Method and system for predicting death risk of hemodialysis patient based on artificial intelligence

Abstract

The invention relates to the technical field of risk prediction, in particular to a method and system for predicting the death risk of a hemodialysis patient based on artificial intelligence. Comprising the following steps: acquiring physiological index data from dialysis equipment, constructing a multi-source data set and processing the multi-source data set to obtain a cleaned data set; time sequence features are extracted, and a basic feature set is constructed; for the basic feature set, analyzing the influence of the index change on the health state, and generating an individualized health portrait; according to the individualized health portrait, potential factors related to risks are extracted, and risk assessment parameters are determined; for the risk assessment parameter, analyzing a risk trigger condition, and judging an index fluctuation condition to obtain risk category distribution; and according to the risk category distribution, obtaining a matched intervention rule, extracting an adjustment strategy, and determining a personalized intervention framework. According to the method, accurate death risk prediction and personalized intervention can be provided, the health management effect of dialysis patients is remarkably improved, and the death risk is reduced.

Description

Technical Field

[0001] This invention relates to the field of risk prediction technology, and in particular to a method and system for predicting the mortality risk of hemodialysis patients based on artificial intelligence. Background Technology

[0002] Hemodialysis is a common treatment for patients with kidney failure. However, during dialysis, a patient's health condition can change drastically, leading to risks such as low blood pressure, cardiac complications, and even increasing the likelihood of death. While current dialysis treatments incorporate some health monitoring methods, the lack of precise mortality risk prediction and a dynamic health management system means that many potential risks fail to be identified and effectively intervened in a timely manner. This results in patients' health conditions not being adequately assessed and adjusted in a timely manner during actual treatment.

[0003] Traditional methods for predicting dialysis risks rely primarily on regular checkups and single physiological indicators, failing to comprehensively reflect the dynamic changes in a patient's health. Even when monitoring physiological data such as blood pressure, heart rate, and body temperature, these data are often static and cannot reflect real-time fluctuations and changes in the patient's health status. Moreover, existing health assessments are typically based on traditional medical experience and rough standards, lacking personalized and real-time feedback mechanisms, and cannot be flexibly adjusted according to individual patient differences and unexpected changes that occur during dialysis.

[0004] Furthermore, traditional risk management systems often lack sufficient flexibility and precision, failing to achieve personalized health assessments. During dialysis treatment, individual differences, medical histories, and treatment responses among patients vary, making it difficult to achieve optimal treatment outcomes using standardized interventions. Therefore, how to generate personalized health profiles based on patients' real-time health data during dialysis, predict mortality risk, and develop personalized interventions is a key challenge currently facing technology.

[0005] To overcome the shortcomings of traditional methods, this invention proposes an artificial intelligence-based mortality risk prediction system for hemodialysis patients. This system utilizes real-time collected physiological and historical data, employing methods such as data analysis, temporal feature extraction, and lag effect modeling to accurately assess patients' mortality risk and implement dynamic health management through a personalized intervention framework. By introducing machine learning technology, the system can automatically learn new intervention strategies based on clinical data and adjust intervention measures in real time, ensuring timely and effective management of patients' health status during dialysis. Summary of the Invention

[0006] In its first aspect, this invention provides a method for predicting the mortality risk of hemodialysis patients based on artificial intelligence, mainly including: Step S1: Obtain physiological indicator data from the dialysis device, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; extract time-series features based on the cleaned dataset and construct a basic feature set; Step S2: For the aforementioned set of basic features, analyze the impact of changes in indicators on health status and generate a personalized health profile; Step S3: Based on the individualized health profile, extract potential risk-related factors and determine risk assessment parameters; Step S4: For the risk assessment parameters, analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution; Step S5: Based on the risk category distribution, obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework.

[0007] As a preferred embodiment of the present invention, the step of extracting time-series features and constructing a basic feature set based on the cleaned dataset includes: For the cleaned dataset, a historical trend analysis method is used to perform time-series decomposition on the physiological indicator data; through the time-series decomposition, the fluctuation characteristics and trend characteristics of the physiological indicators are extracted; based on the fluctuation characteristics and trend characteristics, key time-series feature parameters are determined; the key time-series feature parameters are integrated into a basic feature set; and the basic feature set is standardized to unify the data format.

[0008] As a preferred embodiment of the present invention, generating a personalized health profile includes: For the aforementioned set of basic features, lag effect analysis is used to assess the delayed impact of any change in an indicator on subsequent health status. The correlation between indicator changes and health status is determined through the lag effect analysis; an individualized health status model is constructed based on the correlation; the individualized health status model is compared with historical data to verify the accuracy of the model; if the prediction deviation of the model exceeds a preset threshold, the analysis parameters are adjusted to optimize the model output; finally, an individualized health profile is generated.

[0009] As a preferred embodiment of the present invention, based on the individualized health profile, potential risk-related factors are extracted, including: Based on the individualized health profile, and combined with prediction window setting technology, potential factors related to mortality risk are analyzed; Through the analysis, key risk influencing factors are extracted; based on these key risk influencing factors, an initial parameter set for risk assessment is constructed; weights are allocated to the initial parameter set to highlight the influence of the main factors; and finally, optimized risk assessment parameters are determined.

[0010] As a preferred embodiment of the present invention, extracting potential risk-related factors further includes: The weight allocation of the initial parameter group for risk assessment can be dynamically adjusted based on real-time monitoring data. The dynamic adjustment adopts an adaptive algorithm, which automatically adjusts the weight of each risk factor in the risk assessment parameter group based on the patient's real-time health data and the health change trend within the prediction window.

[0011] As a preferred embodiment of the present invention, obtaining the risk category distribution includes: For the aforementioned risk assessment parameters, the triggering conditions for hypotension risk during dialysis are analyzed; through the analysis, the fluctuation range of key indicators is determined; it is determined whether the fluctuation of the key indicators exceeds a preset threshold; if the fluctuation of the key indicators exceeds the preset threshold, it is marked as a high-risk state; based on the marking results, a risk category distribution is generated.

[0012] As a preferred embodiment of the present invention, determining a personalized intervention framework includes: For the risk category distribution, an intervention rule base matching the individualized health profile is obtained; adjustment strategies applicable to the current risk category are extracted from the intervention rule base; a personalized intervention plan framework is constructed based on the adjustment strategies; the feasibility of the personalized intervention plan framework is verified to ensure the executability of the strategies; if the adjustment strategies do not match the current health status, the intervention rules are re-selected; finally, the optimized personalized intervention framework is determined.

[0013] As a preferred embodiment of the present invention, determining the personalized intervention framework further includes: The intervention rule base automatically learns new intervention strategies from clinical data through machine learning models, updates and expands the rule base, and at the same time, the intervention framework combines real-time health monitoring data and adopts a real-time feedback mechanism to automatically adjust personalized intervention plans according to changes in the patient's health status.

[0014] Secondly, the present invention also provides a system for predicting the mortality risk of hemodialysis patients based on artificial intelligence, for implementing the above-mentioned method, the system comprising: The construction unit is used to acquire physiological index data from dialysis equipment, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; and extract time-series features based on the cleaned dataset to construct a basic feature set. The generation unit is used to analyze the impact of changes in indicators on health status based on the set of basic features, and generate a personalized health profile. The determination unit is used to extract potential risk-related factors and determine risk assessment parameters based on the individualized health profile; it is also used to obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework based on the risk category distribution. The acquisition unit is used to analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution for the risk assessment parameters.

[0015] Thirdly, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for predicting the mortality risk of hemodialysis patients based on artificial intelligence. It involves acquiring physiological indicator data from dialysis equipment in real time, constructing a multi-source dataset, cleaning it, and extracting time-series features to form a basic feature set. For this feature set, the method analyzes the lagged impact of indicator changes on health status, generating an individualized health profile. Based on the health profile, risk-related factors are extracted, risk assessment parameters are determined, and risk triggering conditions such as hypotension are analyzed to obtain the risk category distribution. Intervention rules are matched according to the risk distribution, adjustment strategies are extracted, and a personalized intervention framework is determined. This invention, through multi-dimensional data analysis and modeling, achieves accurate assessment and personalized intervention of patients' health status during dialysis, effectively reducing the risk of dialysis-related complications and improving the safety and effectiveness of dialysis treatment. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for predicting the mortality risk of hemodialysis patients based on artificial intelligence, as described in an embodiment of the present invention. Figure 2 This is a system structure diagram of predicting the mortality risk of hemodialysis patients based on artificial intelligence in an embodiment of the present invention. Detailed Implementation

[0018] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0019] like Figure 1 This embodiment of a method for predicting the mortality risk of hemodialysis patients based on artificial intelligence may specifically include: Step S1: Obtain physiological indicator data from the dialysis device, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; extract time-series features based on the cleaned dataset and construct a basic feature set; The construction of a basic feature set includes: acquiring multiple physiological indicator data in real time; integrating the physiological indicator data to construct a multi-source dataset; using denoising techniques to remove noise interference from the multi-source dataset; cleaning the multi-source dataset to remove outliers and missing values, generating a cleaned dataset; using historical trend analysis to perform time-series decomposition of the physiological indicator data on the cleaned dataset; extracting fluctuation and trend characteristics of the physiological indicators through the time-series decomposition; determining key time-series feature parameters based on the fluctuation and trend characteristics; integrating the key time-series feature parameters into a basic feature set; standardizing the basic feature set to unify the data format; and if redundant features exist in the basic feature set, feature filtering is performed to optimize the structure of the basic feature set.

[0020] Specifically, real-time data on various physiological indicators such as blood pressure, heart rate, and blood oxygen saturation are acquired from patients. Each of these indicators is stream data, originating from different sensors and devices, forming a multi-source dataset. After collection, data integration techniques are used to unify the format of the various physiological data and perform noise reduction to remove noise caused by equipment errors, signal interference, and other factors. The data cleaning process further removes outliers and missing values ​​to ensure data accuracy and completeness. For the remaining normal data, smoothing techniques are used to reduce data fluctuations and make the data more representative. Through these processing steps, a cleaned dataset is obtained, providing a reliable foundation for subsequent analysis. When processing the cleaned dataset, historical trend analysis is used to decompose the physiological indicator data over time to extract the fluctuation and trend characteristics. Time-series decomposition technology separates long-term trends and short-term fluctuations by performing historical analysis on physiological indicator data, such as blood pressure and heart rate. For example, during dialysis, a patient's blood pressure may show a long-term upward trend while also exhibiting periodic fluctuations in the short term. By analyzing these fluctuation and trend characteristics, we can better capture the dynamic changes in patients' health status.

[0021] Based on the fluctuation and trend features extracted from the time-series decomposition, key time-series feature parameters, such as the amplitude of data fluctuations and periodic changes, are further identified. These parameters can accurately reflect changes in the patient's physiological state. These time-series feature parameters are integrated to form a basic feature set, which serves as the basis for subsequent health assessment and risk prediction. To ensure data consistency and comparability, this basic feature set is standardized to unify the data format, enabling physiological data from different sources to be processed within the same analytical framework.

[0022] The aforementioned technical solution, by extracting and standardizing time-series features, can effectively identify fluctuations in health indicators related to mortality risk, thereby providing precise data support for personalized risk assessment and intervention strategies. For example, long-term elevation or fluctuation in blood pressure may indicate a potential risk of hypotension during dialysis, helping doctors adjust treatment plans in a timely manner and reduce the patient's risk of death.

[0023] Step S2: For the aforementioned set of basic features, analyze the impact of changes in indicators on health status and generate a personalized health profile; The process of generating a personalized health profile includes: using lag effect analysis to assess the delayed impact of any indicator change on subsequent health status for the set of basic features; determining the correlation between indicator changes and health status through the lag effect analysis; constructing a personalized health status model based on the correlation; comparing the personalized health status model with historical data to verify the model's accuracy; adjusting the analysis parameters and optimizing the model output if the model's prediction deviation exceeds a preset threshold; and finally generating a personalized health profile.

[0024] Specifically, based on a set of fundamental features, lag effect analysis is used to assess the delayed impact of changes in any physiological indicator on subsequent health status. This analytical method, through the temporal nature of historical data, reveals whether changes in specific physiological indicators, such as blood pressure and heart rate, will have a delayed effect on a patient's health status. Lag effect analysis identifies that the impact of changes in certain indicators on health status is not immediate but rather occurs with a certain time lag, providing an important basis for predicting long-term trends in health status. For example, if a patient's blood pressure gradually increases during dialysis, lag effect analysis can reveal whether this change will lead to increased cardiac burden and thus affect the patient's overall health. Establishing this correlation allows the health profile to accurately reflect the evolution of a patient's health status. Next, an individualized health status model is constructed based on this correlation. This model can dynamically predict the trend of changes in a patient's health status based on their physiological data. The constructed individualized health status model is verified for accuracy by comparing it with historical data. Historical data provides a reference standard for the model, and comparison can evaluate the model's predictive ability in practical applications. If the model's prediction deviation exceeds a preset threshold, the analysis parameters will be adjusted to optimize the model output and further improve the model's prediction accuracy. Through repeated adjustments and optimizations, the model can ultimately reflect the patient's health status more accurately, thereby generating an individualized health profile. This health profile not only considers current physiological data but also incorporates lag effects and historical trends, making it more forward-looking.

[0025] Step S3: Based on the individualized health profile, extract potential risk-related factors and determine risk assessment parameters; Based on the individualized health profile, potential risk-related factors are extracted, including: For the individualized health profile, using prediction window setting technology, potential factors related to mortality risk are analyzed; through this analysis, key risk influencing factors are extracted; based on these key risk influencing factors, an initial parameter set for risk assessment is constructed; weights are allocated to the initial parameter set to highlight the influence of major factors; if redundant parameters exist in the initial parameter set, parameter simplification is performed; finally, optimized risk assessment parameters are determined.

[0026] Specifically, step S3 involves extracting potential factors related to mortality risk based on the individualized health profile, and then determining risk assessment parameters. This is achieved by combining the individualized health profile with predictive window setting technology to analyze potential factors related to mortality risk. In this process, the individualized health profile contains multi-dimensional data, including the patient's physiological state, medical history, and lifestyle habits. This data, after preprocessing and feature extraction, forms a health profile with temporal characteristics. In-depth analysis of these profiles reveals key factors closely related to mortality risk, such as trends in physiological indicators like blood pressure fluctuations, blood sugar levels, and abnormal heart rates, as well as the influence of external factors like patient medical history, medication dependence, and lifestyle. The analysis incorporates predictive windowing techniques—identifying an appropriate time window, such as 30 or 90 days—to assess which factors in the individualized health profile have the greatest impact on mortality risk within that window. Historical data comparison and statistical analysis identify potentially highly correlated factors. For example, if a patient's blood pressure has been consistently higher than normal over the past week, coupled with a history of heart disease, elevated blood pressure is extracted as a key risk factor. Based on these key risk factors, an initial parameter set for risk assessment is constructed. Each factor in the parameter set includes a corresponding numerical threshold to assess the patient's health status. For instance, a heart rate above 100 beats per minute may be a high-risk indicator, and blood sugar exceeding a certain threshold may also be a risk factor. Finally, a weighted allocation of these risk factors highlights their impact on mortality risk, ensuring that the role of key factors is fully reflected in the risk assessment. If certain factors show low correlation in historical data, their corresponding weights will be reduced to ensure the accuracy and efficiency of risk assessment parameters.

[0027] After initially constructing the risk assessment parameter set, the parameters are simplified and redundant factors are eliminated. This process is carried out through correlation analysis, with the aim of reducing computational complexity and improving prediction accuracy. By removing factors with high repetition and small impact on mortality risk, the risk assessment model can be optimized, making it more concise and efficient. Only factors that have a significant impact on the patient's mortality risk are retained, thereby providing a more accurate basis for predicting mortality risk. This allows potential health risks to be identified early, helping doctors to develop timely and effective intervention measures based on the risk assessment results.

[0028] Furthermore, based on the individualized health profile, extracting potential risk-related factors also includes: The weight allocation of the initial parameter group for risk assessment can be dynamically adjusted based on real-time monitoring data. The dynamic adjustment adopts an adaptive algorithm, which automatically adjusts the weight of each risk factor in the risk assessment parameter group based on the patient's real-time health data and the trend of health changes within the prediction window, ensuring that the weight allocation is accurately updated as the patient's health condition changes.

[0029] Specifically, by introducing a dynamic adjustment mechanism, the risk assessment model can be adjusted in real time based on new health data to update the weights in the initial parameter set when a patient's health condition changes. This dynamic adjustment mechanism makes the risk assessment more accurate and able to reflect changes in the patient's health status in real time. To achieve the goal of dynamic adjustment, an adaptive algorithm is used. This algorithm can automatically identify which health factors are more important to the risk assessment at a specific time by analyzing the patient's real-time health data and the trend of health changes within the prediction window, and thus assign higher weights to them in the model. For example, if a patient experiences significant health changes at a certain stage (such as large fluctuations in blood pressure), the adaptive algorithm can automatically increase the weights of risk factors related to that fluctuation. By monitoring real-time data, including the collection of physiological data such as blood pressure, heart rate, and body temperature, and comparing it with historical data, the system helps identify trends in health changes through prediction window settings. It automatically adjusts the weighting of risk factors, ensuring that the weight of each health factor is always proportional to its impact on mortality risk prediction. Since each patient's health condition is different, risk assessment should be adjusted based on individual patient characteristics and real-time health changes. The adaptive algorithm automatically updates weights based on real-time data from each patient, ensuring the accuracy of risk assessment at every point in time. This technical solution addresses the issue that traditional static weighting methods may not reflect the true health risk in a timely manner when patient health changes. By dynamically adjusting the weights of the initial parameter group for risk assessment through real-time monitoring data, the risk assessment can be flexibly adjusted according to changes in the patient's health condition, thereby improving the accuracy and real-time nature of personalized risk assessment.

[0030] Step S4: For the risk assessment parameters, analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution; The acquisition of the risk category distribution includes: analyzing the triggering conditions for hypotension risk during dialysis based on the risk assessment parameters; and determining the fluctuation range of key indicators through the analysis. Determine whether the fluctuation of the key indicator exceeds a preset threshold; if the fluctuation of the key indicator exceeds the preset threshold, mark it as a high-risk state; generate a risk category distribution based on the marking result.

[0031] Specifically, risk category distribution is obtained by judging the fluctuations of key indicators. This process analyzes the dynamic changes of various physiological indicators during hemodialysis to identify triggering conditions that may lead to high-risk states, ensuring effective prediction of patients' mortality risk. Risk assessment parameters are analyzed using multi-source physiological data acquired in real time from the dialysis equipment, such as blood pressure and heart rate, to determine the fluctuation range of various physiological indicators during dialysis. In this process, continuous monitoring of key indicators such as blood pressure can identify change patterns related to the risk of hypotension, such as a sharp drop in systolic and diastolic blood pressure. During data processing, preset thresholds are used to judge these fluctuations. When the fluctuation of key indicators exceeds the set range, the state is automatically marked as high-risk. This judgment logic ensures that alarms are triggered in a timely manner when potential acute risks such as hypotension occur, thereby intervening in the patient's health. This process generates a comprehensive risk category distribution map based on the fluctuation range and risk labeling results, reflecting the patient's health status during dialysis. This classification not only helps clinicians quickly identify high-risk patients but also provides data support for the development of subsequent intervention strategies. For example, suppose that during a dialysis session, a patient's systolic blood pressure gradually drops from the normal value to below 90 mmHg and remains at this level for five minutes. The system determines this as a low-risk condition based on a set threshold and notifies clinicians via a risk label. At this point, the system updates the patient's risk category distribution, identifies them as high-risk, and retrieves the corresponding intervention strategy based on this label, such as adjusting the dialysate temperature or slowing down the fluid removal rate.

[0032] The aforementioned technical solution effectively addresses the shortcomings of existing technologies in monitoring the dynamic changes in the health status of dialysis patients, particularly life-threatening emergencies such as hypotension. It enables timely detection of potential problems through real-time data monitoring and dynamic risk assessment, allowing for appropriate interventions before the patient's health deteriorates. By analyzing and distributing different risk categories, the system can provide personalized health management plans for each patient, avoiding the lag inherent in traditional monitoring methods.

[0033] Step S5: Based on the risk category distribution, obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework.

[0034] The framework for personalized intervention includes: For the risk category distribution, an intervention rule base matching the individualized health profile is obtained; adjustment strategies applicable to the current risk category are extracted from the intervention rule base; a personalized intervention plan framework is constructed based on the adjustment strategies; the feasibility of the personalized intervention plan framework is verified to ensure the executability of the strategies; if the adjustment strategies do not match the current health status, the intervention rules are re-selected; finally, the optimized personalized intervention framework is determined.

[0035] Specifically, intervention rules matching individualized health profiles are obtained based on risk category distribution, adjustment strategies are extracted, and a personalized intervention framework is ultimately determined. Throughout the process, risk category distribution is obtained by analyzing patient health data, including various physiological indicators collected from dialysis equipment, such as blood pressure and heart rate. This data, after processing in the aforementioned steps, accurately reflects changes in the patient's health status. Depending on the risk category, the system extracts intervention rules corresponding to each risk category from a pre-established intervention rule base. These intervention rules are closely integrated with the health profile, enabling personalized intervention plans based on the specific health conditions of different patients. For example, when the risk category distribution indicates a high-risk state, such as low blood pressure, the system automatically selects corresponding adjustment strategies from the intervention rule base, such as reducing the fluid removal rate during dialysis or adjusting the composition of the dialysate. These adjustment strategies are further customized based on the patient's specific health profile, such as age, medical history, and lifestyle habits. If the patient has a history of hypertension, the system may recommend reducing sodium intake and increasing exercise, especially low-intensity aerobic exercise. The intervention framework is generated based on real-time feedback from current health data. It not only helps patients improve their current health status but also reduces future health risks. Throughout the process, the intervention framework is optimized and adjusted based on real-time monitoring data to ensure that the strategies consistently match the patient's health status. If certain adjustment strategies are found to be inconsistent with the patient's health status, the system will re-evaluate the intervention rules. For example, if dietary adjustments fail to effectively control blood pressure, the system will readjust the strategy, potentially adding multi-dimensional interventions such as medication or psychological support.

[0036] The above-mentioned technical solution, through dynamic adaptive algorithms and multi-dimensional health data analysis, realizes a highly personalized health management framework, which can effectively address the different risks that patients may encounter during dialysis, ensuring that each patient can receive targeted health interventions, thereby improving the safety and effectiveness of dialysis treatment.

[0037] Furthermore, defining a personalized intervention framework also includes: The intervention rule base automatically learns new intervention strategies from clinical data through machine learning models, updates and expands the rule base, and at the same time, the intervention framework combines real-time health monitoring data and adopts a real-time feedback mechanism to automatically adjust personalized intervention plans according to changes in the patient's health status.

[0038] Specifically, the intervention rule base automatically learns new intervention strategies from clinical data and updates and expands the rule base through a machine learning model. Based on the physiological indicators and health history of dialysis patients, the machine learning model continuously extracts effective intervention strategies from real-time monitoring and historical data, dynamically adjusting and optimizing the rule base. Specifically, the model identifies health change patterns related to patient mortality risk through analysis of a large amount of clinical data. These patterns are reflected in the data during dialysis, such as blood pressure fluctuations and abnormal heart rate. The machine learning model extracts intervention strategies based on these patterns and adds them to the intervention rule base. This process ensures that the intervention rule base remains updated with the continuous addition of new data, effectively addressing the dynamic health conditions of different patients and providing personalized intervention measures. Simultaneously, the intervention framework combines real-time health monitoring data and employs a real-time feedback mechanism to automatically adjust personalized intervention plans based on changes in the patient's health status. By continuously collecting physiological data such as blood pressure and heart rate, the feedback mechanism analyzes the relationship between these data and the health profile in real time, identifying abnormal fluctuations in health status, such as persistently low blood pressure or accelerated heart rate, and automatically adjusting intervention strategies based on these changes. If the current intervention measures fail to effectively control the patient's health status, the system will immediately adjust the treatment plan, such as adjusting the fluid removal rate during dialysis or recommending medication adjustments, to ensure that the intervention plan can flexibly respond to changes in health status. The above technical solution, by combining machine learning and real-time feedback mechanisms, solves the dynamic and personalized problems in predicting the mortality risk of dialysis patients. It can automatically update the intervention rule base and adjust the intervention framework based on the patient's real-time health data and historical trends, making the treatment plan more flexible and precise, effectively improving the accuracy of risk management during dialysis, and reducing the patient's mortality risk.

[0039] This invention also provides a system for predicting the mortality risk of hemodialysis patients based on artificial intelligence, used to implement the above-mentioned method, such as... Figure 2 As shown, the system includes: The construction unit is used to acquire physiological index data from dialysis equipment, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; and extract time-series features based on the cleaned dataset to construct a basic feature set. The generation unit is used to analyze the impact of changes in indicators on health status based on the set of basic features, and generate a personalized health profile. The determination unit is used to extract potential risk-related factors and determine risk assessment parameters based on the individualized health profile; it is also used to obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework based on the risk category distribution. The acquisition unit is used to analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution for the risk assessment parameters.

[0040] The present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0041] In summary, this invention collects multi-source physiological data from dialysis equipment in real time, including indicators such as blood pressure, heart rate, and blood oxygen saturation, and cleans and denoises the data to ensure its accuracy and completeness. This process provides a reliable data foundation for subsequent health assessments, ensuring that all subsequent steps are built upon high-quality real-time data. Based on the cleaned dataset, the system extracts temporal features and constructs a basic feature set, identifying fluctuations and trends in the patient's health status through historical trend analysis. Through the extraction of these features, the system can reflect the patient's physiological changes during dialysis, thus providing a basis for generating personalized health profiles and helping to identify health factors that may lead to mortality risk. By using lag effect analysis to assess the delayed impact of indicator changes on health status, a personalized health status model is constructed. This step enables the system not only to predict immediate health status but also to identify potential health risks in advance, further improving the accuracy of personalized risk assessment. By analyzing the health profile, potential factors related to mortality risk are extracted, and an initial parameter set for risk assessment is constructed based on these factors. By weighting these parameters and simplifying redundant parameters, the accuracy and efficiency of the risk assessment model are ensured, enabling it to accurately reflect patients' health risks in practical applications. Combining the distribution of patients' risk categories, intervention rules matching their health profiles are extracted, and a framework for personalized intervention plans is constructed. Through a real-time feedback mechanism, the system can automatically adjust the intervention plan, ensuring the real-time nature and targeted nature of intervention measures. This step ensures that the intervention strategy is not only based on static data but can also be optimized according to the patient's real-time health status, reducing the risk of patient mortality. The coordination between these steps forms a closed-loop feedback mechanism, allowing personalized health management and intervention strategies to be adjusted in real time according to health changes during dialysis, thereby effectively improving the safety and efficacy of dialysis treatment.

[0042] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the mortality risk of hemodialysis patients based on artificial intelligence, characterized in that, include: Step S1: Obtain physiological indicator data from the dialysis device, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; Based on the cleaned dataset, extract time-series features and construct a basic feature set; Step S2: For the aforementioned set of basic features, analyze the impact of changes in indicators on health status and generate a personalized health profile; Step S3: Based on the individualized health profile, extract potential risk-related factors and determine risk assessment parameters; Step S4: For the risk assessment parameters, analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution; Step S5: Based on the risk category distribution, obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework.

2. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 1, characterized in that, The step of extracting time-series features and constructing a basic feature set based on the cleaned dataset includes: For the cleaned dataset, a historical trend analysis method is used to perform time-series decomposition on the physiological indicator data; through the time-series decomposition, the fluctuation characteristics and trend characteristics of the physiological indicators are extracted; based on the fluctuation characteristics and trend characteristics, key time-series feature parameters are determined; the key time-series feature parameters are integrated into a basic feature set; and the basic feature set is standardized to unify the data format.

3. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 1, characterized in that, Generate personalized health profiles, including: For the aforementioned set of basic features, lag effect analysis is used to assess the delayed impact of any change in an indicator on subsequent health status. The correlation between indicator changes and health status is determined through the lag effect analysis; an individualized health status model is constructed based on the correlation; the individualized health status model is compared with historical data to verify the accuracy of the model; if the prediction deviation of the model exceeds a preset threshold, the analysis parameters are adjusted to optimize the model output; finally, an individualized health profile is generated.

4. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 1, characterized in that, Based on the individualized health profile, potential risk-related factors are extracted, including: Based on the individualized health profile, and combined with prediction window setting technology, potential factors related to mortality risk are analyzed; Through the analysis, key risk influencing factors are extracted; based on these key risk influencing factors, an initial parameter set for risk assessment is constructed; weights are allocated to the initial parameter set to highlight the influence of the main factors; and finally, optimized risk assessment parameters are determined.

5. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 4, characterized in that, Extracting potential risk-related factors also includes: The weight allocation of the initial parameter group for risk assessment can be dynamically adjusted based on real-time monitoring data. The dynamic adjustment adopts an adaptive algorithm, which automatically adjusts the weight of each risk factor in the risk assessment parameter group based on the patient's real-time health data and the health change trend within the prediction window.

6. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 1, characterized in that, Obtaining the risk category distribution includes: For the aforementioned risk assessment parameters, the triggering conditions for hypotension risk during dialysis are analyzed; through the analysis, the fluctuation range of key indicators is determined; it is determined whether the fluctuation of the key indicators exceeds a preset threshold; if the fluctuation of the key indicators exceeds the preset threshold, it is marked as a high-risk state; based on the marking results, a risk category distribution is generated.

7. The method for predicting the mortality risk of hemodialysis patients based on artificial intelligence as described in claim 1, characterized in that, Define a personalized intervention framework, including: For the risk category distribution, an intervention rule base matching the individualized health profile is obtained; adjustment strategies applicable to the current risk category are extracted from the intervention rule base; a personalized intervention plan framework is constructed based on the adjustment strategies; the feasibility of the personalized intervention plan framework is verified to ensure the executability of the strategies; if the adjustment strategies do not match the current health status, the intervention rules are re-selected; finally, the optimized personalized intervention framework is determined.

8. The method as described in claim 7, characterized in that, Defining a personalized intervention framework also includes: The intervention rule base automatically learns new intervention strategies from clinical data through machine learning models, updates and expands the rule base, and at the same time, the intervention framework combines real-time health monitoring data and adopts a real-time feedback mechanism to automatically adjust personalized intervention plans according to changes in the patient's health status.

9. A system for predicting the mortality risk of hemodialysis patients based on artificial intelligence, used to implement the method as described in any one of claims 1-8, characterized in that, The system includes: The construction unit is used to acquire physiological index data from dialysis equipment, construct a multi-source dataset, merge and process it to obtain a cleaned dataset; and extract time-series features based on the cleaned dataset to construct a basic feature set. The generation unit is used to analyze the impact of changes in indicators on health status based on the set of basic features, and generate a personalized health profile. The determination unit is used to extract potential risk-related factors and determine risk assessment parameters based on the individualized health profile; it is also used to obtain matching intervention rules, extract adjustment strategies, and determine a personalized intervention framework based on the risk category distribution. The acquisition unit is used to analyze the risk triggering conditions, determine the fluctuation of indicators, and obtain the risk category distribution for the risk assessment parameters.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the method as described in any one of claims 1-8.

Images