Early-stage diabetes mellitus reverse treatment evaluation system based on machine learning

Abstract

The invention relates to the technical field of medical data processing, in particular to an early diabetes reverse treatment evaluation system based on machine learning. The system is used for performing multi-source diabetes mellitus index analysis processing on early-stage diabetes mellitus medical archive data of a patient to generate multi-source diabetes mellitus index analysis data; performing empowerment matrix analysis of diabetes mellitus index characteristics based on the multi-source diabetes mellitus index analysis data to generate an empowerment diabetes mellitus index characteristic matrix; performing personalized diabetes state trend analysis according to the empowerment diabetes index feature matrix to generate personalized diabetes state trend data; obtaining an early diabetes reverse treatment intervention strategy; based on the empowerment diabetes mellitus index feature matrix and the personalized diabetes mellitus state trend data, performing early-stage diabetes mellitus reverse treatment evaluation on the early-stage diabetes mellitus reverse treatment intervention benefit simulation data, and generating early-stage diabetes mellitus reverse treatment evaluation data. According to the invention, accurate diabetes reverse treatment evaluation is realized aiming at the diabetes index of the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to an early diabetes reversal treatment evaluation system based on machine learning. Background Art

[0002] Early diabetes refers to the initial stage of the development of diabetes, usually including prediabetes and early diagnosed type 2 diabetes. At this stage, the patient's blood sugar level is higher than the normal range but has not yet reached the diagnostic criteria for severe diabetes, usually manifested as a mild increase in fasting blood sugar or impaired glucose tolerance. Typical characteristics of early diabetes include increased insulin resistance and mild damage to pancreatic β-cell function, but these pathological changes are still in a reversible stage. Advanced artificial intelligence technology and data analysis methods are used to evaluate the reversal treatment of early diabetes, and multi-dimensional analysis and dynamic monitoring of patients' health data are performed to evaluate the effect and progress of their diabetes reversal treatment. Combined with machine learning models and big data technology, accurate quantification of treatment effects and risk warnings are performed to improve the efficiency and success rate of treatment. However, the existing early diabetes reversal treatment evaluation technology is difficult to comprehensively evaluate the diabetic status of early diabetic patients, lacks in-depth modeling of individual diabetic indicators of patients, and the evaluation results are somewhat one-sided and cannot meet the needs of individualized treatment. It is also difficult to capture the dynamic evolution of the diabetic status and long-term trends, and cannot provide a reliable basis for accurate early diabetes reversal treatment evaluation and the formulation of optimized intervention strategies. Summary of the invention

[0003] Based on this, the present invention provides an early diabetes reversal treatment evaluation system based on machine learning to solve at least one of the above technical problems.

[0004] To achieve the above objectives, an early diabetes reversal treatment evaluation system based on machine learning includes the following modules: The diabetes index parsing and weighting module is used to obtain the patient's early diabetes medical file data; perform multi-source diabetes index parsing on the patient's early diabetes medical file data to generate multi-source diabetes index parsing data; perform weighting matrix analysis on diabetes index characteristics based on the multi-source diabetes index parsing data to generate a weighted diabetes index characteristic matrix; The diabetes state trend analysis module is used to perform diabetes state analysis based on the weighted diabetes indicator feature matrix to generate diabetes state data; perform personalized diabetes state trend analysis based on the diabetes state data to generate personalized diabetes state trend data; The diabetes reversal potential assessment module is used to conduct quantitative analysis of the diabetes reversal impact of indicators based on the preset diabetes reversal factors and the weighted diabetes indicator feature matrix, and generate indicator diabetes reversal impact quantitative data; conduct diabetes reversal potential assessment modeling processing based on the indicator diabetes reversal impact quantitative data, and generate a diabetes reversal potential assessment model; The early diabetes reversal treatment evaluation module is used to obtain the early diabetes reversal treatment intervention strategy; use the diabetes reversal potential evaluation model to perform early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategy, and generate early diabetes reversal treatment intervention benefit simulation data; perform early diabetes reversal treatment evaluation processing on the early diabetes reversal treatment intervention benefit simulation data according to personalized diabetes state trend data, and generate early diabetes reversal treatment evaluation data; perform reversal treatment intervention strategy feedback optimization processing on the early diabetes reversal treatment intervention strategy based on personalized diabetes state trend data and early diabetes reversal treatment evaluation data, and generate an optimized early diabetes reversal treatment intervention strategy.

[0005] Furthermore, the diabetes indicator parsing and weighting module includes the following functions: Obtain the patient's early diabetes medical record data; Perform multi-source diabetes index analysis on the patient's early diabetes medical record data to generate multi-source diabetes index analysis data; Performing multi-source diabetes indicator feature screening processing on the multi-source diabetes indicator analysis data to generate multi-source diabetes indicator feature data; A weighted matrix analysis of diabetes indicator characteristics is performed based on multi-source diabetes indicator characteristic data to generate a weighted diabetes indicator characteristic matrix.

[0006] Furthermore, the weighted matrix analysis of diabetes indicator characteristics based on multi-source diabetes indicator characteristic data includes: Based on multi-source diabetes indicator characteristic data, the state impact characterization capability of diabetes indicators is analyzed to generate state impact characterization capability data of diabetes indicators; Perform a weight parameter analysis of the diabetes indicator characteristics according to the diabetes indicator status impact characterization capability data to obtain the diabetes indicator characteristic weight parameter; The weighted matrix of diabetes indicator characteristics is designed through multi-source diabetes indicator characteristic data and corresponding diabetes indicator characteristic weight parameters to generate a weighted diabetes indicator characteristic matrix.

[0007] Furthermore, the state impact characterization capability analysis of diabetes indicators based on multi-source diabetes indicator characteristic data includes: Performing diabetes indicator characteristic disturbance processing on multi-source diabetes indicator characteristic data using preset diabetes indicator interference parameters to generate disturbed diabetes indicator characteristic data; Performing diabetes state change analysis of diabetes indicator disturbance based on the perturbed diabetes indicator characteristic data to generate diabetes indicator disturbance state change data; The state impact characterization capability of the diabetes indicator is analyzed based on the disturbance state change data of the diabetes indicator, and the state impact characterization capability data of the diabetes indicator is generated.

[0008] Furthermore, the diabetes state change analysis of diabetes indicator disturbance based on the perturbed diabetes indicator characteristic data includes: Performing a disturbance change analysis of the diabetes indicator characteristic according to the disturbed diabetes indicator characteristic data to generate the diabetes indicator characteristic disturbance change data; Obtain the patient's historical diabetes indicator status assessment data; Based on the preset XGBoost algorithm and historical diabetes indicator status evaluation data, a mapping relationship between indicator disturbance and diabetes status change is established to obtain an indicator disturbance diabetes status evaluation model; The indicator perturbation diabetes state assessment model is used to analyze the diabetes state change of the diabetes indicator disturbance on the diabetes indicator characteristic perturbation change data to generate diabetes indicator disturbance state change data.

[0009] Furthermore, the diabetes status trend analysis module includes the following functions: Perform diabetes status analysis based on the weighted diabetes indicator feature matrix to generate diabetes status data; Performing diabetes state space definition processing based on the diabetes state data to generate diabetes state space data; Performing diabetes time series state data analysis on the diabetes state data to generate diabetes time series state data; Based on the diabetes state spatial data and the diabetes time series state data, a transfer characteristic trend analysis of the diabetes time series state is performed to generate diabetes time series state transfer characteristic trend data; Performing diabetes time series state periodic change pattern analysis based on diabetes time series state data to generate diabetes state periodic change pattern data; Based on the diabetes time series state transition characteristic trend data and the diabetes state periodic change pattern data, personalized diabetes state trend analysis is performed to generate personalized diabetes state trend data.

[0010] Furthermore, the transfer characteristic trend analysis of the diabetes time series state is performed based on the diabetes state spatial data and the diabetes time series state data: Performing characteristic frequency distribution analysis of diabetes time series state data on diabetes time series state data through diabetes state space data to generate diabetes time series state characteristic frequency distribution data; The transfer characteristic trend of the diabetes time series state is analyzed according to the diabetes time series state characteristic frequency distribution data to generate the diabetes time series state transfer characteristic trend data.

[0011] Furthermore, the transfer characteristic trend analysis of the diabetes time series state according to the diabetes time series state characteristic frequency distribution data includes: Performing a diabetes time series state transition feature analysis based on diabetes time series state feature frequency distribution data to generate diabetes time series state transition feature data; The preset Markov chain model is used to perform a transfer characteristic trend analysis of the diabetes time series state transfer characteristic data to generate the diabetes time series state transfer characteristic trend data.

[0012] Furthermore, the diabetes reversal potential assessment module includes the following functions: According to the preset diabetes reversal factors and the multi-source diabetes indicator analysis data of the weighted diabetes indicator characteristic matrix, the causal relationship analysis between each indicator and the diabetes reversal factor is performed to generate the indicator diabetes reversal factor causal relationship data; According to the diabetes indicator feature weight parameters of the weighted diabetes indicator feature matrix and the causal relationship data of diabetes reversal factors, the diabetes reversal impact quantitative analysis of the indicator is performed to generate the diabetes reversal impact quantitative data of the indicator; The diabetes reversal potential assessment modeling is performed based on the quantitative data of the diabetes reversal impact indicators to generate a diabetes reversal potential assessment model.

[0013] Furthermore, the early diabetes reversal treatment assessment module includes the following functions: Obtain early intervention strategies for diabetes reversal treatment; Use the diabetes reversal potential assessment model to conduct early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategy, and generate early diabetes reversal treatment intervention benefit simulation data; Perform early diabetes reversal treatment evaluation processing on early diabetes reversal treatment intervention benefit simulation data according to personalized diabetes status trend data to generate early diabetes reversal treatment evaluation data; Conduct early diabetes reversal treatment demand analysis based on personalized diabetes status trend data to generate early diabetes reversal treatment demand data; Performing an early diabetes reversal treatment feedback optimization coefficient analysis on the early diabetes reversal treatment evaluation data through the early diabetes reversal treatment demand data, and generating an early diabetes reversal treatment feedback optimization coefficient; According to the early diabetes reversal treatment feedback optimization coefficient, the early diabetes reversal treatment intervention strategy is subjected to reversal treatment intervention strategy feedback optimization processing to generate an optimized early diabetes reversal treatment intervention strategy.

[0014] The beneficial effect of the present application is that the diabetes index analysis and empowerment module can integrate multi-source diabetes indicators and generate accurate diabetes index analysis data by deeply analyzing the patient's early diabetes medical archive data. Through the analysis and screening of multi-source data, the most representative diabetes index features are extracted, thereby providing a basis for subsequent weighted matrix analysis, and the diabetes index feature data is disturbed by interference parameters, and the model established by the XGBoost algorithm is combined to analyze the changes in diabetes state, and the influence of diabetes index features on disease changes is more accurately quantified, which can not only improve the weight calculation accuracy of diabetes index features, but also help accurately identify the potential for diabetes reversal, and then optimize the treatment plan and provide personalized treatment plan feedback. Through the design of the weighted matrix, the diabetes index analysis and empowerment module can provide a more scientific basis for diabetes reversal treatment, ensure the reasonable weight distribution of diabetes indicators, and effectively integrate medical data from different sources, so that the treatment effect evaluation is more accurate and reliable, significantly improve the intelligence and automation level of diabetes reversal treatment evaluation, and help to make more scientific and accurate decisions during the treatment process, thereby improving the success rate of early diabetes reversal. The diabetes state trend analysis module can accurately reflect the dynamic changes of diabetes by conducting a comprehensive diabetes state analysis on the weighted diabetes indicator feature matrix. The definition processing of the diabetes state space and the time series state analysis provide patients with personalized diabetes state trend predictions. Identify the periodic changes of the diabetes state and deeply analyze the fluctuations of diabetes indicators in different time periods, so as to provide a reference for clinical treatment. It also combines the frequency distribution analysis and transition feature trend analysis of the diabetes time series state, and uses the Markov chain model to conduct in-depth research on the state transition trend, which can effectively predict the future changes of the diabetes state, not only helping to identify the risks that arise during the treatment process, but also adjusting the treatment strategy according to the patient's personalized needs, making the treatment process more flexible and adaptable. Through in-depth analysis of the periodic change pattern of the diabetes time series state, the diabetes state trend analysis module can provide a long-term and comprehensive diabetes management plan, provide a more accurate prediction of the patient's diabetes state, and predict the potential change trend before treatment. The diabetes reversal potential assessment module can comprehensively and accurately evaluate the potential impact of different indicators on diabetes reversal by conducting a quantitative analysis of the impact of diabetes reversal based on the preset diabetes reversal factors and the weighted diabetes indicator feature matrix. Through in-depth analysis of these data, the module can generate quantitative data on the impact of diabetes reversal indicators, help identify which factors play a key role in the process of diabetes reversal, and reveal which diabetes indicators and factors have a significant impact on reversal potential, thereby providing patients with personalized treatment path recommendations.Through the modeling and processing of diabetes reversal potential assessment, the module further generates a diabetes reversal potential assessment model, which can not only objectively evaluate the benefits of various diabetes indicators for diabetes reversal, but also dynamically adjust according to the specific conditions of the patient, making the treatment process more accurate and personalized. This module can effectively predict the treatment effect of patients, avoid ineffective treatment, save medical resources, and improve the success rate of treatment. The early diabetes reversal treatment assessment module obtains the early diabetes reversal treatment intervention strategy and combines the diabetes reversal potential assessment model to simulate the intervention benefit analysis to predict the effect of different treatment intervention strategies in advance. This simulation analysis can provide patients with accurate treatment expectations, thereby improving the efficiency and accuracy of treatment. Based on personalized diabetes status trend data, the early diabetes reversal treatment assessment module can further evaluate the intervention benefit and generate early diabetes reversal treatment assessment data. By combining the individual differences of patients and the dynamic changes of diabetes, it provides a more targeted and personalized treatment plan. In addition, based on personalized diabetes status trend data and early treatment evaluation data, the early diabetes reversal treatment assessment module can feedback optimize the treatment strategy. Through the feedback optimization processing of the reversal treatment intervention strategy, the intervention strategy is dynamically adjusted during the treatment process to ensure that the treatment plan matches the patient's actual condition. This optimization process can continuously improve the optimization of intelligent technology for early diabetes reversal treatment intervention and ensure the best effect of diabetes reversal treatment. Overall, the early treatment assessment module significantly improves the accuracy and effectiveness of diabetes treatment, provides strong decision-making support for clinicians, and ensures flexibility and adaptability during the treatment process.

[0015] Therefore, the existing early diabetes reversal treatment evaluation system based on machine learning can comprehensively evaluate the diabetic status of early diabetes patients based on intelligence. By introducing machine learning technology, it can extract key information from massive medical data, dynamically track the progression of patients' diseases, and deeply model individual diabetes indicators of patients, thereby quantifying the potential and treatment effects of diabetes reversal to meet individualized treatment needs. And by analyzing the characteristic trends and periodic change patterns of diabetes time series state transitions, it can capture the dynamic evolution laws and long-term trends of diabetes states, as well as simulate and analyze early diabetes reversal treatment intervention strategies and optimize feedback, thereby providing a reliable basis for accurate early diabetes reversal treatment evaluation and the formulation of optimized intervention strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a module flow diagram of an early diabetes reversal treatment evaluation system based on machine learning of the present invention; Figure 2 for Figure 1 Detailed implementation functional flow chart of diabetes indicator parsing and empowerment module; Figure 3 for Figure 1 Detailed implementation functional flow chart of diabetes status trend analysis module; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0017] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0018] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0019] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0020] To achieve this, please refer to Figures 1 to 3 The present invention provides an early diabetes reversal treatment evaluation system based on machine learning. In the embodiments of the present invention, please refer to Figure 1 FIG. 1 is a schematic diagram of a module flow of an early diabetes reversal treatment evaluation system based on machine learning according to the present invention. The early diabetes reversal treatment evaluation system based on machine learning includes the following modules: S1: Diabetes index parsing and weighting module, used to obtain the patient's early diabetes medical record data; perform multi-source diabetes index parsing on the patient's early diabetes medical record data to generate multi-source diabetes index parsing data; perform weighting matrix analysis on diabetes index characteristics based on the multi-source diabetes index parsing data to generate a weighted diabetes index characteristic matrix; In the embodiment of the present invention, the process of obtaining the patient's early diabetes medical record data needs to rely on a unified medical data collection platform. By establishing a connection with the hospital's electronic health record system (EHR) interface, the patient's medical record data is extracted in batches. Including calling the patient's basic diagnosis and treatment records, laboratory test results, physical examination reports, and historical medical course information. Through standardized data processing procedures, ensure that the extracted medical record data covers key indicators such as the patient's blood sugar level, glycosylated hemoglobin (HbA1c), blood lipids, liver and kidney function, as well as related information such as family history of diabetes and lifestyle records. Multi-source diabetes indicator analysis and processing is based on data fusion and feature engineering methods, and a unified algorithm framework is used. The different source data in the patient's medical file are analyzed through the feature decomposition model, including numerical indicators (such as blood sugar, HbA1c, etc.), classification indicators (such as gender, diabetes classification) and time series data (such as continuous monitoring of blood sugar values). For numerical data, the data range is normalized using a standardization algorithm to eliminate the impact of dimensional differences; for categorical data, the One-Hot Encoding technique is used to convert it into numerical features that can be used for machine learning algorithms; for time series data, the time window sliding and segmented aggregation analysis methods are used to extract key statistical features (such as mean, maximum, change trend, etc.). After the analysis is completed, structured multi-source diabetes indicator analysis data is generated to provide input for subsequent feature screening. The multi-source diabetes indicator feature screening process is based on the machine learning feature selection algorithm, and the XGBoost and LASSO regression models are used to achieve feature screening. The multi-source diabetes indicator analysis data is input into the XGBoost model, and the indicators with significant predictive ability for the reversal state of diabetes are screened out through feature importance score calculation; the screening results are further input into the LASSO regression model, and the redundant features are further compressed through the regularization constraint mechanism to ensure that the screened indicator feature set has the best information representativeness and prediction effect. In the feature screening process, the performance indicators of the feature set (such as AUC value and prediction accuracy) are evaluated through a cross-validation strategy to ensure that the selected features meet the actual application requirements. The generated multi-source diabetes indicator characteristic data contains highly correlated characteristic variables, which provides an accurate input basis for weighted matrix analysis. According to the screened multi-source diabetes indicator characteristic data, the importance comparison matrix of characteristic indicators is constructed to quantify the relative weights of different indicators in the assessment of diabetes status. The weight value of the comparison matrix is ​​statistically determined by the impact of the relevant data of historical indicators on the diabetes status to ensure the scientific nature of the weight distribution. Using the fuzzy comprehensive evaluation method, the importance weights of different indicators are combined through weighted average calculation to generate a weighted diabetes indicator characteristic matrix. The analysis results can clearly characterize the degree of influence of each diabetes indicator characteristic on the disease status, and provide data support for the subsequent status trend analysis module.

[0021] S2: Diabetes state trend analysis module, used to perform diabetes state analysis based on the weighted diabetes indicator feature matrix to generate diabetes state data; perform personalized diabetes state trend analysis based on the diabetes state data to generate personalized diabetes state trend data; In an embodiment of the present invention, the diabetes state analysis helps identify the patient's diabetes control level and potential risks by conducting an in-depth analysis of the weighted diabetes indicator feature matrix. Based on the weighted diabetes indicator feature matrix, a multivariate regression analysis method is used to comprehensively evaluate multiple diabetes-related indicators (such as blood sugar, blood pressure, BMI, etc.) and establish a mathematical model. The weight assigned to each indicator feature is determined by the aforementioned weighting process to ensure that the impact of each indicator on the diabetes state is reasonably reflected. The model predicts the patient's current diabetes control level (such as good control, mild out of control, severe out of control, etc.). This analysis helps generate diabetes state data, including key indicators such as the patient's diabetes classification and risk score, to provide a basis for the selection of subsequent treatment plans. The diabetes state space definition process is used to convert the multidimensional state information of diabetes into a standardized mathematical model, and define different state space dimensions according to the diabetes state data (including blood sugar level, insulin use, weight change, etc.), such as blood sugar control level, body mass index, frequency of complications, etc. Use data clustering technology (such as K-means clustering) to classify similar states into the same spatial area to generate diabetes state space data. Diabetes time series state data analysis is used to observe the state change trend of diabetic patients over a period of time. The diabetes state data is organized into time series data in chronological order. Time series data are modeled using time series data analysis methods, such as the autoregressive moving average model (ARMA) or the long short-term memory neural network (LSTM). The model identifies the law of changes in diabetes state, such as whether the diabetes state is gradually worsening or under control. The generated diabetes time series state data not only reflects the diabetes control trend of patients within a certain period of time, but also predicts the future development trend of diabetes state, providing data support for the evaluation of treatment effects and the selection of intervention timing. The diabetes time series state transition feature trend analysis aims to identify the transition trend of the state of diabetic patients from one stage to another. Based on the diabetes state space data and diabetes time series state data, the Markov chain model is used to describe the transition process of diabetes state from one state to another. By calculating the transition probability between each diabetes state, the transition frequency and trend between different states are analyzed, such as the probability of turning from mild diabetes out of control to moderate out of control. According to the results of the transition feature trend analysis, the diabetes time series state transition feature trend data is generated. The diabetes time series state periodic change pattern analysis is used to discover the periodic fluctuations of diabetes state over a period of time. Perform Fourier transform or wavelet transform on the diabetes time series data, convert the time series data from the time domain to the frequency domain, and extract the periodic components. Use periodic analysis methods, such as periodic autocorrelation analysis, to identify the periodic pattern of diabetes state changes. The generated diabetes state periodic change pattern data shows the law of state changes in diabetic patients, helping to determine the intervention measures required by patients in different time periods and optimize treatment plans.Personalized diabetes state trend analysis combines diabetes time series state transition feature trend data and diabetes state periodic change pattern data. The diabetes time series state transition feature trend data and periodic change pattern data are combined, and based on individual historical data, a personalized diabetes state trend analysis model is established using machine learning algorithms (such as support vector machines, random forests, etc.). This model can not only analyze the patient's state change trend in different time periods, but also predict the future diabetes state based on the patient's individual characteristics (such as age, gender, lifestyle habits, etc.). The generated personalized diabetes state trend data provides diabetic patients with targeted health management plans and treatment recommendations, making intervention measures more accurate and the treatment effect more significant.

[0022] S3: Diabetes reversal potential assessment module, used to perform quantitative analysis of the diabetes reversal impact of indicators based on preset diabetes reversal factors and the weighted diabetes indicator feature matrix, and generate indicator diabetes reversal impact quantitative data; perform diabetes reversal potential assessment modeling processing based on the indicator diabetes reversal impact quantitative data, and generate a diabetes reversal potential assessment model; In the embodiment of the present invention, in the process of evaluating the potential for diabetes reversal, a causal relationship analysis is first performed based on the preset diabetes reversal factors and the multi-source diabetes indicator analysis data in the weighted diabetes indicator feature matrix to identify how the key indicators of diabetes affect the potential for diabetes reversal. Select factors related to diabetes reversal, such as blood sugar control level, islet function, weight management, and lifestyle, and extract these diabetes indicator features from multi-source data. Statistical analysis methods (such as correlation analysis, regression analysis, or causal inference algorithm) are used to analyze the relationship between each diabetes indicator and the reversal factor to generate causal relationship data between the diabetes indicator and the reversal factor. A quantitative analysis of the impact of diabetes reversal is performed based on the diabetes indicator feature weight parameters in the weighted diabetes indicator feature matrix and the causal relationship data of the diabetes reversal factor. The diabetes indicator feature weight parameters reflect the importance of each indicator to the diabetic state. According to these weights, the influence of different diabetes indicators is weighted. Combined with the causal relationship data, determine which indicators have a greater impact on diabetes reversal. Using mathematical modeling methods, such as multiple regression analysis, structural equation modeling, etc., the relationship between each indicator and the potential for diabetes reversal is quantified, the degree of influence of each diabetes indicator on diabetes reversal is obtained, and the quantitative data of the indicator diabetes reversal impact is generated. Based on the generated indicator diabetes reversal impact quantitative data, diabetes reversal potential assessment modeling is performed. The goal of this process is to build a model for analyzing the impact of each indicator on diabetes reversal potential, which can evaluate the potential of diabetes reversal based on the patient's specific diabetes indicators and reversal impact quantitative data. By using machine learning algorithms, such as support vector machines (SVMs), random forests, etc., a multivariate regression model or classification model is constructed to model the relationship between each diabetes indicator and reversal potential. When training the model, a large amount of historical data related to diabetes indicators and diabetes reversal treatment is used, combined with different indicator characteristics and reversal impact quantitative data of diabetic patients to improve the accuracy and predictive ability of the model. The model can evaluate the potential of diabetes reversal based on the patient's specific situation and generate a diabetes reversal potential assessment model. This model can not only help doctors better understand the patient's condition, but also provide decision support for the formulation of personalized treatment plans.

[0023] S4: Early diabetes reversal treatment evaluation module, used to obtain early diabetes reversal treatment intervention strategies; use the diabetes reversal potential evaluation model to perform early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategies, and generate early diabetes reversal treatment intervention benefit simulation data; perform early diabetes reversal treatment evaluation processing on the early diabetes reversal treatment intervention benefit simulation data based on personalized diabetes state trend data, and generate early diabetes reversal treatment evaluation data; perform reversal treatment intervention strategy feedback optimization processing on the early diabetes reversal treatment intervention strategies based on personalized diabetes state trend data and early diabetes reversal treatment evaluation data, and generate optimized early diabetes reversal treatment intervention strategies.

[0024] In the embodiment of the present invention, in the diabetes reversal treatment evaluation module, it is first necessary to obtain the early diabetes reversal treatment intervention strategy. These intervention strategies are pre-set medical intervention strategies for patients, including diet adjustment, exercise intervention, drug treatment, etc. After obtaining the early diabetes reversal treatment intervention strategy, these strategies are simulated and analyzed using the diabetes reversal potential evaluation model. By inputting various diabetes indicators of the patient and their corresponding diabetes reversal potential evaluation data, the potential evaluation model established by machine learning algorithms (such as neural networks, etc.) is used to predict the effects of different intervention strategies on diabetes reversal. The model will generate intervention benefit simulation data by simulating the implementation effects of different treatment plans. These data will include the effects of each intervention strategy in different time periods, such as changes in blood sugar levels, improvements in pancreatic islet function, changes in weight, etc. This process is based on a large amount of historical data and individualized patient data to ensure the accuracy of the simulation analysis. After obtaining the early diabetes reversal treatment intervention benefit simulation data, the intervention benefit is evaluated and processed according to the personalized diabetes state trend data. The personalized diabetes state trend data reflects the current and historical diabetes development trends of the patient, including blood sugar control levels, changes in pancreatic islet function, living habits, etc. By combining these data with intervention benefit simulation data, regression analysis, classification analysis and other methods are used to evaluate the effect of each intervention strategy in different patient groups. The long-term effect of treatment is predicted based on the trend of blood sugar level changes, and the specific efficacy evaluation data of each strategy is obtained through the model. Based on personalized diabetes status trend data, early diabetes reversal treatment demand analysis is carried out. By analyzing the patient's diabetes status at various time points and combining historical treatment data, it is evaluated whether the patient currently needs additional treatment intervention. By using time series data analysis methods, it is identified whether the patient's diabetes status is worsening and whether it is necessary to strengthen intervention. The patient's condition deterioration trend will become an important basis for formulating treatment needs, and effectively identify the treatment needs that arise in different patients during the treatment process. After obtaining the early diabetes reversal treatment demand data, the early diabetes reversal treatment evaluation data is analyzed for feedback optimization coefficients. By using feedback control theory or optimization algorithms to process the evaluation data, it is determined which factors (such as intervention strategies, treatment intensity, treatment cycle, etc.) affect the patient's potential for diabetes reversal. Using model analysis, an optimization coefficient is generated to reflect how to adjust the treatment plan to improve the efficacy in specific situations. The optimization coefficient analysis can combine the treatment strategy with the specific needs of the patient to ensure the flexibility and accuracy of the treatment plan in a dynamic environment. After the optimization coefficients are generated, the early diabetes reversal treatment intervention strategy is fed back and optimized through these optimization coefficients. Based on the feedback optimization coefficients, mathematical optimization methods (such as optimization algorithms) are used to adjust the treatment intervention plan, including adjusting the drug dosage, exercise plan, diet plan and other treatment parameters according to the patient's personalized treatment needs.The optimized intervention strategy will improve the effect of diabetes reversal, reduce treatment risks, ensure the continuity and effectiveness of treatment, and generate an optimized early diabetes reversal treatment intervention strategy. This strategy provides patients with a personalized, dynamically adjusted treatment plan and improves the success rate of reversal treatment.

[0025] Further, as an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed functional flow chart of the diabetes index parsing and weighting module. In this embodiment, the diabetes index parsing and weighting module includes the following functions: S11: Obtain the patient's early diabetes medical record data; In an embodiment of the present invention, the process of obtaining the patient's early diabetes medical archive data needs to rely on a unified medical data collection platform. By establishing a connection with the hospital's electronic health record system (EHR) interface, the patient's medical record data is extracted in batches. This includes calling the patient's basic diagnosis and treatment records, laboratory test results, physical examination reports, and historical medical information. Through standardized data processing procedures, it is ensured that the extracted medical archive data covers key indicators such as the patient's blood sugar level, glycosylated hemoglobin (HbA1c), blood lipids, liver and kidney function, as well as related information such as family history of diabetes and lifestyle records. Using data quality control mechanisms (such as missing value filling, outlier detection and correction), the collected archive data is reviewed for completeness and accuracy to ensure that the generated medical archive data meets the analysis requirements.

[0026] S12: performing multi-source diabetes indicator analysis processing on the patient's early diabetes medical record data to generate multi-source diabetes indicator analysis data; In an embodiment of the present invention, the multi-source diabetes index analysis and processing is based on data fusion and feature engineering methods, and is performed using a unified algorithm framework. The data from different sources in the patient's medical records are analyzed through a feature decomposition model, including numerical indicators (such as blood sugar, HbA1c, etc.), categorical indicators (such as gender, diabetes classification), and time series data (such as continuous monitoring of blood sugar values). For numerical data, a standardization algorithm is used to normalize the data range to eliminate the impact of dimensional differences; for categorical data, the One-Hot Encoding technology is used to convert it into numerical features that can be used for machine learning algorithms; for time series data, the time window sliding and segmented aggregation analysis methods are used to extract key statistical features (such as mean, maximum value, change trend, etc.). After the analysis is completed, structured multi-source diabetes index analysis data is generated to provide input for subsequent feature screening.

[0027] S13: performing multi-source diabetes indicator feature screening processing on the multi-source diabetes indicator analysis data to generate multi-source diabetes indicator feature data; In an embodiment of the present invention, the feature screening process of multi-source diabetes indicators is performed based on a machine learning feature selection algorithm, and feature screening is achieved using XGBoost and LASSO regression models. The multi-source diabetes indicator analysis data is input into the XGBoost model, and the indicators with significant predictive ability for the diabetes reversal state are screened out through feature importance score calculation; the screening results are further input into the LASSO regression model, and the redundant features are further compressed through the regularization constraint mechanism to ensure that the screened indicator feature set has the best information representativeness and prediction effect. In the feature screening process, the performance indicators of the feature set (such as AUC value, prediction accuracy) are evaluated through a cross-validation strategy to ensure that the selected features meet the actual application requirements. The generated multi-source diabetes indicator feature data contains highly correlated feature variables, which provides an accurate input basis for weighted matrix analysis.

[0028] S14: Performing a weighted matrix analysis of diabetes indicator characteristics based on multi-source diabetes indicator characteristic data to generate a weighted diabetes indicator characteristic matrix.

[0029] In an embodiment of the present invention, an importance comparison matrix of characteristic indicators is constructed based on the screened multi-source diabetes indicator characteristic data to quantify the relative weights of different indicators in the assessment of diabetes status. The weight values ​​of the comparison matrix are statistically determined by the effect of the relevant data of historical indicators on the diabetes status to ensure the scientific nature of the weight distribution. The fuzzy comprehensive evaluation method is used to combine the importance weights of different indicators through weighted average calculation to generate a weighted diabetes indicator characteristic matrix. The analysis results can clearly characterize the degree of influence of each diabetes indicator characteristic on the disease status and provide data support for the subsequent status trend analysis module.

[0030] Furthermore, the weighted matrix analysis of diabetes indicator characteristics based on multi-source diabetes indicator characteristic data includes: Based on multi-source diabetes indicator characteristic data, the state impact characterization capability of diabetes indicators is analyzed to generate state impact characterization capability data of diabetes indicators; Perform a weight parameter analysis of the diabetes indicator characteristics according to the diabetes indicator status impact characterization capability data to obtain the diabetes indicator characteristic weight parameter; The weighted matrix of diabetes indicator characteristics is designed through multi-source diabetes indicator characteristic data and corresponding diabetes indicator characteristic weight parameters to generate a weighted diabetes indicator characteristic matrix.

[0031] In an embodiment of the present invention, the multi-source diabetes indicator feature data of the patient is used to extract the mapping relationship between the indicator and the change of diabetes state through the correlation analysis of the historical diabetes state annotation data. The random forest feature importance algorithm is used to calculate the contribution of each indicator to the diabetes state, and the Kendall correlation coefficient is used to further verify the correlation strength between the indicator and the state change. According to the analysis results, a scoring model for the characterization ability of the feature on the diabetes state is constructed. The model normalizes the characterization ability scores of different indicators to the [0,1] interval through normalization calculation, generates the state impact characterization ability data of the diabetes indicator, ensures that the data is clear and convenient for subsequent weight analysis. Using the state impact characterization ability data of the diabetes indicator, the hierarchical analysis method (AHP) is used to construct a pairwise comparison matrix of each indicator to determine the relative importance of each pair of indicators in the assessment of diabetes state. The matrix eigenvalue decomposition method is used to calculate the weight of each indicator to ensure that the weight distribution meets the consistency test requirements. The weighted regression model is introduced to verify the actual performance of the weight parameters of each indicator in predicting the change of diabetes state through historical patient data, and optimize the initial weight distribution. The final generated diabetes indicator feature weight parameters can accurately reflect the role of each feature in the assessment of diabetes status, providing a basis for the design of the weight matrix. The multi-source diabetes indicator feature data and weight parameters are combined, and the data features of each indicator and its corresponding weight are matrix-operated to generate the initial weight matrix. The weight matrix is ​​processed by principal component analysis (PCA) to reduce the dimension, remove redundant information, and enhance the matrix's ability to express state analysis. Subsequently, the fuzzy analytic hierarchy process (Fuzzy AHP) is used to optimize the initial weight matrix, and the weights of key indicators are further adjusted to make the matrix more in line with the actual needs of diabetes reversal assessment. The final output weighted diabetes indicator feature matrix achieves a balance between the sparsity of the matrix and the ability to express information, and can be directly used in the subsequent diabetes status trend analysis module.

[0032] Furthermore, the state impact characterization capability analysis of diabetes indicators based on multi-source diabetes indicator characteristic data includes: Performing diabetes indicator characteristic disturbance processing on multi-source diabetes indicator characteristic data using preset diabetes indicator interference parameters to generate disturbed diabetes indicator characteristic data; Performing diabetes state change analysis of diabetes indicator disturbance based on the perturbed diabetes indicator characteristic data to generate diabetes indicator disturbance state change data; The state impact characterization capability of the diabetes indicator is analyzed based on the disturbance state change data of the diabetes indicator, and the state impact characterization capability data of the diabetes indicator is generated.

[0033] In an embodiment of the present invention, based on known clinical experience and medical literature, a group of interference parameters related to the development of diabetes are selected, such as factors such as diet, exercise, and drug intervention. These interference parameters can be preset through data collection or expert knowledge base. According to the amplitude of different interference parameters, the patient's diabetes indicator characteristic data is disturbed. For example, by setting a certain disturbance range (such as ±10%), the weight, blood sugar, blood lipids and other indicators are increased or decreased to generate disturbed diabetes indicator characteristic data, ensuring that the disturbed data can reflect the potential impact of different clinical interventions or environmental factors on the diabetic state, and provide basic data for subsequent analysis. After the perturbed diabetes indicator characteristic data is generated, the impact of different interference factors on the diabetic state is observed by analyzing the changes in the diabetic state of these data. Use statistical models (such as support vector machines, logistic regression models, etc.) to model the perturbed diabetes indicator characteristic data and predict the state changes of diabetic patients. For example, a mapping relationship is established between the changes in diabetes indicators and specific states such as blood sugar control and diabetic complications through training models. Use the model to predict the perturbed data and generate diabetes state change data under different perturbation conditions. The results of the analysis show how the changes in various diabetes indicators affect the overall diabetes control state and how different interference factors lead to fluctuations in the diabetes state. The generated diabetes indicator disturbance state change data provides a quantitative basis for the subsequent analysis of the influence of diabetes indicators. Based on the diabetes indicator disturbance state change data, the weighted regression analysis method is used for regression modeling to evaluate the influence of each disturbance diabetes indicator feature on the change of diabetes state. Through model calculation, the ability to characterize the influence of each disturbance indicator on the diabetes state can be obtained. In this process, regression models such as ridge regression and LASSO regression are used to ensure that the model can accurately identify important features and avoid overfitting when the data dimension is high. Subsequently, based on the regression results, the state influence characterization ability data of diabetes indicators is generated, reflecting the contribution and performance of different diabetes indicators in the change of diabetes state. Finally, these data provide a scientific basis for the design of the weighted diabetes indicator feature matrix and the evaluation of diabetes reversal potential, helping the system to accurately identify which indicators have a greater influence on diabetes reversal.

[0034] Furthermore, the diabetes state change analysis of diabetes indicator disturbance based on the perturbed diabetes indicator characteristic data includes: Performing a disturbance change analysis of the diabetes indicator characteristic according to the disturbed diabetes indicator characteristic data to generate the diabetes indicator characteristic disturbance change data; Obtain the patient's historical diabetes indicator status assessment data; Based on the preset XGBoost algorithm and historical diabetes indicator status evaluation data, a mapping relationship between indicator disturbance and diabetes status change is established to obtain an indicator disturbance diabetes status evaluation model; The indicator perturbation diabetes state assessment model is used to analyze the diabetes state change of the diabetes indicator disturbance on the diabetes indicator characteristic perturbation change data to generate diabetes indicator disturbance state change data.

[0035] In an embodiment of the present invention, the perturbation change analysis of the diabetes indicator characteristics quantifies the impact of the diabetes indicator changes on the patient's health status by performing a detailed analysis of the diabetes indicator characteristic data after the disturbance. Based on the perturbed diabetes indicator characteristic data, data mining techniques such as cluster analysis and principal component analysis (PCA) are used to analyze the change amplitude and correlation of each diabetes indicator characteristic. Identify the relationship between different indicator changes and diabetes control status. For example, for the perturbation change of blood sugar level, by comparing the perturbed blood sugar data with the patient's diabetes control results (such as fasting blood sugar, postprandial blood sugar, etc.), observe the impact of the perturbation on diabetes control, and then generate diabetes indicator characteristic perturbation change data. These change data reflect how the diabetes indicators fluctuate under different intervention conditions, and provide a basis for subsequent analysis of diabetes state changes. Obtaining the patient's historical diabetes indicator status assessment data is achieved through the patient's long-term health records and diagnosis and treatment records. Obtain the patient's previous diabetes-related examination data through the patient's medical information system (such as an electronic medical record system EMR), including blood sugar level, blood pressure, weight, insulin usage and other related indicators. These data are systematically organized into time series data and cleaned and processed through data preprocessing techniques (such as removing outliers, filling missing data, and standardizing data). The acquisition of this data not only provides information on the patient's historical health status, but also provides the necessary historical basis for establishing a diabetes status assessment model. When constructing the mapping relationship of diabetes status changes, the XGBoost algorithm is used to analyze the relationship between historical diabetes indicator status assessment data and diabetes status. The diabetes indicators (such as blood sugar, blood lipids, weight, etc.) in the historical data and the corresponding diabetes status (such as diabetes control level, whether there are complications, etc.) are used as training data to train the XGBoost model. The model sets multiple decision trees to maximize information gain and gradually optimizes the predictive ability of diabetes indicators for diabetes status changes. Using the trained model, the perturbed diabetes indicator feature data is used as input to map the relationship between these perturbed indicators and diabetes status changes, and then the indicator perturbation diabetes status assessment model is obtained. This model can predict the changes in diabetes status of patients under different interventions by inputting new diabetes indicator data. After the indicator perturbation diabetes status assessment model is constructed, the model is used to analyze the changes in diabetes status of diabetes indicator feature perturbation change data. The diabetes indicator characteristic disturbance change data is input into the XGBoost model for predictive analysis. Based on the input diabetes indicator data, the model will output the specific impact of each indicator change on the diabetes control state, including whether it will lead to an increase in blood sugar levels and whether it may cause diabetes complications. By analyzing each indicator, the diabetes indicator disturbance state change data is generated, which describes in detail the impact of diabetes indicators on the overall state change under different disturbance conditions.The analysis results can provide a basis for subsequent treatment decisions and help doctors develop more accurate treatment plans.

[0036] Further, as an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 FIG. 1 is a schematic diagram of a detailed functional flow of a diabetes state trend analysis module in FIG. 1 . In this embodiment, the diabetes state trend analysis module includes the following functions: S21: performing diabetes status analysis according to the weighted diabetes indicator feature matrix to generate diabetes status data; In an embodiment of the present invention, the diabetes status analysis helps identify the patient's diabetes control level and potential risks by conducting an in-depth analysis of the weighted diabetes indicator feature matrix. Based on the weighted diabetes indicator feature matrix, a multivariate regression analysis method is used to comprehensively evaluate multiple diabetes-related indicators (such as blood sugar, blood pressure, BMI, etc.) and establish a mathematical model. The weight assigned to each indicator feature is determined by the aforementioned weighting process to ensure that the impact of each indicator on the diabetes status is reasonably reflected. The model predicts the patient's current diabetes control level (such as good control, mild out of control, severe out of control, etc.). This analysis helps generate diabetes status data, including key indicators such as the patient's diabetes grade and risk score, providing a basis for the selection of subsequent treatment plans.

[0037] S22: performing diabetes state space definition processing based on the diabetes state data to generate diabetes state space data; In an embodiment of the present invention, the diabetes state space definition process is used to convert the multidimensional state information of diabetes into a standardized mathematical model, and define different state space dimensions according to the diabetes state data (including blood sugar level, insulin use, weight change, etc.), such as blood sugar control level, body mass index, frequency of complications, etc. Use data clustering technology (such as K-means clustering) to classify similar states into the same spatial area to generate diabetes state space data. This state space data visualization represents the health status of diabetic patients under different clinical indicators, helping doctors to more intuitively understand the patient groups in different states, and then carry out personalized treatment.

[0038] S23: performing diabetes time series state data analysis on the diabetes state data to generate diabetes time series state data; In an embodiment of the present invention, the diabetes time series state data analysis is used to observe the state change trend of diabetic patients over a period of time. The diabetes state data is sorted into time series data in chronological order. The time series data is modeled using a time series data analysis method, such as an autoregressive moving average model (ARMA) or a long short-term memory neural network (LSTM). The model identifies the law of changes in the diabetes state, such as whether the diabetes state is gradually worsening or under control. The generated diabetes time series state data not only reflects the patient's diabetes control trend over a certain period of time, but also can predict the future development trend of the diabetes state, providing data support for the evaluation of treatment effects and the selection of intervention timing.

[0039] S24: performing a transfer characteristic trend analysis of the diabetes time series state based on the diabetes state spatial data and the diabetes time series state data to generate diabetes time series state transfer characteristic trend data; In an embodiment of the present invention, the diabetes time series state transition feature trend analysis is intended to identify the transition trend of the state of a diabetic patient from one stage to another. Based on the diabetes state space data and the diabetes time series state data, a Markov chain model is used to describe the transition process of the diabetes state from one state to another. By calculating the transition probability between each diabetes state, the transition frequency and trend between different states are analyzed, such as the probability of turning from mild diabetes out of control to moderate diabetes out of control. Based on the results of the transition feature trend analysis, diabetes time series state transition feature trend data is generated. This data helps identify whether a diabetic patient is at risk of worsening his or her state within a specific time period, and provides a basis for doctors to adjust treatment.

[0040] S25: performing diabetes time series state periodic change pattern analysis according to the diabetes time series state data to generate diabetes state periodic change pattern data; In an embodiment of the present invention, the periodic change pattern analysis of the diabetes time series state is used to discover the periodic fluctuations of the diabetes state over a period of time. The diabetes time series state data is subjected to Fourier transform or wavelet transform, the time series data is converted from the time domain to the frequency domain, and the periodic components are extracted. The periodic analysis method, such as periodic autocorrelation analysis, is used to identify the periodic pattern of changes in the diabetes state, such as the volatility of blood sugar control each quarter. The generated diabetes state periodic change pattern data shows the regularity of the state changes of diabetic patients, helps determine the intervention measures required by patients in different time periods, and optimizes the treatment plan.

[0041] S26: Perform personalized diabetes state trend analysis based on the diabetes time series state transition characteristic trend data and the diabetes state periodic change pattern data to generate personalized diabetes state trend data.

[0042] In an embodiment of the present invention, the personalized diabetes state trend analysis combines the diabetes time series state transition characteristic trend data and the diabetes state periodic change pattern data to tailor a diabetes management plan for the patient. The diabetes time series state transition characteristic trend data and the periodic change pattern data are used in combination, and based on the individual's historical data, a personalized diabetes state trend analysis model is established using machine learning algorithms (such as support vector machines, random forests, etc.). This model can not only analyze the patient's state change trend in different time periods, but also predict the future diabetes state based on the patient's individual characteristics (such as age, gender, living habits, etc.). The generated personalized diabetes state trend data provides diabetic patients with targeted health management plans and treatment recommendations, making intervention measures more accurate and the treatment effect more significant.

[0043] Furthermore, the transfer characteristic trend analysis of the diabetes time series state is performed based on the diabetes state spatial data and the diabetes time series state data: Performing characteristic frequency distribution analysis of diabetes time series state data on diabetes time series state data through diabetes state space data to generate diabetes time series state characteristic frequency distribution data; The transfer characteristic trend of the diabetes time series state is analyzed according to the diabetes time series state characteristic frequency distribution data to generate the diabetes time series state transfer characteristic trend data.

[0044] In an embodiment of the present invention, the characteristic frequency distribution analysis of the diabetes time series state reveals important patterns in diabetes management by analyzing the frequency of state changes of diabetic patients at different time points, and collects diabetes time series state data, which records the patient's state over a period of time (such as blood sugar level, weight, drug use, etc.). The frequency analysis method (such as histogram or probability density function) is used to classify the diabetes state data at each time point, and the frequency of occurrence of each state is calculated. The purpose of frequency analysis is to identify the most common intervals of diabetes state and which states frequently appear in patient management. Using these frequency distribution data, the characteristic frequency distribution data of diabetes time series state is generated, thereby providing basic data for subsequent analysis of diabetes state changes to ensure the accuracy and operability of the analysis. The transition characteristic trend analysis of diabetes time series state reveals the possibility and law of diabetes patients transferring from one state to another by analyzing the transition trend between diabetes states. Using the aforementioned generated diabetes time series state characteristic frequency distribution data, a diabetes state transition matrix is ​​constructed. The matrix records the probability of diabetic patients transferring from one state to another at different time nodes. For example, if the probability of a patient's blood sugar changing from well-controlled to slightly out of control is 30%, this data will be reflected in the transition matrix. The generation of the transition matrix depends on the statistics and analysis of historical data. Based on these transition matrices, the Markov chain model is used to model the transition trend of the diabetes time series state, analyze the change rules between different diabetes states, and generate diabetes time series state transition characteristic trend data. This data helps identify which states are most likely to transfer and which factors lead to the deterioration or improvement of the diabetes state, providing guidance for the formulation of subsequent personalized intervention strategies.

[0045] Furthermore, the transfer characteristic trend analysis of the diabetes time series state according to the diabetes time series state characteristic frequency distribution data includes: Performing a diabetes time series state transition feature analysis based on diabetes time series state feature frequency distribution data to generate diabetes time series state transition feature data; The preset Markov chain model is used to perform a transfer characteristic trend analysis of the diabetes time series state transfer characteristic data to generate the diabetes time series state transfer characteristic trend data.

[0046] In an embodiment of the present invention, the transition characteristic analysis of the diabetes time series state reveals the transition relationship between the states of the diabetic patient on the time axis by conducting an in-depth study on the frequency distribution data of the time series state. A state transition table is constructed by collecting the diabetes time series state data, which contains the transition probability of each diabetes state at different time points. For each pair of states (such as from hyperglycemia to normal blood sugar, or from mild diabetes to severe diabetes), the number of transitions from one state to another is calculated, and the transition probability is calculated based on these times. These transition probabilities use frequency statistics methods, such as normalized transition frequencies, to generate diabetes time series state transition characteristic data. This data can help identify whether there is a regularity or trend in the health management of diabetic patients, and which state transitions are more frequent, thereby providing a basis for the optimization of diabetes intervention strategies. The Markov chain model is used to perform a transition characteristic trend analysis on the diabetes time series state transition characteristic data, and the future changes in diabetes state are predicted by establishing a state transition model. Based on the transition characteristic data of the diabetes time series state, a Markov chain model is constructed, which defines a series of states and their transition probabilities. For example, if the current state of a diabetic patient is "good blood sugar control", based on the statistics of historical data, the model will predict the probability of the patient transferring to the state of "mild diabetes" or "out of blood sugar control" in the next period of time. The establishment of the model requires the use of a certain initial data set, and the transition probability between each state is learned through the training algorithm. After the model training is completed, this Markov chain model is used to predict the future transition trend of the diabetes time series state, and based on the current patient's diabetes state, the possible state transition path in the future time period is predicted to generate diabetes time series state transition characteristic trend data. This data helps to formulate personalized intervention plans, predict the changing trend of patients' diabetes status, and provide decision support for clinical treatment.

[0047] Furthermore, the diabetes reversal potential assessment module includes the following functions: According to the preset diabetes reversal factors and the multi-source diabetes indicator analysis data of the weighted diabetes indicator characteristic matrix, the causal relationship analysis between each indicator and the diabetes reversal factor is performed to generate the indicator diabetes reversal factor causal relationship data; According to the diabetes indicator feature weight parameters of the weighted diabetes indicator feature matrix and the causal relationship data of diabetes reversal factors, the diabetes reversal impact quantitative analysis of the indicator is performed to generate the diabetes reversal impact quantitative data of the indicator; The diabetes reversal potential assessment modeling is performed based on the quantitative data of the diabetes reversal impact indicators to generate a diabetes reversal potential assessment model.

[0048] In the embodiment of the present invention, in the process of evaluating the potential for diabetes reversal, a causal relationship analysis is first performed based on the preset diabetes reversal factors and the multi-source diabetes indicator analysis data in the weighted diabetes indicator feature matrix to identify how the key indicators of diabetes affect the potential for diabetes reversal. Factors related to diabetes reversal are selected, such as blood sugar control level, islet function, weight management, and lifestyle, and these diabetes indicator features are extracted from multi-source data. Statistical analysis methods (such as correlation analysis, regression analysis, or causal inference algorithm) are used to analyze the relationship between each diabetes indicator and the reversal factor to generate causal relationship data between the diabetes indicator and the reversal factor. A quantitative analysis of the impact of diabetes reversal is performed based on the diabetes indicator feature weight parameters in the weighted diabetes indicator feature matrix and the causal relationship data of the diabetes reversal factor. The diabetes indicator feature weight parameters reflect the importance of each indicator to the diabetic state. According to these weights, the influence of different diabetes indicators is weighted. Combined with the causal relationship data, determine which indicators have a greater impact on diabetes reversal. Mathematical modeling methods, such as multiple regression analysis, structural equation modeling, etc., are used to quantify the relationship between each indicator and the potential for diabetes reversal, and the degree of influence of each diabetes indicator on diabetes reversal is obtained. Generate quantitative data of the diabetes reversal effect of indicators, which provide a quantitative basis for evaluating the possibility of diabetes reversal and help further optimize the reversal treatment strategy. Based on the generated quantitative data of the diabetes reversal effect of indicators, the diabetes reversal potential assessment modeling process is carried out. The goal of this process is to build a model for analyzing the impact of each indicator on the diabetes reversal potential, which can evaluate the diabetes reversal potential of the patient based on the patient's specific diabetes indicators and reversal effect quantitative data. By using machine learning algorithms, such as support vector machines (SVMs), random forests, etc., a multivariate regression model or classification model is constructed to model the relationship between each diabetes indicator and reversal potential. When training the model, a large amount of historical data related to diabetes indicators and diabetes reversal treatment is used, combined with different indicator characteristics and reversal effect quantitative data of diabetic patients to improve the accuracy and predictive ability of the model. The model can evaluate the potential for diabetes reversal based on the patient's specific situation and generate a diabetes reversal potential assessment model. This model can not only help doctors better understand the patient's condition, but also provide decision support for the formulation of personalized treatment plans.

[0049] Furthermore, the early diabetes reversal treatment assessment module includes the following functions: Obtain early intervention strategies for diabetes reversal treatment; Use the diabetes reversal potential assessment model to conduct early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategy, and generate early diabetes reversal treatment intervention benefit simulation data; Perform early diabetes reversal treatment evaluation processing on early diabetes reversal treatment intervention benefit simulation data according to personalized diabetes status trend data to generate early diabetes reversal treatment evaluation data; Conduct early diabetes reversal treatment demand analysis based on personalized diabetes status trend data to generate early diabetes reversal treatment demand data; Performing an early diabetes reversal treatment feedback optimization coefficient analysis on the early diabetes reversal treatment evaluation data through the early diabetes reversal treatment demand data, and generating an early diabetes reversal treatment feedback optimization coefficient; According to the early diabetes reversal treatment feedback optimization coefficient, the early diabetes reversal treatment intervention strategy is subjected to reversal treatment intervention strategy feedback optimization processing to generate an optimized early diabetes reversal treatment intervention strategy.

[0050] In the embodiment of the present invention, in the diabetes reversal treatment evaluation module, it is first necessary to obtain the early diabetes reversal treatment intervention strategy. These intervention strategies are pre-set medical intervention strategies for patients, including diet adjustment, exercise intervention, drug treatment, etc. After obtaining the early diabetes reversal treatment intervention strategy, these strategies are simulated and analyzed using the diabetes reversal potential evaluation model. By inputting various diabetes indicators of the patient and their corresponding diabetes reversal potential evaluation data, the potential evaluation model established by machine learning algorithms (such as neural networks, etc.) is used to predict the effects of different intervention strategies on diabetes reversal. The model will generate intervention benefit simulation data by simulating the implementation effects of different treatment plans. These data will include the effects of each intervention strategy in different time periods, such as changes in blood sugar levels, improvements in pancreatic islet function, changes in weight, etc. This process is based on a large amount of historical data and individualized patient data to ensure the accuracy of the simulation analysis. After obtaining the early diabetes reversal treatment intervention benefit simulation data, the intervention benefit is evaluated and processed according to the personalized diabetes state trend data. The personalized diabetes state trend data reflects the current and historical diabetes development trends of the patient, including blood sugar control levels, changes in pancreatic islet function, living habits, etc. By combining these data with intervention benefit simulation data, regression analysis, classification analysis and other methods are used to evaluate the effect of each intervention strategy in different patient groups. The long-term effect of treatment is predicted based on the trend of blood sugar level changes, and the specific efficacy evaluation data of each strategy is obtained through the model. This analysis process can provide personalized treatment plan optimization suggestions. Based on personalized diabetes status trend data, early diabetes reversal treatment demand analysis is carried out. By analyzing the patient's diabetes status at various time points and combining historical treatment data, it is evaluated whether the patient currently needs additional treatment intervention. By using time series data analysis methods, it is identified whether the patient's diabetes status is worsening and whether it is necessary to strengthen intervention. The patient's condition deterioration trend will become an important basis for formulating treatment needs, effectively identifying the treatment needs of different patients during the treatment process, and providing data support for the next step of treatment plan optimization. After obtaining the early diabetes reversal treatment demand data, the early diabetes reversal treatment evaluation data is analyzed for feedback optimization coefficients. By using feedback control theory or optimization algorithms to process the evaluation data, it is determined which factors (such as intervention strategies, treatment intensity, treatment cycle, etc.) affect the patient's diabetes reversal potential. Using model analysis, an optimization coefficient is generated to reflect how to adjust the treatment plan to improve the efficacy in specific situations. Optimization coefficient analysis can combine treatment strategies with the specific needs of patients, ensuring the flexibility and accuracy of treatment plans in a dynamic environment. After the optimization coefficients are generated, these optimization coefficients are used to feedback and optimize the early diabetes reversal treatment intervention strategy.Based on the feedback optimization coefficient, mathematical optimization methods (such as optimization algorithms) are used to adjust the treatment intervention plan, including adjusting treatment parameters such as drug dosage, exercise plan, and diet plan according to the patient's personalized treatment needs. The optimized intervention strategy will improve the effect of diabetes reversal, reduce treatment risks, and ensure the continuity and effectiveness of treatment. For example, during the treatment process, if it is found that the patient's blood sugar level has not dropped as expected, the model will adjust the intervention intensity or improve the treatment plan according to the optimization coefficient, and finally generate an optimized early diabetes reversal treatment intervention strategy. This strategy provides patients with a personalized and dynamically adjusted treatment plan, which improves the success rate of reversal treatment.

[0051] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0052] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An early diabetes reversal treatment evaluation system based on machine learning, characterized in that: Includes the following modules: The diabetes index parsing and weighting module is used to obtain the patient's early diabetes medical file data; perform multi-source diabetes index parsing on the patient's early diabetes medical file data to generate multi-source diabetes index parsing data; perform weighting matrix analysis on diabetes index characteristics based on the multi-source diabetes index parsing data to generate a weighted diabetes index characteristic matrix; The diabetes state trend analysis module is used to perform diabetes state analysis based on the weighted diabetes indicator feature matrix to generate diabetes state data; perform personalized diabetes state trend analysis based on the diabetes state data to generate personalized diabetes state trend data; The diabetes reversal potential assessment module is used to conduct quantitative analysis of the diabetes reversal impact of indicators based on the preset diabetes reversal factors and the weighted diabetes indicator feature matrix, and generate indicator diabetes reversal impact quantitative data; conduct diabetes reversal potential assessment modeling processing based on the indicator diabetes reversal impact quantitative data, and generate a diabetes reversal potential assessment model; The early diabetes reversal treatment evaluation module is used to obtain the early diabetes reversal treatment intervention strategy; use the diabetes reversal potential evaluation model to perform early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategy, and generate early diabetes reversal treatment intervention benefit simulation data; perform early diabetes reversal treatment evaluation processing on the early diabetes reversal treatment intervention benefit simulation data according to personalized diabetes state trend data, and generate early diabetes reversal treatment evaluation data; perform reversal treatment intervention strategy feedback optimization processing on the early diabetes reversal treatment intervention strategy based on personalized diabetes state trend data and early diabetes reversal treatment evaluation data, and generate an optimized early diabetes reversal treatment intervention strategy.

2. The early diabetes reversal treatment evaluation system based on machine learning according to claim 1, characterized in that: The diabetes indicator analysis and empowerment module includes the following functions: Obtain the patient's early diabetes medical record data; Perform multi-source diabetes index analysis on the patient's early diabetes medical record data to generate multi-source diabetes index analysis data; Performing multi-source diabetes indicator feature screening processing on the multi-source diabetes indicator analysis data to generate multi-source diabetes indicator feature data; A weighted matrix analysis of diabetes indicator characteristics is performed based on multi-source diabetes indicator characteristic data to generate a weighted diabetes indicator characteristic matrix.

3. The early diabetes reversal treatment evaluation system based on machine learning according to claim 2, characterized in that: The weighted matrix analysis of diabetes indicator characteristics based on multi-source diabetes indicator characteristic data includes: Based on multi-source diabetes indicator characteristic data, the state impact characterization capability of diabetes indicators is analyzed to generate state impact characterization capability data of diabetes indicators; Perform a weight parameter analysis of the diabetes indicator characteristics according to the diabetes indicator status impact characterization capability data to obtain the diabetes indicator characteristic weight parameter; The weighted matrix of diabetes indicator characteristics is designed through multi-source diabetes indicator characteristic data and corresponding diabetes indicator characteristic weight parameters to generate a weighted diabetes indicator characteristic matrix.

4. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 3, characterized in that: The state impact characterization capability analysis of diabetes indicators based on multi-source diabetes indicator characteristic data includes: Performing diabetes indicator characteristic disturbance processing on multi-source diabetes indicator characteristic data using preset diabetes indicator interference parameters to generate disturbed diabetes indicator characteristic data; Performing diabetes state change analysis of diabetes indicator disturbance based on the perturbed diabetes indicator characteristic data to generate diabetes indicator disturbance state change data; The state impact characterization capability of the diabetes indicator is analyzed based on the disturbance state change data of the diabetes indicator, and the state impact characterization capability data of the diabetes indicator is generated.

5. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 4, characterized in that: The diabetes state change analysis of the diabetes index disturbance based on the disturbed diabetes index characteristic data includes: Performing a disturbance change analysis of the diabetes indicator characteristic according to the disturbed diabetes indicator characteristic data to generate the diabetes indicator characteristic disturbance change data; Obtain the patient's historical diabetes indicator status assessment data; Based on the preset XGBoost algorithm and historical diabetes indicator status evaluation data, a mapping relationship between indicator disturbance and diabetes status change is established to obtain an indicator disturbance diabetes status evaluation model; The indicator perturbation diabetes state assessment model is used to analyze the diabetes state change of the diabetes indicator disturbance on the diabetes indicator characteristic perturbation change data to generate diabetes indicator disturbance state change data.

6. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 1, characterized in that: The diabetes status trend analysis module includes the following functions: Perform diabetes status analysis based on the weighted diabetes indicator feature matrix to generate diabetes status data; Performing diabetes state space definition processing based on the diabetes state data to generate diabetes state space data; Performing diabetes time series state data analysis on the diabetes state data to generate diabetes time series state data; Based on the diabetes state spatial data and the diabetes time series state data, a transfer characteristic trend analysis of the diabetes time series state is performed to generate diabetes time series state transfer characteristic trend data; Performing diabetes time series state periodic change pattern analysis based on diabetes time series state data to generate diabetes state periodic change pattern data; Based on the diabetes time series state transition characteristic trend data and the diabetes state periodic change pattern data, personalized diabetes state trend analysis is performed to generate personalized diabetes state trend data.

7. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 6, characterized in that: The transfer characteristic trend analysis of the diabetes time series state based on the diabetes state spatial data and the diabetes time series state data is as follows: Performing characteristic frequency distribution analysis of diabetes time series state data on diabetes time series state data through diabetes state space data to generate diabetes time series state characteristic frequency distribution data; The transfer characteristic trend of the diabetes time series state is analyzed according to the diabetes time series state characteristic frequency distribution data to generate the diabetes time series state transfer characteristic trend data.

8. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 7, characterized in that: The method of performing the transfer characteristic trend analysis of the diabetes time series state according to the diabetes time series state characteristic frequency distribution data comprises: Performing a diabetes time series state transition feature analysis based on diabetes time series state feature frequency distribution data to generate diabetes time series state transition feature data; The preset Markov chain model is used to perform a transfer characteristic trend analysis of the diabetes time series state transfer characteristic data to generate the diabetes time series state transfer characteristic trend data.

9. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 3, characterized in that: The Diabetes Reversal Potential Assessment Module includes the following features: According to the preset diabetes reversal factors and the multi-source diabetes indicator analysis data of the weighted diabetes indicator characteristic matrix, the causal relationship analysis between each indicator and the diabetes reversal factor is performed to generate the indicator diabetes reversal factor causal relationship data; According to the diabetes indicator feature weight parameters of the weighted diabetes indicator feature matrix and the causal relationship data of diabetes reversal factors, the diabetes reversal impact quantitative analysis of the indicator is performed to generate the diabetes reversal impact quantitative data of the indicator; The diabetes reversal potential assessment modeling is performed based on the quantitative data of the diabetes reversal impact indicators to generate a diabetes reversal potential assessment model.

10. The early stage diabetes reversal treatment evaluation system based on machine learning according to claim 1, characterized in that: The Early Diabetes Reversal Treatment Assessment Module includes the following features: Obtain early intervention strategies for diabetes reversal treatment; Use the diabetes reversal potential assessment model to conduct early diabetes reversal treatment intervention benefit simulation analysis on the early diabetes reversal treatment intervention strategy, and generate early diabetes reversal treatment intervention benefit simulation data; Perform early diabetes reversal treatment evaluation processing on early diabetes reversal treatment intervention benefit simulation data according to personalized diabetes status trend data to generate early diabetes reversal treatment evaluation data; Conduct early diabetes reversal treatment demand analysis based on personalized diabetes status trend data to generate early diabetes reversal treatment demand data; Performing an early diabetes reversal treatment feedback optimization coefficient analysis on the early diabetes reversal treatment evaluation data through the early diabetes reversal treatment demand data, and generating an early diabetes reversal treatment feedback optimization coefficient; According to the early diabetes reversal treatment feedback optimization coefficient, the early diabetes reversal treatment intervention strategy is subjected to reversal treatment intervention strategy feedback optimization processing to generate an optimized early diabetes reversal treatment intervention strategy.

Images