Effect evaluation method for non-dialysis diabetic kidney disease

Abstract

The invention discloses an effect evaluation method for a non-dialysis diabetic kidney disease, which comprises the following steps: monitoring the change trend of data by collecting renal function related data of a patient, and generating a basic data set; a dynamic nonlinear weighted regression algorithm is adopted to analyze the multi-dimensional renal function data, data weights are dynamically adjusted, and influence weights of different time periods are distributed according to the change mode of the illness state; the data influence is adjusted in real time through a time correlation mechanism, and a disease assessment curve is generated; according to the method, historical data and real-time data are combined, a patient condition trend line and a prediction result are generated, personalized treatment feedback is output, and intervention suggestions are provided for clinicians based on long-term prognosis prediction. The method is suitable for disease monitoring and treatment feedback of patients with non-dialysis diabetic kidney diseases.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing and analysis, and in particular to an effect evaluation method for non-dialysis diabetic nephropathy. Background Art

[0002] Diabetic nephropathy is a common complication among diabetic patients. As the disease progresses, renal function is gradually impaired, and in severe cases it can lead to end-stage renal disease. For non-dialysis diabetic nephropathy patients, timely and accurate assessment of changes in renal function and provision of effective treatment feedback are the key to slowing disease progression and delaying the need for dialysis or transplantation.

[0003] Existing technologies rely heavily on single, static measurements of physiological indicators for data collection and processing. The indicators usually collected include blood sugar, urine protein, serum creatinine, etc. Although these indicators can reflect the current state of the patient's renal function, the condition of non-dialysis diabetic kidney disease patients usually fluctuates slowly and complexly, and static data is difficult to capture the changing trend of the condition. Traditional analysis methods often process each data point independently, ignoring the time correlation between these data and the continuity of the disease changes, and are unable to fully reflect the fluctuations of the patient's condition.

[0004] Most of the existing prediction models are based on linear regression or fixed-weight algorithms. Although such models can make preliminary predictions of the disease, for chronic diseases such as diabetic nephropathy with a complex course and slow changes, the fixed-weight algorithms cannot be flexibly adjusted, which affects the accuracy of the prediction results. Especially when the patient's condition changes slowly or presents nonlinear changes, the traditional linear regression model is difficult to adapt to this complexity and cannot adjust the prediction parameters in time, resulting in a lag in the assessment of the disease. This fixed model cannot effectively deal with patients with large fluctuations in their condition and is prone to misjudgment and even delayed treatment.

[0005] The disease assessment of existing technologies often relies on single or limited physiological indicators for analysis, ignoring the fusion processing of multi-dimensional data. The disease changes of diabetic kidney disease are affected by many factors, including blood sugar levels, urine protein content, serum creatinine concentration and other physiological parameters; the existing methods lack unified processing and comprehensive analysis of these multi-dimensional data, and cannot form a comprehensive assessment of the disease. Even if there are some technical means of multi-parameter analysis, they often simply linearly combine different indicators and fail to fully utilize the interrelationships between various physiological parameters. This processing method leads to the one-sidedness of the evaluation results and cannot fully reflect the overall development of the patient's disease.

[0006] In terms of treatment feedback, existing technologies also have major deficiencies. Since traditional predictive models cannot be adjusted dynamically, personalized treatment feedback mostly relies on the doctor's experience and judgment and lacks precise data-driven support. Doctors usually give treatment recommendations based on the patient's historical condition and single assessment results, but this method cannot respond to changes in the patient's condition in a timely manner, especially when the condition changes suddenly or gradually. Treatment feedback is often delayed and the treatment plan cannot be adjusted in real time. The personalized feedback in existing technologies lacks long-term condition monitoring and dynamic adjustment capabilities, especially when facing patients with chronic diseases. It is impossible to generate accurate long-term treatment plans based on the patient's real-time data and historical condition. This experience-based feedback method is particularly insufficient when dealing with complex condition fluctuations and cannot meet the personalized management needs of patients with diabetic kidney disease.

[0007] Existing disease prediction technologies also have limitations in processing the time dimension. The condition fluctuations of patients with chronic diseases are often long-term and cyclical. Existing technologies lack effective mechanisms when processing time-related data and are unable to dynamically adjust the prediction model according to the frequency and amplitude of disease fluctuations. Time weighting or time-correlation analysis is not fully applied in many existing technologies, resulting in the prediction model being able to only make static predictions based on historical data, which makes it difficult to reflect the long-term development trend of the disease. Due to the lack of a time-related adjustment mechanism, existing models are difficult to cope with long-term fluctuations in the disease, especially when the patient's condition gradually worsens or undergoes sudden changes, the deviation of the prediction results will be further increased.

[0008] Therefore, how to provide a method for evaluating the effects of non-dialysis diabetic nephropathy is an urgent problem to be solved by those skilled in the art. Summary of the invention

[0009] One purpose of the present invention is to propose a method for evaluating the effect of non-dialysis diabetic nephropathy. The present invention utilizes a dynamic nonlinear weighted regression algorithm and a time-correlation dynamic adjustment mechanism, and describes in detail the entire process from the collection of patient renal function data, real-time monitoring of disease fluctuations, dynamic adjustment of data weights, to disease prediction and generation of personalized treatment feedback. By combining multidimensional data analysis and an adaptive adjustment mechanism, the present invention can capture the nonlinear changes in the patient's condition and provide a real-time personalized treatment plan, with the advantages of high monitoring accuracy, timely feedback, and strong personalization.

[0010] A method for evaluating the effect of non-dialysis diabetic nephropathy according to an embodiment of the present invention comprises the following steps:

[0011] S1. Collect the patient's renal function-related data, and continuously monitor data fluctuations during the collection process to generate a basic data set;

[0012] S2. Use dynamic nonlinear weighted regression algorithm to analyze and process the multidimensional renal function data of patients, adjust the weight of data points, dynamically allocate data weights of different time periods and characteristics based on the magnitude of changes in the condition, trends and correlation with historical data, and capture the nonlinear changes in the patient's condition;

[0013] S3. Introduce a time-related dynamic adjustment mechanism to adjust the influence of data in different time periods in real time based on the frequency of fluctuations in the patient's condition and generate a condition assessment curve;

[0014] S4. Generate a prediction of renal function changes based on the output results of the dynamic nonlinear weighted regression algorithm, and output a disease trend line by combining the patient's multidimensional historical data and current disease fluctuations;

[0015] S5. Based on the prediction results, dynamically adjust the weights to generate a treatment feedback plan, and optimize the treatment strategy by combining the patient's real-time data and long-term medical history;

[0016] S6. Through the long-term prognosis prediction of dynamic nonlinear weighted regression algorithm, combined with time correlation and disease fluctuation trend, a forecast report on future disease development is generated and intervention suggestions are provided.

[0017] Optionally, the S1 specifically includes:

[0018] S11. Collect the patient's renal function-related data, including blood glucose G(t), urine protein P(t), serum creatinine C(t) and estimated glomerular filtration rate E(t), attach a timestamp t to each data point, and generate a preliminary time series data set D t ;

[0019] S12. Based on the preliminary dataset D t , at each time point t, the real-time fluctuation monitoring of the collected physiological indicators is performed, and the fluctuation value of each data point is calculated ΔX(t)=X(t) where X(t-1) is the data at the current time point, and X(t) is the data at the previous time point;

[0020] S13. According to the fluctuation monitoring result ΔX(t), when the fluctuation amplitude exceeds the threshold, the sampling frequency will be adjusted according to the formula f(t) = α×e -β×|ΔX(t)| Dynamic adjustment is used to increase the sampling frequency f(t) when the data fluctuates greatly, shorten the collection time interval Δt, and generate a dense data set D dense ;

[0021] S14. Dense dataset D generated after the time interval is shortened dense Input into the adaptive filtering algorithm to perform noise filtering and smoothing to eliminate outliers and noise interference;

[0022] S15, apply adaptive filtering algorithm to Ddense The filter formula is where w i is the weight of the first N data points;

[0023] S16, the final processed basic data set D base output through an adaptive filter.

[0024] Optionally, the S2 specifically includes:

[0025] S21. Use dynamic nonlinear weighted regression algorithm to analyze and process the multidimensional renal function data of patients, and combine it with the generated basic data set D base Conduct analysis;

[0026] S22, by adaptively adjusting the weights of data points, combining the changing patterns of the condition and historical data, dynamically assigning weights according to data changes in different time periods and characteristics, and capturing nonlinear changes in the patient's condition;

[0027] S23. Time weighting for disease condition changes:

[0028]

[0029] Among them, ΔX(t) represents the change of the current data point, Δt is the time interval, and λ is the time-related adjustment factor;

[0030] S24, the disease fluctuation detection mechanism is used to identify abnormal fluctuation points and adjust the weight distribution accordingly to identify key disease change trends;

[0031] S25. Combine the time-dependent function T(t) to make long-term trend predictions, analyze the patient's renal function changes by adjusting the time window T(t), and make long-term evaluations and predictions.

[0032] Optionally, the S3 specifically includes:

[0033] S31, introduce time-related dynamic adjustment mechanism to process the multi-dimensional renal function data of patients, based on the basic data set D base And weight W(t) for data processing;

[0034] S32. Adjust the influence of data in different time periods according to the frequency of fluctuations in the patient's condition:

[0035] Wadj(t)=W(t)×(1+μ·f(Δt,B(t)));

[0036] Among them, W adj(t) is the adjusted weight, W(t) is the initial weight at the current time point, μ is the adjustment coefficient, and the function f(Δt, B(t)) represents the function of the time interval Δt and the disease fluctuation frequency B(t), which is used to dynamically adjust the weight distribution. It combines the characteristics of disease fluctuation and time interval changes to achieve dynamic adjustment of weights;

[0037] S33. The calculation of the disease fluctuation frequency B(t) is obtained by analyzing the fluctuation amplitude and trend in the historical data:

[0038]

[0039] Where N is the number of past time points, X(ti) is the data collected at time point ti, and the disease fluctuation frequency B(t) is used to identify sudden and gradual changes in the disease. By analyzing past data fluctuations, the changing trend of the disease can be identified, providing a basis for weight adjustment;

[0040] S34, after calculating the weight, using the dynamic nonlinear weighted regression algorithm according to the adjusted weight W adj (t) Analyze the disease data and generate a continuous disease assessment curve to reflect the change trend of the patient's renal function in different time periods;

[0041] S35. Adjust the weight distribution of the model in real time to enable it to respond flexibly when the condition changes suddenly or gradually, and generate continuous condition assessment results in the short term. The condition fluctuation detection mechanism identifies abnormal fluctuation points, dynamically adjusts their weights, responds promptly to current condition changes, and provides assessments based on short-term fluctuations;

[0042] S36. Combined with the time-dependent function T(t), the patient's current and recent changes in renal function are dynamically evaluated. The model adjusts the time window T(t) to deal with recent fluctuations in the condition and generates an evaluation result of the stable trend of the condition.

[0043] Optionally, the S4 specifically includes:

[0044] S41. Generate a long-term trend prediction of renal function changes based on a dynamic nonlinear weighted regression algorithm, generate a trend line of the patient's condition changes based on the patient's multidimensional historical data, combined with the current condition fluctuation data and short-term and recent evaluation results;

[0045] S42, combining the time-dependent function T(t) and the real-time adjustment mechanism to further generate a long-term disease prediction model:

[0046]

[0047] Among them, P(t) represents the long-term disease prediction result at time t, W adj(ti) is the dynamic adjustment weight at time point ti, X(ti) is the historical data collected at time point ti, δ is the time adjustment parameter, and T(t) is the time-dependent function that reflects the long-term trend of the disease;

[0048] S43. The calculation formula of the disease fluctuation frequency B(t) is:

[0049]

[0050] Where B(t) represents the frequency of disease fluctuation, which reflects the speed of disease change at time t, N represents the number of historical data points used for calculation, X(ti) is the historical data at time point ti, and Δt is the time interval between adjacent time points;

[0051] S44. Combining the fluctuation frequency of the disease and the adaptive adjustment mechanism, the model optimizes the data in different time periods to generate a long-term disease prediction curve to reflect the long-term disease change trend of the patient;

[0052] S45. The generated long-term trend line provides doctors with disease management advice and helps formulate long-term treatment and intervention measures. By combining short-term, recent and long-term disease data, the model can respond to changes in complex diseases and provide a reliable basis for long-term disease prediction.

[0053] Optionally, the S5 specifically includes:

[0054] S51. Generate a treatment feedback plan based on the patient's condition prediction results, combine the real-time data R(t) and historical medical history data H(t), comprehensively evaluate the patient's current health status, and generate treatment recommendations;

[0055] S52. Through a dynamic adjustment mechanism, weights are allocated according to the amplitude and frequency of disease fluctuations, using a nonlinear weight adjustment formula:

[0056]

[0057] Among them, W opt (t) represents the optimal weight after adjustment, W(t) is the initial weight, ΔR(t) represents the real-time data increment at the current time point t, ΔH(t) represents the increment of historical medical history data, γ is the weight adjustment coefficient, and β is the nonlinear scaling coefficient of data difference;

[0058] S53, based on the optimized weight W opt (t), combined with the speed of change of the patient's condition, generates customized treatment feedback. The model monitors the fluctuations of the condition in real time and adjusts the treatment plan to cope with the dynamic changes of the condition.

[0059] S54. Combine the disease fluctuation frequency B(t) and the patient's long-term data trend to optimize the long-term treatment plan and adjust the feedback mechanism:

[0060]

[0061] Where B(t) represents the frequency of disease fluctuation, N is the number of data points, Δt is the time interval, and δ is the scaling factor;

[0062] S55. Generate a treatment feedback curve based on feedback data to reflect the patient's treatment progress and provide a basis for adjusting future treatment plans to ensure that the plan can continue to adapt to changes in the patient's condition.

[0063] Optionally, the S6 specifically includes:

[0064] S61. Use a dynamic nonlinear weighted regression algorithm to generate a long-term disease prognosis prediction, combine the patient's multidimensional historical data and real-time data, analyze the fluctuation trend of kidney function, and generate a prediction model for future disease development;

[0065] S62, introducing time weighting function T weight (t), assign weights to data in different time periods:

[0066]

[0067] Among them, T weight (t) is used to adjust the impact of different time periods on the disease prediction results, α is the adjustment coefficient, and t0 is the key time point;

[0068] S63. Combine the patient's real-time data fluctuations and historical condition changes to generate future condition predictions:

[0069]

[0070] Among them, P future (t) represents the predicted value of future disease condition, W adj (ti) is the adjusted weight, X(ti) is the patient's historical data, T weight (t) Time weights are introduced to reflect the impact of key time points;

[0071] S64. The model analyzes the long-term data characteristics of patients with diabetic kidney disease, identifies mutation points, generates early warning signals, and provides prediction reports based on time correlation and disease fluctuations;

[0072] S65. Based on the disease prediction report and real-time data feedback, generate intervention recommendations for diabetic nephropathy patients. The model uses the time weighted function T weight (t) is dynamically adjusted to continuously update the forecast report.

[0073] The beneficial effects of the present invention are:

[0074] By introducing a dynamic nonlinear weighted regression algorithm, the present invention can dynamically adjust the weight distribution of different time periods according to the change amplitude, trend and historical data of the patient's condition, and capture the nonlinear changes of the patient's condition. The time-related dynamic adjustment mechanism enhances the real-time and adaptability of the present invention. By analyzing the frequency of condition fluctuations, the present invention can dynamically adjust the data influence of different time periods when the patient's condition changes faster or slower, ensuring sensitive capture of condition fluctuations. Compared with traditional fixed models, the present invention can more flexibly adjust the prediction model in the time dimension, generate a continuous condition assessment curve, and reflect the patient's condition change trend in different time periods. This mechanism greatly improves the accuracy and timeliness of condition prediction, especially in the chronic management of diabetic nephropathy, which helps doctors to adjust treatment strategies in a timely manner.

[0075] The present invention can comprehensively analyze multiple renal function-related data such as blood sugar, urine protein, serum creatinine, etc. through the combination of multi-dimensional data, and generate a more accurate disease prediction report. This multi-dimensional data fusion method overcomes the problem of one-sided evaluation results in the prior art, provides a more comprehensive disease assessment, and generates long-term trend forecasts by adaptively adjusting weights, helping clinicians to intervene in the patient's disease progression in advance to avoid disease deterioration or delayed treatment.

[0076] The present invention also has the advantage of treatment feedback. By dynamically adjusting the weight results and combining the patient's real-time data and long-term medical history, the present invention can generate a customized treatment feedback plan to help doctors adjust treatment strategies in time when the condition fluctuates. This feedback not only improves the accuracy of treatment, but also provides data support for long-term condition management, allowing doctors to tailor the best treatment plan for each patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0078] Figure 1 This is a flow chart of a method for evaluating the effect of non-dialysis diabetic nephropathy proposed by the present invention;

[0079] Figure 2 This is a schematic diagram of disease assessment and weight adjustment based on a dynamic nonlinear weighted regression algorithm;

[0080] Figure 3 Schematic diagram of the dynamic regulation mechanism of time correlation. DETAILED DESCRIPTION

[0081] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0082] refer to Figure 1-3 , a method for evaluating the effect of non-dialysis diabetic nephropathy, comprising the following steps:

[0083] S1. Collect the patient's renal function-related data, and continuously monitor data fluctuations during the collection process to generate a basic data set;

[0084] S2. Use dynamic nonlinear weighted regression algorithm to analyze and process the multidimensional renal function data of patients, adjust the weight of data points, dynamically allocate data weights of different time periods and characteristics based on the magnitude of changes in the condition, trends and correlation with historical data, and capture the nonlinear changes in the patient's condition;

[0085] S3. Introduce a time-related dynamic adjustment mechanism to adjust the influence of data in different time periods in real time based on the frequency of fluctuations in the patient's condition and generate a condition assessment curve;

[0086] S4. Generate a prediction of renal function changes based on the output results of the dynamic nonlinear weighted regression algorithm, and output a disease trend line by combining the patient's multidimensional historical data and current disease fluctuations;

[0087] S5. Based on the prediction results, dynamically adjust the weights to generate a treatment feedback plan, and optimize the treatment strategy by combining the patient's real-time data and long-term medical history;

[0088] S6. Through the long-term prognosis prediction of dynamic nonlinear weighted regression algorithm, combined with time correlation and disease fluctuation trend, a forecast report on future disease development is generated and intervention suggestions are provided.

[0089] In this implementation, S1 specifically includes:

[0090] S11. Collect the patient's renal function-related data, including blood glucose G(t), urine protein P(t), serum creatinine C(t) and estimated glomerular filtration rate E(t), attach a timestamp t to each data point, and generate a preliminary time series data set D t ;

[0091] S12. Based on the preliminary dataset D t , at each time point t, the real-time fluctuation monitoring of the collected physiological indicators is performed, and the fluctuation value of each data point is calculated ΔX(t)=X(t) where X(t-1) is the data at the current time point, and X(t) is the data at the previous time point;

[0092] S13. According to the fluctuation monitoring result ΔX(t), when the fluctuation amplitude exceeds the threshold, the sampling frequency will be adjusted according to the formula f(t) = α×e -β×|ΔX(t)| Dynamic adjustment is used to increase the sampling frequency f(t) when the data fluctuates greatly, shorten the collection time interval Δt, and generate a dense data set D dense ;

[0093] S14. Dense dataset D generated after the time interval is shortened dense Input into the adaptive filtering algorithm to perform noise filtering and smoothing to eliminate outliers and noise interference;

[0094] S15, apply adaptive filtering algorithm to D dense The filter formula is F adaptive (X(t))= where w i is the weight of the first N data points;

[0095] S16, the final processed basic data set D base output through an adaptive filter.

[0096] In this implementation, S2 specifically includes:

[0097] S21. Use dynamic nonlinear weighted regression algorithm to analyze and process the multidimensional renal function data of patients, and combine it with the generated basic data set D base Conduct analysis;

[0098] S22, by adaptively adjusting the weights of data points, combining the changing patterns of the condition and historical data, dynamically assigning weights according to data changes in different time periods and characteristics, and capturing nonlinear changes in the patient's condition;

[0099] S23. Time weighting for disease condition changes:

[0100]

[0101] Among them, ΔX(t) represents the change of the current data point, Δt is the time interval, and λ is the time-related adjustment factor;

[0102] S24, the disease fluctuation detection mechanism is used to identify abnormal fluctuation points and adjust the weight distribution accordingly to identify key disease change trends;

[0103] S25. Combine the time-dependent function T(t) to make long-term trend predictions, analyze the patient's renal function changes by adjusting the time window T(t), and make long-term evaluations and predictions.

[0104] In this implementation, S3 specifically includes:

[0105] S31, introduce time-related dynamic adjustment mechanism to process the multi-dimensional renal function data of patients, based on the basic data set D base And weight W(t) for data processing;

[0106] S32. Adjust the influence of data in different time periods according to the frequency of fluctuations in the patient's condition:

[0107] Wadj(t)=W(t)×(1+μ·f(Δt,B(t)));

[0108] Among them, W adj (t) is the adjusted weight, W(t) is the initial weight at the current time point, μ is the adjustment coefficient, and the function f(Δt, B(t)) represents the function of the time interval Δt and the disease fluctuation frequency B(t), which is used to dynamically adjust the weight distribution. It combines the characteristics of disease fluctuation and time interval changes to achieve dynamic adjustment of weights;

[0109] S33. The calculation of the disease fluctuation frequency B(t) is obtained by analyzing the fluctuation amplitude and trend in the historical data:

[0110]

[0111] Where N is the number of past time points, X(ti) is the data collected at time point ti, and the disease fluctuation frequency B(t) is used to identify sudden and gradual changes in the disease. By analyzing past data fluctuations, the changing trend of the disease can be identified, providing a basis for weight adjustment;

[0112] S34, after calculating the weight, using the dynamic nonlinear weighted regression algorithm according to the adjusted weight W adj (t) Analyze the disease data and generate a continuous disease assessment curve to reflect the change trend of the patient's renal function in different time periods;

[0113] S35. Adjust the weight distribution of the model in real time to enable it to respond flexibly when the condition changes suddenly or gradually, and generate continuous condition assessment results in the short term. The condition fluctuation detection mechanism identifies abnormal fluctuation points, dynamically adjusts their weights, responds promptly to current condition changes, and provides assessments based on short-term fluctuations;

[0114] S36. Combined with the time-dependent function T(t), the patient's current and recent changes in renal function are dynamically evaluated. The model adjusts the time window T(t) to deal with recent fluctuations in the condition and generates an evaluation result of the stable trend of the condition.

[0115] In this implementation manner, the S4 specifically includes:

[0116] S41. Generate a long-term trend prediction of renal function changes based on a dynamic nonlinear weighted regression algorithm, generate a trend line of the patient's condition changes based on the patient's multidimensional historical data, combined with the current condition fluctuation data and short-term and recent evaluation results;

[0117] S42, combining the time-dependent function T(t) and the real-time adjustment mechanism to further generate a long-term disease prediction model:

[0118]

[0119] Among them, P(t) represents the long-term disease prediction result at time t, W adj (ti) is the dynamic adjustment weight at time point ti, X(ti) is the historical data collected at time point ti, δ is the time adjustment parameter, and T(t) is the time-dependent function that reflects the long-term trend of the disease;

[0120] S43. The calculation formula of the disease fluctuation frequency B(t) is:

[0121]

[0122] Where B(t) represents the frequency of disease fluctuation, which reflects the speed of disease change at time t, N represents the number of historical data points used for calculation, X(ti) is the historical data at time point ti, and Δt is the time interval between adjacent time points;

[0123] S44. Combining the fluctuation frequency of the disease and the adaptive adjustment mechanism, the model optimizes the data in different time periods to generate a long-term disease prediction curve to reflect the long-term disease change trend of the patient;

[0124] S45. The generated long-term trend line provides doctors with disease management advice and helps formulate long-term treatment and intervention measures. By combining short-term, recent and long-term disease data, the model can respond to changes in complex diseases and provide a reliable basis for long-term disease prediction.

[0125] In this implementation manner, S5 specifically includes:

[0126] S51. Generate a treatment feedback plan based on the patient's condition prediction results, combine the real-time data R(t) and historical medical history data H(t), comprehensively evaluate the patient's current health status, and generate treatment recommendations;

[0127] S52. Through a dynamic adjustment mechanism, weights are allocated according to the amplitude and frequency of disease fluctuations, using a nonlinear weight adjustment formula:

[0128]

[0129] Among them, W opt(t) represents the optimal weight after adjustment, W(t) is the initial weight, ΔR(t) represents the real-time data increment at the current time point t, ΔH(t) represents the increment of historical medical history data, γ is the weight adjustment coefficient, and β is the nonlinear scaling coefficient of data difference;

[0130] S53, based on the optimized weight W opt (t), combined with the speed of change of the patient's condition, generates customized treatment feedback. The model monitors the fluctuations of the condition in real time and adjusts the treatment plan to cope with the dynamic changes of the condition.

[0131] S54. Combine the disease fluctuation frequency B(t) and the patient's long-term data trend to optimize the long-term treatment plan and adjust the feedback mechanism:

[0132]

[0133] Where B(t) represents the frequency of disease fluctuation, N is the number of data points, Δt is the time interval, and δ is the scaling factor;

[0134] S55. Generate a treatment feedback curve based on feedback data to reflect the patient's treatment progress and provide a basis for adjusting future treatment plans to ensure that the plan can continue to adapt to changes in the patient's condition.

[0135] In this implementation manner, S6 specifically includes:

[0136] S61. Use a dynamic nonlinear weighted regression algorithm to generate a long-term disease prognosis prediction, combine the patient's multidimensional historical data and real-time data, analyze the fluctuation trend of kidney function, and generate a prediction model for future disease development;

[0137] S62, introducing time weighting function T weight (t), assign weights to data in different time periods:

[0138]

[0139] Among them, T weight (t) is used to adjust the impact of different time periods on the disease prediction results, α is the adjustment coefficient, and t0 is the key time point;

[0140] S63. Combine the patient's real-time data fluctuations and historical condition changes to generate future condition predictions:

[0141]

[0142] Among them, P future (t) represents the predicted value of future disease condition, W adj (ti) is the adjusted weight, X(ti) is the patient's historical data, T weight(t) Time weights are introduced to reflect the impact of key time points;

[0143] S64. The model analyzes the long-term data characteristics of patients with diabetic kidney disease, identifies mutation points, generates early warning signals, and provides prediction reports based on time correlation and disease fluctuations;

[0144] S65. Based on the disease prediction report and real-time data feedback, generate intervention recommendations for diabetic nephropathy patients. The model uses the time weighted function T weight (t) is dynamically adjusted to continuously update the forecast report.

[0145] Embodiment 1:

[0146] In order to verify the feasibility of the present invention in implementation, the present invention was applied to a large tertiary hospital, and a one-year clinical trial of disease monitoring and treatment optimization was conducted. A total of 300 patients diagnosed with non-dialysis diabetic nephropathy participated in the trial. These patients had different conditions, but they all faced the same challenge: renal function gradually deteriorated, and traditional monitoring and treatment plans often missed the best time for intervention due to lags or inaccurate predictions. This trial aims to verify whether the method proposed in the present invention can help patients achieve more timely and effective disease management through dynamic monitoring, data analysis and personalized feedback.

[0147] In this experimental scenario, all patients' renal function-related data, including blood glucose, urine protein, serum creatinine and estimated glomerular filtration rate, are remotely collected and transmitted to the hospital's data system. The data of each patient is input into the dynamic nonlinear weighted regression algorithm of the present invention. The system dynamically adjusts the data weight according to the real-time data and the patient's historical condition, and analyzes the changes in the condition in real time. Unlike traditional methods, the present invention can capture subtle changes in the patient's condition when there are minor fluctuations in the patient's condition, and flexibly adjust the influence on data in different time periods according to the time correlation mechanism.

[0148] During the experiment, the blood sugar of a 65-year-old male patient rose sharply from 7.8mmol / L to 13.5mmol / L in just two weeks. Although the urine protein level did not change significantly during this period, the system automatically adjusted the data weights to focus on blood sugar changes. By comparing the real-time data with the patient's historical data over the past two years, the system generated a new condition assessment curve, indicating that the patient may be at risk of early renal function deterioration. Based on this assessment, the system issued a warning to the doctor and recommended an adjustment to the patient's drug dosage. The doctor quickly adjusted the patient's treatment plan and the condition was effectively controlled. This process demonstrates the flexibility and timeliness of the present invention in responding to sudden changes in the disease, while the traditional static model may take a week to make a similar judgment.

[0149] In the case of another female patient, her condition changed relatively slowly, her serum creatinine level remained at around 80 μmol / L, and her eGFR was relatively stable. Through the time-related dynamic adjustment mechanism of the present invention, the system found that her eGFR showed a slow downward trend in the past three years, from 72 mL / min to 65 mL / min, indicating that her renal function was gradually deteriorating. The long-term prediction model generated by the system showed that the patient's eGFR would drop to 60 mL / min in the next six months, indicating that more active intervention was needed. In this case, the system recommended to the doctor to monitor the patient's renal function more frequently and adjusted the treatment plan. Through early intervention, the patient's condition remained relatively stable in the next six months and further deterioration was avoided.

[0150] From the data, among the 300 patients in the trial, about 84% of the patients received early warning of disease changes through the disease monitoring and prediction system of the present invention, and the intervention time when the disease suddenly changed was shortened by 48 hours on average; compared with the traditional static regression model, the dynamic nonlinear weighted regression algorithm can more accurately reflect the complexity of the patient's disease fluctuations. For example, a patient's urine protein level increased from 0.6g / 24h to 1.3g / 24h in three months, and the system generated a new assessment report within 24 hours after detecting the change, and recommended adjusting the patient's drug dosage.

[0151] The accuracy of long-term prognosis prediction is also an advantage of the present invention. Through the time-related adjustment mechanism, the system can generate a long-term disease development trend report to help doctors make intervention decisions before the disease worsens. In a one-year trial, the long-term disease trend prediction of 79% of patients was highly consistent with the actual situation, with an error range of less than 5%, while the error rate of the traditional model was as high as 12%. For example, a patient's eGFR was 58mL / min at the beginning of the trial, and the system predicted that it would drop to 54mL / min after half a year, while the actual result showed that it dropped to 53mL / min, with an error of only 1mL / min.

[0152] Table 1 Clinical trial data of non-dialysis diabetic nephropathy

[0153]

[0154] The data in the table show that the dynamic nonlinear weighted regression algorithm proposed in the present invention has achieved remarkable results in disease assessment and treatment feedback. The initial eGFR of the five patients varied at the beginning of the experiment, ranging from 58 to 72 mL / min. After 6 months of monitoring, the predicted eGFR was highly consistent with the actual eGFR, with an error of 1 mL / min, showing the high-precision performance of the algorithm in disease prediction. The effect of shortening intervention time has also been fully verified. Compared with traditional methods, the treatment intervention time of patients was shortened by an average of 24 to 48 hours, indicating that the real-time monitoring and feedback mechanism of the present invention can capture changes in the disease in a timely manner and provide rapid treatment feedback. The timeliness of such feedback is particularly important in clinical intervention, especially for chronic diseases such as diabetic nephropathy. The frequency of fluctuations in the patient's condition is low, but once a change occurs, a rapid response can significantly improve the prognosis.

[0155] In summary, the application of the present invention in this clinical trial effectively solves many problems of the prior art in disease assessment. It not only improves the accuracy of disease prediction, but also shortens the time of treatment feedback, helping doctors make more personalized treatment decisions based on the dynamic changes of patients' real-time data and historical data. It has significant advantages in the long-term management of patients with diabetic kidney disease, and can achieve accurate monitoring of the disease, timely early warning and personalized treatment feedback, helping patients get the most timely intervention when their condition fluctuates, thereby greatly improving the treatment effect.

[0156] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating the effect of non-dialysis diabetic nephropathy, characterized in that: The steps include: S1. Collect the patient's renal function-related data, and continuously monitor data fluctuations during the collection process to generate a basic data set; S2. Use dynamic nonlinear weighted regression algorithm to analyze and process the basic data set, adjust the weight of data points, dynamically allocate data weights of different time periods and characteristics based on the magnitude of disease changes, trends and correlation with historical data, and capture nonlinear changes in patients' conditions; S3. Introduce a time-related dynamic adjustment mechanism to adjust the influence of data in different time periods in real time based on the frequency of fluctuations in the patient's condition and generate a condition assessment curve; S4. Generate a prediction of renal function changes based on the output results of the dynamic nonlinear weighted regression algorithm, and output a disease trend line by combining the patient's multidimensional historical data and current disease fluctuations; S5. Based on the prediction of renal function changes, dynamically adjust the weights to generate a treatment feedback plan, and optimize the treatment strategy by combining the patient's real-time data and long-term medical history; S6. Through the long-term prognosis prediction of dynamic nonlinear weighted regression algorithm, combined with time correlation and disease fluctuation trend, a forecast report on future disease development is generated and intervention suggestions are provided.

2. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S1 specifically includes: S11. Collect the patient's renal function-related data, including blood glucose G(t), urine protein P(t), serum creatinine C(t) and estimated glomerular filtration rate E(t), attach a timestamp t to each data point, and generate a preliminary time series data set D t ; S12. Based on the preliminary dataset D t , at each time point t, the real-time fluctuation monitoring of the collected physiological indicators is performed, and the fluctuation value of each data point is calculated ΔX(t)=X(t) where X(t-1) is the data at the current time point, and X(t) is the data at the previous time point; S13. According to the fluctuation monitoring result ΔX(t), when the fluctuation amplitude exceeds the threshold, the sampling frequency will be adjusted according to the formula f(t) = α×e -β×|ΔX(t)| Dynamic adjustment is used to increase the sampling frequency f(t) when the data fluctuates greatly, shorten the collection time interval Δt, and generate a dense data set D dense ; S14. Dense dataset D generated after the time interval is shortened dense Input into the adaptive filtering algorithm to perform noise filtering and smoothing to eliminate outliers and noise interference; S15, apply adaptive filtering algorithm to D dense The filter formula is where w i is the weight of the first N data points; S16, the final processed basic data set D base output through an adaptive filter.

3. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S2 specifically includes: S21. Use dynamic nonlinear weighted regression algorithm to analyze and process the multidimensional renal function data of patients, and combine it with the generated basic data set D base Conduct analysis; S22, by adaptively adjusting the weights of data points, combining the changing patterns of the condition and historical data, dynamically assigning weights according to data changes in different time periods and characteristics, and capturing nonlinear changes in the patient's condition; S23. Time weighting for disease condition changes: Among them, ΔX(t) represents the change of the current data point, Δt is the time interval, and λ is the time-related adjustment factor; S24, the disease fluctuation detection mechanism is used to identify abnormal fluctuation points and adjust the weight distribution accordingly to identify key disease change trends; S25. Combine the time-dependent function T(t) to make long-term trend predictions, analyze the patient's renal function changes by adjusting the time window T(t), and make long-term evaluations and predictions.

4. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S3 specifically includes: S31, introduce time-related dynamic adjustment mechanism to process the multi-dimensional renal function data of patients, based on the basic data set D base And weight W(t) for data processing; S32. Adjust the influence of data in different time periods according to the frequency of fluctuations in the patient's condition: Wadj(t)=W(t)×(1+μ·f(Δt,B(t))); Among them, W adj (t) is the adjusted weight, W(t) is the initial weight at the current time point, μ is the adjustment coefficient, and the function f(Δt, B(t)) represents the function of the time interval Δt and the disease fluctuation frequency B(t); S33. The calculation of the disease fluctuation frequency B(t) is obtained by analyzing the fluctuation amplitude and trend in the historical data: Where N is the number of past time points, and X(ti) is the data collected at time point ti; S34, after calculating the weight, using the dynamic nonlinear weighted regression algorithm according to the adjusted weight W adj (t) Analyze disease data and generate continuous disease assessment curves; S35. Adjust the weight distribution of the model in real time to enable it to respond flexibly when the condition changes suddenly or gradually, and generate continuous condition assessment results in the short term. The condition fluctuation detection mechanism identifies abnormal fluctuation points and responds to current condition changes in a timely manner by dynamically adjusting their weights, providing assessment based on short-term fluctuations; S36. Combined with the time-dependent function T(t), the patient's current and recent changes in renal function are dynamically evaluated. The model adjusts the time window T(t) to deal with recent fluctuations in the condition and generates an evaluation result of the stable trend of the condition.

5. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S4 specifically includes: S41. Generate a long-term trend prediction of renal function changes based on a dynamic nonlinear weighted regression algorithm, generate a trend line of the patient's condition changes based on the patient's multidimensional historical data, combined with the current condition fluctuation data and short-term and recent evaluation results; S42, combining the time-dependent function T(t) and the real-time adjustment mechanism to further generate a long-term disease prediction model: Among them, P(t) represents the long-term disease prediction result at time t, W adj (ti) is the dynamic adjustment weight at time point ti, X(ti) is the historical data collected at time point ti, δ is the time adjustment parameter, and T(t) is the time-dependent function that reflects the long-term trend of the disease; S43. The calculation formula of the disease fluctuation frequency B(t) is: Where B(t) represents the frequency of disease fluctuation, which reflects the speed of disease change at time t, N represents the number of historical data points used for calculation, X(ti) is the historical data at time point ti, and Δt is the time interval between adjacent time points; S44. Combining the disease fluctuation frequency and the adaptive adjustment mechanism, the model optimizes the data in different time periods to generate a long-term disease prediction curve; S45. The generated long-term trend line provides doctors with disease management advice and helps formulate long-term treatment and intervention measures.

6. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S5 specifically includes: S61. Generate a treatment feedback plan based on the patient's condition prediction results, combine the real-time data R(t) and historical medical history data H(t), comprehensively evaluate the patient's current health status, and generate treatment recommendations; S62. Through a dynamic adjustment mechanism, weights are allocated according to the amplitude and frequency of disease fluctuations, using a nonlinear weight adjustment formula: Among them, W opt (t) represents the optimal weight after adjustment, W(t) is the initial weight, ΔR(t) represents the real-time data increment at the current time point t, ΔH(t) represents the increment of historical medical history data, γ is the weight adjustment coefficient, and β is the nonlinear scaling coefficient of data difference; S63, based on the optimized weight W opt (t), combined with the speed of change of the patient's condition, generates customized treatment feedback. The model monitors the fluctuations of the condition in real time and adjusts the treatment plan to cope with the dynamic changes of the condition. S64. Combine the disease fluctuation frequency B(t) and the patient's long-term data trend to optimize the long-term treatment plan and adjust the feedback mechanism: Where B(t) represents the frequency of disease fluctuation, N is the number of data points, Δt is the time interval, and δ is the scaling factor; S65. Generate a treatment feedback curve based on the feedback data to reflect the patient's treatment progress.

7. The method for evaluating the effect of non-dialysis diabetic nephropathy according to claim 1, characterized in that: The S6 specifically includes: S71. Use a dynamic nonlinear weighted regression algorithm to generate a long-term disease prognosis prediction, combine the patient's multidimensional historical data and real-time data, analyze the fluctuation trend of kidney function, and generate a prediction model for future disease development; S72, introducing time weighting function T weight (t), assign weights to data in different time periods: Among them, T weight (t) is used to adjust the impact of different time periods on the disease prediction results, α is the adjustment coefficient, and t0 is the key time point; S73. Combine the patient's real-time data fluctuations and historical condition changes to generate future condition predictions: Among them, P future (t) represents the predicted value of future disease condition, W adj (ti) is the adjusted weight, X(ti) is the patient's historical data, T weight (t) Time weights are introduced to reflect the impact of key time points; S74. The model analyzes the long-term data characteristics of patients with diabetic kidney disease, identifies mutation points, generates early warning signals, and provides prediction reports based on time correlation and disease fluctuations; S75. Based on the disease prediction report and real-time data feedback, generate intervention recommendations for diabetic nephropathy patients. The model uses the time weighted function T weight (t) is dynamically adjusted to continuously update the forecast report.

Images