located, providing strong support for personalized treatment. In addition, the addition of cost-effectiveness analysis helps to optimize resource allocation, ensure the effective use of social resources while meeting the individual needs of patients, and ultimately improve the overall quality and efficiency of medical services, reduce long-term medical costs, and promote improvements in medical equity and accessibility.

[0059] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 A flowchart of a medical decision support method based on big data provided in an embodiment of the present application;

[0062] Figure 2 A schematic diagram of the structure of a medical decision support system based on big data provided in an embodiment of the present application;

[0063] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0065] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0067] Figure 1 A flowchart of a medical decision support method based on big data is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0068] 101. Integrate and merge patient information from different medical institutions, personal health monitoring devices, and public health databases to obtain a dynamic update library, which contains health management records of the patient throughout his or her life cycle;

[0069] In this step, the patient information from different medical institutions, personal health monitoring devices and public health databases is integrated and fused to obtain a dynamic update library. The dynamic update library contains the health management records of the patient's entire life cycle, including but not limited to medical records, laboratory test results, imaging data, daily health monitoring data (such as blood pressure, heart rate, etc.) and public health statistics, which are used to build a comprehensive and real-time updated patient health record.

[0070] In the embodiments of the present application, firstly, the consistency and accuracy of multi-source data are ensured through pre-processing methods such as data cleaning, standardization and feature engineering; then, the scattered data are integrated into a unified platform using data integration technology to form a dynamic update library, and an incremental update mechanism is used to ensure the real-time and integrity of the data, thereby providing a solid data foundation for subsequent analysis.

[0071] Suppose a hospital wants to build a comprehensive patient health management platform that integrates the in-hospital electronic medical record system, data interfaces of external cooperative clinics, user health monitoring data uploaded by smart wearable devices, and epidemiological data in the national public health database. Through this platform, doctors can obtain the complete health history of each patient, from birth to all medical records to better understand and manage the patient's health status.

[0072] 102. Applying a graph neural network algorithm to analyze the relationships between patients and between patients and environmental factors in the dynamically updated library to predict individualized disease development trends, and combining real-time streaming data analysis technology to instantly evaluate newly acquired patient information and generate dynamic risk warning signals;

[0073] In this step, the graph neural network (GNN) algorithm is used to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases. At the same time, combined with real-time streaming data analysis technology, newly acquired patient information is instantly evaluated to generate dynamic risk warning signals. This step aims to use complex network structures to capture the correlation between patients and the influence of environmental factors, and provide more accurate personalized disease predictions.

[0074] In the embodiment of the present application, a weighted graph model of patients and their environmental factors is constructed based on a dynamically updated library, and a GNN algorithm is used to learn node representations and predict disease progression patterns. At the same time, real-time streaming data analysis technology continuously monitors new input data streams, and once an abnormality or key change point is detected, an immediate assessment process is immediately triggered, and a dynamic risk warning signal is output so that preventive measures can be taken in a timely manner.

[0075] Suppose a diabetic patient wears a continuous glucose monitor, whose blood sugar level and other physiological parameters are transmitted to a central server in real time. When the GNN model detects that the patient's blood sugar fluctuations are similar to those of other high-risk people, the system will immediately issue an alarm to remind medical staff to pay attention to the patient's condition changes and recommend further examination or adjustment of treatment plans.

[0076] 103. Based on the dynamic risk warning signal, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm. Through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed. The health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and a medical advice report is generated in combination with the individual needs of patients and socioeconomic impacts.

[0077] In this step, based on dynamic risk warning signals, an explanatory artificial intelligence (XAI) algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm. Through feature importance analysis and local interpretable models (such as LIME or SHAP), the key factors and interaction mechanisms affecting disease progression are revealed. In addition, health economic evaluation methods are applied to quantify the cost-effectiveness ratio of different treatment options, and a medical advice report is generated based on the individual needs of patients and socioeconomic impacts. This process enhances the transparency and credibility of the model's prediction results, helping doctors make more informed decisions.

[0078] In the embodiment of this application, after receiving the dynamic risk warning, the XAI algorithm analyzes the output of the GNN model in detail, identifies which features are most important for disease prediction, and explains how these features work together to promote the development of the disease. Subsequently, the optimal treatment path is determined through health economics evaluation, taking into account cost-effectiveness and personalized needs, and finally generating a detailed medical advice report to guide clinical practice.

[0079] Assuming that a heart disease patient is predicted to have a higher risk of recurrence, the XAI algorithm analyzes and finds that smoking history and family inheritance are the main risk factors. Based on this analysis result, the system recommends smoking cessation counseling and support services, and suggests a drug treatment plan that is both in line with the patient's financial ability and can significantly reduce the risk of recurrence in the short term.

[0080] 104. In combination with the medical advice report, a dynamic learning and feedback mechanism is established to optimize the personalized suggestions in the medical advice report and build a medical decision support system.

[0081] In this step, a dynamic learning and feedback mechanism is established in combination with the medical advice report to optimize the personalized advice in the medical advice report and build a medical decision support system. This mechanism allows the system to continuously improve itself based on the actual treatment plan adopted and its effect, ensuring that the advice provided is always in the best state. In this way, not only the quality of medical services is improved, but also the continuous advancement of knowledge and technology is promoted.

[0082] In the embodiment of the present application, the system collects data feedback on the treatment plans adopted by doctors and the actual implementation effects to evaluate the effectiveness of personalized recommendations. Based on these feedbacks, machine learning algorithms are applied to adjust model parameters and optimize future personalized recommendations. In addition, the updated knowledge base and recommendation strategies enable the system to automatically adapt to new data and feedback, thereby maintaining long-term accuracy and reliability.

[0083] Suppose a medical institution has used the above medical decision support system for a period of time and found that certain types of cancer treatments have achieved better results in practice. The system automatically updates the relevant model through a feedback mechanism, and the next time it encounters a similar case, it can more confidently recommend these proven and effective treatment plans, thereby improving the overall treatment success rate.

[0084] In summary, steps 101 to 104 cover the entire process from data integration to the generation of personalized medical advice, aiming to provide a comprehensive, dynamic and personalized medical decision support system to meet the needs of efficiency, accuracy and transparency in modern medical services.

[0085] In order to meet the needs of more accurate personalized disease prediction and real-time risk assessment, in some embodiments, the graph neural network algorithm is applied in step 102 to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and the newly acquired patient information is instantly evaluated in combination with real-time streaming data analysis technology to generate dynamic risk warning signals, including:

[0086] Applying a graph neural network algorithm, the relationships between patients and between patients and environmental factors in the dynamically updated library are modeled to obtain simulation results of disease progression patterns; based on the simulation results of the disease progression patterns, real-time streaming data analysis technology is used to instantly evaluate newly acquired patient information and generate an instant evaluation report; utilizing the instant evaluation report in combination with the simulation results of the disease progression patterns, the patient's future health risks are quantified to obtain a potential health risk score; based on the potential health risk score, patients at different risk levels are classified through comprehensive evaluation and setting of risk thresholds to generate dynamic risk warning signals.

[0087] In this embodiment, the graph neural network (GNN) algorithm is used to model the relationship between patients and environmental influences in the dynamically updated library to obtain the simulation results of the disease progression pattern. This simulation result not only covers the patient's medical record data and daily health monitoring data, but also includes environmental factors such as air quality and seasonal changes, which are used to provide a comprehensive perspective on the development of the disease.

[0088] In an embodiment of the present application, a GNN algorithm is first used to construct a complex network model that can capture the correlation between patients and environmental influences; secondly, based on the disease progression pattern simulation results output by this model, real-time streaming data analysis technology is used to quickly evaluate the newly collected data to form an instant evaluation report; again, based on the latest information provided by the instant evaluation report and combined with the simulation results of the disease progression pattern, the patient's future health risks are quantified to derive a potential health risk score; finally, risk thresholds are set through comprehensive evaluation, patients are classified according to different risk levels, and dynamic risk warning signals are generated.

[0089] Here is a specific example:

[0090] Suppose a regional health department wants to improve the early warning capabilities of chronic respiratory diseases through advanced analytical methods; to achieve this goal, they adopted the above technical solutions. First, the data uploaded by all hospitals, clinics and personal health devices in the region were integrated to establish a dynamically updated library containing detailed historical records; second, the GNN algorithm was used to analyze these data to identify which patient groups are most vulnerable to local air pollution and simulate the development trend of chronic respiratory diseases; third, in daily operations, whenever new health monitoring data flows into the system, it is immediately evaluated to quickly determine whether there are abnormal conditions or high-risk factors; finally, the potential health risk score of each patient is calculated based on the instant evaluation results, and different risk thresholds are set to automatically send personalized risk warning signals to doctors and patients. Through the above steps, the regional health department has significantly improved the early detection rate of chronic respiratory diseases and effectively reduced the deterioration of the disease.

[0091] In order to meet the need for more accurate real-time health status assessment, in some embodiments, the simulation results based on the disease progression pattern apply real-time streaming data analysis technology to perform real-time assessment processing on newly acquired patient information and generate a real-time assessment report, including:

[0092] Based on the simulation results of the disease progression pattern, newly acquired patient information is preprocessed to obtain preprocessed patient information; real-time streaming data analysis technology is applied to instantly analyze and process the preprocessed patient information, identify abnormalities or key change points related to disease progression, and obtain instant analysis results; based on the instant analysis results and in combination with the simulation results of the disease progression pattern, the patient's current health status is comprehensively evaluated and processed to obtain an instant health status evaluation; based on the instant health status evaluation, an instant evaluation report is generated, and the instant evaluation report includes the patient's current health status, risk factors, and possible development trends.

[0093] In this embodiment, the newly acquired patient information is first preprocessed to obtain the preprocessed patient information. The data involved in this step includes the latest physiological parameters, symptom descriptions, and any health-related update records to ensure the accuracy and timeliness of subsequent analysis.

[0094] In the embodiment of the present application, first, the system receives and cleans new patient data from various sources, removes noise and inconsistencies, and ensures data quality; second, real-time streaming data analysis technology is applied to conduct in-depth mining of pre-processed patient information to identify key change points or abnormalities that may affect disease progression; third, combined with the previously constructed disease progression model simulation results, the patient's current health status is comprehensively evaluated to determine his or her health status, risk factors, and future development trends; finally, based on the results of the comprehensive evaluation, a detailed instant evaluation report is generated and provided to doctors as a decision support tool.

[0095] Here is a specific example:

[0096] Suppose a community medical center wants to improve its early intervention capabilities for cardiovascular disease; to achieve this goal, they use the above technical solution. First, when a patient known to have hypertension uploads the latest heart rate and blood pressure data through a smartwatch, the system immediately preprocesses it to ensure that the data format is uniform and correct; second, the real-time streaming data analysis engine quickly analyzes the data and detects that the patient's heart rate suddenly increases beyond the normal range, accompanied by symptoms of mild chest pain; third, the system combines these new findings with the previously established cardiovascular disease progression model to assess that the patient is at increased risk of acute attack and predicts that if the condition continues to deteriorate, it may lead to a heart attack; finally, the system automatically generates an instant assessment report that details the patient's current health status (such as abnormal heart rate), potential risk factors (such as chest pain) and its future development possibilities (such as heart attack), and sends this report to the responsible doctor for reference. Through the above steps, the community medical center can take preventive measures in a timely manner, significantly improving the monitoring level and response speed of high-risk patients.

[0097] In order to meet the demand for more accurate quantification of future health risks, in some embodiments, the instant assessment report is used in combination with the simulation results of the disease progression pattern to quantify the patient's future health risks and obtain a potential health risk score, including:

[0098] Using the instant assessment report, the newly acquired patient information is analyzed in detail to identify key indicators and abnormal changes in the current health status, and generate a current health status assessment; based on the current health status assessment, combined with the simulation results of the disease progression pattern, the patient's future health trends are predicted and processed to generate a future health trend forecast; based on the future health trend forecast, statistical models and machine learning algorithms are used to quantitatively analyze the patient's long-term health risk factors to obtain a preliminary health risk score; using the preliminary health risk score, combined with the patient's individual characteristics, personalized adjustments are made to generate a personalized health risk score; based on the personalized health risk score, a risk assessment system is introduced to set different risk thresholds, and patients at different risk levels are classified to obtain a potential health risk score.

[0099] In this embodiment, the newly acquired patient information is first analyzed in detail using the data in the instant assessment report to identify key indicators of the current health status and any abnormal changes, and generate a current health status assessment. These key indicators may include physiological parameters such as blood pressure, blood sugar level, heart rate, as well as symptom descriptions and other relevant health records to fully reflect the patient's current physical condition.

[0100] In the embodiments of the present application, first, based on the latest data provided by the instant assessment report, the system deeply analyzes the current health status of each patient to identify important health indicators and abnormal change points; secondly, combined with the disease progression pattern simulation results previously obtained through the graph neural network algorithm, the patient's future health trend is predicted, and a detailed future health trend forecast is generated; thirdly, based on the future health trend forecast, statistical models and machine learning algorithms are used to conduct a comprehensive quantitative analysis of factors affecting long-term health, and a preliminarily calculated health risk score for each patient; finally, taking into account the individual characteristics of each patient (such as age, gender, genetic background, etc.), the preliminary health risk score is personalized, and a risk assessment system is introduced to set different risk thresholds, and finally the potential health risk score of each patient is determined, thereby achieving classified management of patients at different risk levels.

[0101] Here is a specific example:

[0102] Suppose that the cardiology department of a hospital wants to improve the long-term health management of patients with coronary heart disease and adopts the above technical solution. First, when a patient with coronary heart disease uploads the latest physical examination report and daily monitoring data, the system immediately analyzes it in detail and finds that his cholesterol level has increased slightly and he has had chest tightness symptoms recently, and generates an assessment of his current health status; secondly, combined with the simulation results of the coronary heart disease progression model constructed previously, the system predicts that the patient has an increased risk of acute cardiovascular events in the next 6 months; thirdly, based on this prediction, the system applies statistical models and machine learning algorithms, comprehensively considers a variety of long-term health risk factors (such as smoking history, family history, etc.), and calculates the patient's preliminary health risk score; finally, considering the patient's older age and the complexity of his personal medical history, the system makes personalized adjustments to his preliminary score, sets the corresponding risk threshold, classifies the patient into a high-risk category, and generates a detailed potential health risk score report. Through the above steps, cardiologists can take preventive measures in time, such as adjusting drug treatment plans or recommending lifestyle changes to reduce patients' future health risks.

[0103] This application takes into account that in the prior art, due to the lack of effective integration and deep mining of multi-source data, it is difficult for medical decision support systems to provide accurate disease predictions and personalized treatment recommendations. In addition, existing models often ignore the complex relationships between patients and between patients and environmental factors, making the prediction results not comprehensive enough and difficult to explain. Therefore, an embodiment of the present invention proposes an optional solution, which aims to solve the above problems by introducing advanced technologies such as graph neural network (GNN) algorithms and nonlinear activation functions, provide more accurate disease progression pattern simulation results, and provide a scientific basis for clinical decision-making.

[0104] Optionally, the application of a graph neural network algorithm to model the relationships between patients and between patients and environmental factors in the dynamically updated library to obtain simulation results of the disease progression pattern includes:

[0105] The data is cleaned and preprocessed, and feature vectors are extracted through feature engineering. Then, feature vectors are compared using similarity algorithms or distance metrics to generate scores representing similarity. At the same time, the impact of disease progression patterns is considered, and R is calculated by combining nonlinear activation functions and other factors. ij , where the strength of the relationship between patient i and patient j is R ij Calculated by the following formula:

[0106] R ij =σ(α(S ij +βP i (t)+γP j (t))-δ ij )

[0107] Among them, R ij represents the strength of the relationship between patient i and patient j; σ is the sigmoid function used to simulate nonlinear activation; S ij is the similarity score between patient i and patient j; α controls the steepness of the function; β is the coefficient of the patient's own disease progression pattern on the strength of the relationship; P i (t) is the simulation result of the disease progression pattern of the patient at time; γ is the influence coefficient of the patient's own disease progression pattern on the relationship strength; P j (t) is the simulation result of the disease progression pattern of patient j at time t; δ ij is a random noise term to simulate the uncertainty in real-world data;

[0108] Based on the calculated relationship strength R ij , construct a weighted graph to reflect the interaction between the patient population and its environmental factors, use a graph neural network to update the node representation, learn deep representation, and then calculate the disease progression pattern simulation result P(t) at each time point to predict the disease development trend. The simulation result P(t) is calculated by the following formula:

[0109]

[0110] Where P(t) represents the simulation result of the predicted disease progression pattern at time t; N is the number of patients; M is the number of environmental factors; w ij (t) is the weight of the patient's environmental factors at time t, which can be obtained based on historical data analysis; g(E i (t),E j (t)) is a nonlinear interaction function used to describe the interaction between patient i and environmental factors, E i (t) and E j (t) are the states or attributes of patient I and environmental factors at time; γ is the coefficient for adjusting the scale of the overall output; θ is the power exponent, which is used to enhance or weaken the nonlinear effect of the comprehensive impact assessment;

[0111] The simulation results are optimized through smoothing and trend analysis, and expert knowledge is introduced to analyze the model output, understand the key influencing factors and their causal relationships, adjust the model parameters or retrain according to the analysis results, and evaluate the cost-effectiveness of different treatment options to provide a scientific basis for clinical decision-making, assist doctors in formulating the best treatment plan, and obtain simulation results of the disease progression pattern.

[0112] This formula is designed to improve the accuracy of disease prediction and enhance the interpretability of the model. This solution uses the graph neural network algorithm to model the patient information in the dynamically updated library, generate the simulation results P(t) of the disease progression pattern, and optimize these results through smoothing and trend analysis. This method not only takes into account the similarities between patients and the influence of their respective disease progression patterns, but also combines the role of environmental factors to ensure the comprehensiveness and meticulousness of the prediction. At the same time, by introducing expert knowledge to parse the model output and understand the key influencing factors and their causal relationships, a closed-loop optimization process from data to decision-making is achieved.

[0113] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0114] R ij =σ(α(S ij +βP i (t)+γP j (t))-δ ij )

[0115] The nonlinear activation function σ is used to simulate nonlinear activation; S ij +βP i (t)+γP j (t) is to comprehensively consider the similarities between patients and the impact of their respective disease progression patterns on the strength of the relationship between them; the random noise term δ ij To simulate the uncertainty in real-world data;

[0116] The following is a brief introduction to how to obtain the parameters of the formula:

[0117] Similarity score S ij Obtained by comparing eigenvectors; the disease progression model influence coefficients β and γ are obtained based on historical data analysis; the disease progression model simulation result P i (t) and P j (t) is derived from previous predictions of disease progression patterns; random noise term δ ij Randomly generated according to actual data distribution;

[0118] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0119]

[0120] The overall output scale factor γ adjusts the scale of the final output; weighted sum The interactive effects of all patient and environmental factors are integrated; the power exponent θ is used to enhance or weaken the nonlinear effect of the comprehensive impact assessment;

[0121] The following is a brief introduction to how to obtain the parameters of the formula:

[0122] The overall output scale coefficient γ is set according to the actual situation; the weight w ij (t) Based on historical data analysis; relationship strength R ij Obtained by calculation; nonlinear interaction function g(E i (t),E j (t)) describes the interaction between patients and environmental factors. The specific form is determined according to the application field. The power exponent θ selects the optimal value based on experimental results.

[0123] Suppose a hospital wants to optimize the long-term management of diabetic patients through this solution. First, the hospital collects the electronic medical record data of all diabetic patients in the past five years, including basic information such as age, gender, body mass index, blood sugar level, and real-time physiological parameters uploaded by daily health monitoring equipment. Then, the data is cleaned and preprocessed, and the feature vector of each patient is extracted through feature engineering, for example:

[0124] S ij =CosineSimilarity(v i ,v j )

[0125] Among them, v i and v j are the feature vectors of patient i and patient j respectively. Then, the similarity score S between all patient pairs is calculated using the similarity algorithm ij , and combined with the simulation results of their respective disease progression models P i (t) and P j (t), the relationship strength R is calculated using the following formula ij :

[0126] R ij =σ(0.5(S ij +0.8P i (t)+0.8P j (t))-0.1)

[0127] Assume that after calculation, a set of relationship strength matrices is obtained. Based on this, a weighted graph is constructed to reflect the interaction between the patient population and environmental factors, such as air quality, seasonal changes, etc. Next, the graph neural network is applied to update the node representation, learn the deep representation, and then calculate the disease progression pattern simulation result P(t) at each time point, such as:

[0128]

[0129] Assuming that the threshold is set to 0.6, if the calculated result is greater than the threshold, it indicates that some patients have a higher risk of disease progression and need to strengthen monitoring and intervention measures. The simulation results are optimized through smoothing and trend analysis, and expert knowledge is introduced to analyze the model output to understand the key influencing factors and their causal relationships. The model parameters are adjusted or retrained according to the analysis results, and the cost-effectiveness of different treatment options is evaluated to assist doctors in formulating the best treatment plan, thereby significantly improving the quality and efficiency of medical services.

[0130] In order to meet the needs of more in-depth analysis of model prediction results and optimization of personalized medical advice, in some embodiments, based on the dynamic risk warning signal described in step 103, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm, and through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed, and the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library, and combined with the individual needs of patients and socioeconomic impacts, a medical advice report is generated, including:

[0131] Based on the dynamic risk warning signal, the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm are deeply analyzed and processed to obtain preliminary analysis results; using the preliminary analysis results, the feature importance analysis in the explanatory artificial intelligence algorithm is introduced to reveal the key factors affecting disease progression and generate a list of key factors; based on the list of key factors, a local interpretable model is applied to analyze the causal relationship and interaction mechanism of the key factors, and a detailed analysis report is generated; based on the detailed analysis report, the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and a cost-effectiveness analysis report is generated; combined with the cost-effectiveness analysis report, the individual needs of patients and socioeconomic impacts are considered, and personalized adjustments are made to generate personalized medical recommendations; based on the personalized medical recommendations, the detailed analysis report and the cost-effectiveness analysis report are combined to generate a medical recommendation report.

[0132] In this embodiment, based on the dynamic risk warning signal, the model prediction results generated by the graph neural network (GNN) algorithm are first analyzed and processed to obtain preliminary analysis results. These data cover the probability of disease progression, risk assessment results, and potential risk factors, and are used to reveal the patterns hidden behind complex data.

[0133] In the embodiment of the present application, first, the system uses an explanatory artificial intelligence algorithm to conduct an in-depth analysis of the results output by the GNN model, identifies which factors are most critical to the progression of the disease, and generates a list of key factors; secondly, based on these key factors, a local interpretable model is applied to further analyze their causal relationships and their interaction mechanisms to form a detailed analysis report; thirdly, based on this detailed analysis report, a health economics evaluation method is used to quantify the cost-effectiveness ratio of different treatment options to ensure the effective allocation of resources; finally, based on the cost-effectiveness analysis report and personalized needs and socioeconomic impacts, medical advice is personalized, and finally a comprehensive medical advice report is generated to guide clinical decision-making.

[0134] Here is a specific example:

[0135] Suppose a specialized oncology hospital wants to improve the personalized treatment effect of lung cancer patients and adopts the above technical solution. First, when the system receives a dynamic risk warning signal from a lung cancer patient, it immediately conducts an in-depth analysis of the disease progression probability and risk assessment results predicted by the GNN model, and finds that smoking history, gene mutation type and environmental exposure are the main risk factors; secondly, by introducing an explanatory AI algorithm, the system generates a list of key factors and applies a local interpretable model to reveal the causal relationship and interaction between these factors, such as how specific gene mutations increase sensitivity to certain chemotherapy drugs; thirdly, based on this detailed analysis report, the system uses health economics evaluation methods to compare a variety of possible treatment options (such as surgery, radiotherapy, targeted therapy, etc.), quantifies the cost-effectiveness ratio of each option, and obtains the best choice; finally, taking into account the patient's specific situation (such as age, economic status, personal preferences, etc.), the system makes personalized adjustments to its recommended treatment plan and generates a detailed medical advice report for the attending physician as a reference. Through the above steps, the hospital can formulate the most appropriate treatment plan for each patient, which not only improves the treatment effect, but also effectively controls medical costs.

[0136] In order to meet the need for more accurate cost-effectiveness evaluation, in some embodiments, based on the detailed analysis report, a health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library to generate a cost-effectiveness analysis report, including:

[0137] Based on the detailed analysis report, the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library to generate a preliminary cost-effectiveness evaluation; using the preliminary cost-effectiveness evaluation, combined with the key factors and their interaction mechanisms in the detailed analysis report, the cost and expected effect of each treatment option are refined to generate a refined cost-effectiveness evaluation; based on the refined cost-effectiveness evaluation, socioeconomic influencing factors are introduced, and different treatment options are comprehensively adjusted to generate a comprehensive cost-effectiveness evaluation; based on the comprehensive cost-effectiveness evaluation, cost-effectiveness thresholds for different treatment options are set, and each option is classified to obtain a cost-effectiveness classification result; combined with the cost-effectiveness classification result, the individual needs of patients are taken into consideration, and optimization configuration recommendations are made for the selection of treatment options and resource allocation to generate a cost-effectiveness analysis report.

[0138] In this example, the cost-effectiveness ratio of different treatment options integrated from the dynamic update library is quantified based on the key factors and their interaction mechanisms provided by the detailed analysis report to generate a preliminary cost-effectiveness assessment. These data cover the direct and indirect costs of different treatment options, as well as the expected treatment effects, and are used to evaluate the economic feasibility and clinical value of each option.

[0139] In the embodiment of the present application, first, the system applies the health economics evaluation method based on the information in the detailed analysis report, and performs preliminary cost-effectiveness calculations on a variety of possible treatment options to form a preliminary evaluation result; secondly, combined with the key factors and their interaction mechanisms revealed in the detailed analysis report, the cost structure and expected effects of each treatment option are further refined to ensure the accuracy and comprehensiveness of the evaluation; thirdly, socioeconomic influencing factors (such as the patient's family income level, regional medical resource distribution, etc.) are introduced to comprehensively adjust different treatment options and generate a comprehensive cost-effectiveness evaluation; finally, based on the comprehensive evaluation results, a cost-effectiveness threshold is set, each option is classified, and the individual needs of the patient are considered, and optimization configuration suggestions are proposed to finally generate a detailed cost-effectiveness analysis report.

[0140] Here is a specific example:

[0141] Suppose a large general hospital wants to optimize the long-term management plan for diabetic patients and adopts the above technical solution. First, when the system receives a detailed analysis report, it immediately applies the health economics evaluation method to quantify the cost-effectiveness ratio of several common diabetes treatment plans (such as oral medication, insulin injection, lifestyle intervention, etc.) and generates preliminary evaluation results; secondly, combined with the key factors revealed in the analysis report (such as blood sugar control, complication risk, etc.), the system refines the specific cost and expected effect of each treatment plan. For example, although a certain oral medication has a lower initial cost, it may lead to a higher complication rate in the long run; thirdly, considering the patient's socioeconomic background (such as family income, medical resources in the place of residence, etc.), the system makes comprehensive adjustments to different treatment plans and generates a comprehensive cost-effectiveness evaluation. It is found that for patients in low-income areas, lifestyle intervention may be a more economical and effective option; finally, based on the evaluation results, the cost-effectiveness threshold is set, and the treatment plans are divided into three categories: high, medium, and low efficiency. In combination with the individual needs of each patient (such as age, occupation, living habits, etc.), specific optimization configuration suggestions are proposed, such as recommending specific treatment paths or resource allocation plans. Through the above steps, the hospital can develop the most appropriate long-term management plan for each diabetic patient, which not only improves the treatment effect but also effectively controls medical costs.

[0142] This application takes into account that in the prior art, due to the lack of comprehensive analysis of feature importance and change sensitivity, it is difficult for medical decision support systems to provide transparent and reliable disease prediction results. In addition, existing models usually ignore the specific contribution of feature changes to the prediction results, making the identification of key factors inaccurate, affecting the effectiveness of personalized medical advice. Therefore, an embodiment of the present invention proposes an optional solution, which aims to solve the above problems by introducing feature importance analysis and comprehensive impact index calculation in an explanatory artificial intelligence algorithm, reveal the key factors that affect disease progression, and provide a basis for generating a list of key factors.

[0143] Optionally, the preliminary analysis results are used to introduce feature importance analysis in an explanatory artificial intelligence algorithm to reveal key factors affecting disease progression and generate a list of key factors, including:

[0144] Patient data is cleaned and standardized, and the impact of each feature change on disease progression prediction is evaluated through perturbation experiments. Key features are identified and feature importance scores are calculated. The feature importance score FI i It is calculated by the following formula:

[0145]

[0146] Among them, FI irepresents the importance score of the feature; N is the number of samples; M is the number of all features; ρ j is the representative weight of the jth sample, reflecting its influence in model training; h(x i ,x j ) is the effect of feature i combined with the jth sample on the prediction of disease progression; x' i It is a new data set after the features are randomly perturbed, which is used to evaluate the impact of feature changes on the prediction results;

[0147] The explanatory AI algorithm is used to analyze the specific contribution of features, evaluate the sensitivity of feature changes, combine importance and sensitivity analysis, and calculate the comprehensive impact index to provide a basis for generating a list of key factors. The comprehensive impact index is calculated using the following formula:

[0148]

[0149] Among them, KI c is the comprehensive impact index of the key factor c; C is the set of factors identified as key; η is a balancing factor used to adjust the relative contribution between feature importance and direct explanatory measures; q i DI is the sensitivity coefficient of the characteristic change, which reflects the influence of a small change in the characteristic value on the prediction result of disease progression; i is a direct measure of the importance of the feature obtained from the local interpretable model; |C| is the size of the key factor set C;

[0150] Set thresholds to screen key factors, sort them by comprehensive impact index, make personalized adjustments based on the patient's individual needs and socioeconomic impact, and prepare a detailed analytical report.

[0151] This formula aims to improve the transparency and credibility of the disease prediction model while ensuring the accuracy of key factor identification. This solution first cleans and standardizes the patient data, evaluates the impact of each feature change on disease progression prediction through perturbation experiments, identifies key features, and calculates the feature importance score FI. i Next, we use the explanatory AI algorithm to analyze the specific contribution of the features, evaluate the sensitivity of feature changes, and calculate the comprehensive impact index KI by combining importance and sensitivity analysis. c This process not only reveals which features are most important for disease prediction, but also provides a more detailed and personalized list of key factors by quantifying the impact of feature changes, providing strong scientific support for clinical decision-making.

[0152] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0153]

[0154] Measure the impact of feature changes on the prediction results of all samples; A normalization factor to ensure that the importance scores of different features are comparable;

[0155] The following is a brief introduction to how to obtain the parameters of the formula:

[0156] The number of samples N is directly obtained from the preprocessed data set; the number of features M is determined according to the results of feature engineering extraction; the representative weight ρ j Calculated through the distribution of training set samples; the function h(x i ,x j ) is the effect of feature G combined with the jth sample on the prediction of disease progression, obtained through model prediction; x' i It is a new data set after the features are randomly perturbed, generated through perturbation experiments;

[0157] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0158]

[0159] Consider the product of feature importance and change sensitivity to reflect the comprehensive impact of feature changes on prediction results; Introducing direct importance measures for local interpretable models to balance feature importance and interpretability;

[0160] The following is a brief introduction to how to obtain the parameters of the formula:

[0161] Comprehensive impact index KI c Used to measure the overall influence of key factor c; the key factor set C is obtained by screening the preliminary analysis results; the balance factor η is set according to actual needs; the characteristic change sensitivity coefficient q i Derived through experiments or historical data analysis; direct importance measurement DI i Obtained from local interpretable models (such as LIME or SHAP); the key factor set size |C| is counted from the filtered key factor list;

[0162] Suppose a hospital wants to optimize the early warning of cardiovascular disease through this solution. First, the hospital collects and cleans the electronic medical record data of all cardiovascular disease patients in the past three years, including basic information such as age, blood pressure, cholesterol level, and real-time physiological parameters uploaded by daily health monitoring equipment. Then, the data is standardized and the impact of each feature change on the prediction of disease progression is evaluated through perturbation experiments. For example:

[0163]

[0164] Suppose that after calculation, we get the importance scores of a set of features. Then, we use the interpretive A1 algorithm to analyze the specific contribution of the features and evaluate the sensitivity of feature changes, such as the impact of small changes in blood pressure and cholesterol levels on the prediction results of disease progression. Combined with importance and sensitivity analysis, we calculate the comprehensive impact index KI c :

[0165]

[0166] Assuming the threshold is set to 0.8, if the calculated result is greater than the threshold, it indicates that certain characteristics are key factors affecting disease progression and require special attention. By setting the threshold to screen key factors and sorting them by comprehensive impact index, a detailed analysis report is compiled to make personalized adjustments based on the individual needs and socioeconomic impact of the patient. Ultimately, the hospital is able to develop the most appropriate treatment plan for each patient, improving the quality and efficiency of medical services.

[0167] In order to solve the problem of continuous optimization of medical advice reports and improvement of the accuracy of personalized advice, in some embodiments, the medical advice report is combined in step 104 to establish a dynamic learning and feedback mechanism, optimize the personalized advice in the medical advice report, and build a medical decision support system, including:

[0168] The medical advice report is used to collect data feedback on the treatment plans adopted by doctors and the actual implementation effects, and a feedback data set is generated; based on the feedback data set, the personalized suggestions in the medical advice report are evaluated in combination with the individual needs and socioeconomic impact of the patients, and an evaluation result is generated; based on the evaluation result, a machine learning algorithm is applied to optimize the personalized suggestions in the medical advice report, and an optimized personalized suggestion is generated; using the optimized personalized suggestion, a knowledge base of the medical decision support system is updated to generate an updated knowledge base; based on the updated knowledge base, the recommendation strategy of the system is adjusted so that the recommendation strategy can be automatically updated according to new data and feedback, and an updated recommendation strategy is generated; in combination with the updated recommendation strategy, a dynamic learning and feedback mechanism is established to improve the personalized suggestions in the medical advice report and build a medical decision support system.

[0169] In this embodiment, the medical advice report is first used to collect data feedback on the treatment plan adopted by the doctor and its actual implementation effect to generate a feedback data set. This data includes the specific content of the treatment plan, the adjustment during the implementation process, and the final efficacy evaluation results, which are used to evaluate the effectiveness and applicability of the existing medical advice.

[0170] In the embodiment of the present application, first, the system collects the treatment plans adopted by doctors and their effectiveness feedback through medical advice reports to form a detailed feedback data set; secondly, based on this feedback data set, combined with the patient's individual needs and socioeconomic impact, the personalized suggestions in the medical advice report are comprehensively evaluated to ensure their rationality and feasibility; thirdly, based on the evaluation results, the personalized suggestions are optimized using machine learning algorithms to improve the accuracy and effectiveness of the suggestions; finally, the optimized personalized suggestions are updated to the knowledge base of the medical decision support system, and the recommendation strategy is adjusted so that the system can be automatically updated according to new data and feedback, thereby building a dynamic learning and feedback mechanism to continuously improve the quality of medical advice reports.

[0171] Here is a specific example:

[0172] Suppose a community hospital wants to improve the effectiveness of chronic disease management by introducing a dynamic learning and feedback mechanism. First, when the doctor adopts the personalized suggestions in the medical advice report, the system will automatically record the specific implementation of the treatment plan and the actual recovery effect of the patient to form a feedback data set; secondly, based on these feedback data, the system combines each patient's individual needs (such as age, lifestyle, etc.) and socioeconomic background (such as family income, medical insurance coverage, etc.) to evaluate the effectiveness of the original personalized suggestions and identify which aspects need improvement; thirdly, based on the evaluation results, the system applies machine learning algorithms to optimize personalized suggestions, such as adjusting drug dosage or recommending more appropriate rehabilitation activities; finally, the optimized personalized suggestions are updated to the knowledge base of the hospital's medical decision support system, and the recommendation strategy is adjusted so that the system can automatically update its suggestions based on the latest data and feedback. Through the above steps, the community hospital not only improves the efficiency and effectiveness of chronic disease management, but also enhances patients' trust and compliance with treatment plans.

[0173] Figure 2 A schematic diagram of the structure of a medical decision support system based on big data is provided for the embodiment of the present application, such as Figure 2 As shown, the device comprises:

[0174] Integration module 21, used to integrate and merge patient information from different medical institutions, personal health monitoring devices and public health databases to obtain a dynamic update library containing health management records of the patient throughout his or her life cycle;

[0175] A prediction module 22, for applying a graph neural network algorithm to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and combining real-time streaming data analysis technology to instantly evaluate newly acquired patient information and generate dynamic risk warning signals;

[0176] A quantification module 23 is used to introduce an explanatory artificial intelligence algorithm to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm based on the dynamic risk warning signal, reveal the key factors and interaction mechanisms affecting disease progression through feature importance analysis and local interpretable models in the artificial intelligence algorithm, apply health economics evaluation methods to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and generate a medical advice report in combination with the individual needs of patients and socioeconomic impacts;

[0177] The optimization module 24 is used to establish a dynamic learning and feedback mechanism in combination with the medical advice report, optimize the personalized advice in the medical advice report, and build a medical decision support system.

[0178] Figure 2 The medical decision support system based on big data can be executed Figure 1 The implementation principle and technical effect of the medical decision support method based on big data described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the medical decision support system based on big data in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0179] In one possible design, Figure 2 A medical decision support system based on big data in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0180] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0181] The processing component 32 is used to: integrate and fuse patient information from different medical institutions, personal health monitoring devices and public health databases to obtain a dynamic update library, which contains health management records of patients throughout their life cycle; apply a graph neural network algorithm to analyze the relationship between patients and between patients and environmental factors in the dynamic update library to predict the development trend of individualized diseases, and combine real-time streaming data analysis technology to instantly evaluate newly acquired patient information and generate dynamic risk warning signals; based on the dynamic risk warning signals, introduce an explanatory artificial intelligence algorithm to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm, reveal the key factors and interaction mechanisms that affect disease progression through feature importance analysis and local interpretable models in the artificial intelligence algorithm, apply health economics evaluation methods to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and generate a medical advice report in combination with the individualized needs of patients and socioeconomic impacts; and establish a dynamic learning and feedback mechanism in combination with the medical advice report to optimize the personalized advice in the medical advice report and build a medical decision support system.

[0182] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0183] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0184] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0185] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0186] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0187] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0188] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 A medical decision support method based on big data in the illustrated embodiment.

[0189] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0190] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0191] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A medical decision support method based on big data, characterized in that: include: Integrate and merge patient information from different medical institutions, personal health monitoring devices, and public health databases to obtain a dynamic update library containing health management records of patients throughout their life cycle; Applying a graph neural network algorithm to analyze the relationships between patients and between patients and environmental factors in the dynamically updated library to predict individualized disease development trends, and combining real-time streaming data analysis technology to instantly evaluate newly acquired patient information and generate dynamic risk warning signals; Based on the dynamic risk warning signal, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm. Through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed. The health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library, and a medical advice report is generated in combination with the individual needs of patients and socioeconomic impacts. In combination with the medical advice report, a dynamic learning and feedback mechanism is established to optimize the personalized suggestions in the medical advice report and build a medical decision support system.

2. The method according to claim 1, characterized in that The application of the graph neural network algorithm analyzes the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and combines the real-time streaming data analysis technology to instantly evaluate the newly acquired patient information and generate dynamic risk warning signals, including: Applying a graph neural network algorithm to model the relationships between patients and between patients and environmental factors in the dynamically updated library, to obtain simulation results of disease progression patterns; Based on the simulation results of the disease progression pattern, real-time streaming data analysis technology is applied to instantly evaluate and process newly acquired patient information and generate an instant evaluation report; Utilizing the instant assessment report and combining it with the simulation results of the disease progression pattern, the patient's future health risk is quantified to obtain a potential health risk score; According to the potential health risk score, patients at different risk levels are classified through comprehensive evaluation and setting risk thresholds to generate dynamic risk warning signals.

3. The method according to claim 2, characterized in that The simulation results based on the disease progression model use real-time streaming data analysis technology to instantly evaluate and process newly acquired patient information and generate an instant evaluation report, including: Based on the simulation result of the disease progression pattern, preprocessing the newly acquired patient information to obtain preprocessed patient information; Applying real-time streaming data analysis technology to instantly analyze and process the pre-processed patient information, identify abnormalities or key change points related to disease progression, and obtain instant analysis results; Based on the instant analysis results and the simulation results of the disease progression model, a comprehensive evaluation is performed on the patient's current health status to obtain an instant health status evaluation; Based on the instant health status assessment, an instant assessment report is generated, and the instant assessment report includes the patient's current health status, risk factors and possible development trends.

4. The method according to claim 2, characterized in that: The instant assessment report is used in combination with the simulation results of the disease progression model to quantify the patient's future health risks and obtain a potential health risk score, including: Utilizing the instant assessment report, the newly acquired patient information is analyzed and processed in detail to identify key indicators and abnormal changes in the current health status and generate a current health status assessment; According to the current health status assessment, combined with the simulation results of the disease progression pattern, the patient's future health trend is predicted and processed to generate a future health trend prediction; Based on the future health trend prediction, the patient's long-term health risk factors are quantitatively analyzed and processed using statistical models and machine learning algorithms to obtain a preliminary health risk score; Using the preliminary health risk score, combined with the individual characteristics of the patient, to perform personalized adjustment processing to generate a personalized health risk score; According to the personalized health risk score, a risk assessment system is introduced to set different risk thresholds, and patients at different risk levels are classified and processed to obtain a potential health risk score.

5. The method according to claim 1, characterized in that Based on the dynamic risk warning signal, an explanatory artificial intelligence algorithm is introduced to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm. Through the feature importance analysis and local interpretable model in the artificial intelligence algorithm, the key factors and interaction mechanisms affecting disease progression are revealed. The health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamic update library, and combined with the individual needs of patients and socioeconomic impacts, a medical advice report is generated, including: Based on the dynamic risk warning signal, the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm are deeply analyzed and processed to obtain preliminary analysis results; Using the preliminary analysis results, feature importance analysis in the explanatory artificial intelligence algorithm is introduced to reveal the key factors affecting disease progression and generate a list of key factors; According to the list of key factors, a local interpretable model is applied to analyze the causal relationship and interaction mechanism of the key factors, and a detailed analysis report is generated; Based on the detailed analysis report, applying health economics evaluation methods, quantifying the cost-effectiveness ratio of different treatment plans integrated from the dynamically updated library, and generating a cost-effectiveness analysis report; Combined with the cost-effectiveness analysis report, taking into account the individual needs of patients and socioeconomic impacts, personalized adjustments are made to generate personalized medical recommendations; Based on the personalized medical advice, the detailed analysis report and the cost-effectiveness analysis report are combined to generate a medical advice report.

6. The method according to claim 5, characterized in that Based on the detailed analysis report, the health economics evaluation method is applied to quantify the cost-effectiveness ratio of different treatment plans integrated from the dynamic update library, and generate a cost-effectiveness analysis report, including: Based on the detailed analysis report, applying health economics evaluation methods, quantifying the cost-effectiveness ratio of different treatment plans integrated from the dynamically updated library, and generating a preliminary cost-effectiveness evaluation; Using the preliminary cost-effectiveness assessment, combined with the key factors and their interaction mechanisms in the detailed analytical report, the costs and expected effects of each treatment option are refined to generate a refined cost-effectiveness assessment; According to the detailed cost-effectiveness evaluation, socioeconomic factors are introduced to comprehensively adjust different treatment plans and generate a comprehensive cost-effectiveness evaluation; Based on the comprehensive cost-effectiveness evaluation, cost-effectiveness thresholds of different treatment options are set, and each option is classified to obtain a cost-effectiveness classification result; In combination with the cost-effectiveness classification results, the individual needs of patients are taken into consideration, and optimization recommendations are made on the selection of treatment options and resource allocation to generate a cost-effectiveness analysis report.

7. The method according to claim 1, characterized in that The medical advice report is combined with a dynamic learning and feedback mechanism to optimize the personalized advice in the medical advice report and build a medical decision support system, including: Using the medical advice report, data feedback on the treatment plan adopted by the doctor and the actual implementation effect is collected to generate a feedback data set; According to the feedback data set, combined with the individual needs and socioeconomic impact of the patient, the personalized recommendations in the medical recommendation report are evaluated to generate an evaluation result; Based on the evaluation results, applying a machine learning algorithm to optimize the personalized recommendations in the medical recommendation report to generate optimized personalized recommendations; Using the optimized personalized suggestions, updating the knowledge base of the medical decision support system to generate an updated knowledge base; According to the updated knowledge base, adjusting the recommendation strategy of the system so that the recommendation strategy can be automatically updated according to new data and feedback to generate an updated recommendation strategy; In combination with the updated recommendation strategy, a dynamic learning and feedback mechanism is established to improve the personalized recommendations in the medical advice report and build a medical decision support system.

8. A medical decision support system based on big data, characterized in that: include: An integration module is used to integrate and fuse patient information from different medical institutions, personal health monitoring devices, and public health databases to obtain a dynamic update library containing health management records of patients throughout their life cycle; A prediction module, which is used to apply a graph neural network algorithm to analyze the relationship between patients and between patients and environmental factors in the dynamically updated library to predict the development trend of individualized diseases, and to instantly evaluate newly acquired patient information in combination with real-time streaming data analysis technology to generate dynamic risk warning signals; A quantification module is used to introduce an explanatory artificial intelligence algorithm to conduct in-depth analysis of the disease progression probability and risk assessment results predicted by the model generated by the graph neural network algorithm based on the dynamic risk warning signal, reveal the key factors and interaction mechanisms that affect disease progression through feature importance analysis and local interpretable models in the artificial intelligence algorithm, apply health economics evaluation methods to quantify the cost-effectiveness ratio of different treatment options integrated from the dynamically updated library, and generate a medical advice report in combination with the individual needs of patients and socioeconomic impacts; The optimization module is used to combine the medical advice report, establish a dynamic learning and feedback mechanism, optimize the personalized suggestions in the medical advice report, and build a medical decision support system.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a medical decision support method based on big data as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a medical decision support method based on big data as described in any one of claims 1 to 7 is implemented.

Images