Nephropathy auxiliary diagnosis decision-making method and system based on artificial intelligence

Abstract

The invention discloses a nephropathy auxiliary diagnosis decision-making method and system based on artificial intelligence, and relates to the technical field of nephropathy auxiliary diagnosis, and the method comprises the steps: extracting the multi-dimensional features of nephropathy according to the nephropathy detection type, and building a nephropathy feature data set; determining importance rates of different clinical index types; establishing a nephropathy diagnosis model based on a grey theory objective evaluation algorithm; comprehensively judging whether the nephropathy detection model has credibility and reliability or not; collecting clinical detection data of a patient, and establishing a nephropathy detection data set; and determining a comprehensive nephropathy diagnosis result of the patient and generating a nephropathy diagnosis sheet. The method has the advantages that key diagnosis features are effectively extracted through principal component analysis dimensionality reduction and multi-dimensional feature analysis, the limitation that a traditional method depends on a single index is solved, the deviation degree of patient data and health benchmark is quantified through grey theory correlation analysis, and doctors can visually understand model auxiliary diagnosis decision logic.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary diagnosis of kidney disease, and in particular to a method and system for auxiliary diagnosis of kidney disease based on artificial intelligence. Background Art

[0002] Chronic kidney disease (CKD) and acute kidney injury (AKI) are highly prevalent diseases worldwide. Early diagnosis and targeted intervention are key to improving patient outcomes. Traditional kidney disease diagnosis relies primarily on physician judgment and static measurements of single biomarkers (such as serum creatinine and urine protein). Existing methods often rely on a single test parameter (such as serum creatinine levels) and lack systematic integration of multidimensional clinical data (such as blood biochemistry, urinalysis, and imaging features), resulting in insufficient exploration of potential diagnostic information. Furthermore, dimensional differences in different test parameters and data gaps further complicate analysis. Although artificial intelligence technologies (such as deep learning) have made progress in medical image analysis, their "black box" nature makes it difficult for physicians to understand the underlying basis of model decisions, limiting clinical trust. For example, models based on convolutional neural networks may not clearly explain why a patient's eGFR value is classified as high-risk. Existing systems often generate one-time diagnostic results based on static data, unable to track changes in patient parameters in real time (such as rapid worsening of acute kidney injury). Furthermore, they lack automated review mechanisms for consecutive inconsistent diagnostic results, which can easily lead to missed or misdiagnosed cases. Most studies evaluate model performance solely through accuracy or AUC, neglecting to verify the model's clinical rationale (e.g., whether the contribution of key indicators aligns with medical consensus). This can lead to bias in actual model applications. Existing diagnostic tools are often limited to data entry and output, lacking visual interfaces (e.g., 3D kidney model rendering or dynamic risk trend display), making it difficult for physicians to quickly identify problems and develop personalized treatment plans.

[0003] In recent years, grey system theory has been gradually introduced into the field of medical data analysis due to its advantages in modeling small sample sizes and uncertain data. However, integrating it with multidimensional feature fusion, dynamic parameter optimization, and rigorous clinical validation remains a pressing technical challenge. Therefore, there is an urgent need to develop a kidney disease auxiliary diagnosis system that integrates multi-source data, possesses self-learning capabilities, and meets clinical needs to improve diagnostic efficiency and accuracy. Summary of the Invention

[0004] In order to solve the above technical problems, a method and system for auxiliary diagnosis and decision-making of kidney disease based on artificial intelligence is provided. This technical solution aims to propose a kidney disease auxiliary diagnosis system that integrates multi-source data and combines gray system theory to improve diagnostic efficiency and accuracy.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] An artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease, comprising:

[0007] According to the type of kidney disease detected, the multidimensional features of kidney disease are extracted and a kidney disease feature dataset is established;

[0008] Based on the renal disease characteristic dataset, the importance rates of different clinical indicator types were determined based on the Pearson correlation coefficient;

[0009] According to the importance rate of different clinical indicator types, a kidney disease diagnosis model was established based on the objective evaluation algorithm of grey theory;

[0010] Establish evaluation indicators for the accuracy of kidney disease detection models and comprehensively evaluate whether the kidney disease detection models are credible and reliable;

[0011] Collect the patient's clinical test data and establish a kidney disease test data set according to the clinical test data type of kidney disease;

[0012] Substitute the kidney disease test data set into the kidney disease diagnosis model to determine the patient's comprehensive kidney disease diagnosis results and generate a kidney disease diagnosis form;

[0013] Establish a human-computer interaction platform to receive, store and analyze patients' clinical test data, and display kidney disease diagnosis results through a visual interface.

[0014] Preferably, extracting multidimensional features of kidney disease according to the kidney disease detection type and establishing a kidney disease feature data set specifically includes:

[0015] Fill missing values, remove outliers and normalize the original clinical data;

[0016] Extract multidimensional features from blood biochemical indicators, urine test indicators, imaging features and patient medical history;

[0017] Principal component analysis was used to reduce the dimensionality of multidimensional features, and the principal component features with cumulative contribution rates greater than the analysis threshold were retained to form the kidney disease feature dataset.

[0018] Preferably, determining the importance rates of different clinical indicator types based on the renal disease characteristic dataset and the Pearson correlation coefficient specifically includes:

[0019] The correlation between different clinical indicator types and clinical diagnosis of kidney disease was calculated based on Pearson correlation coefficient;

[0020] Determine empirical coefficients and bias terms for different clinical indicator types based on historical diagnostic data;

[0021] Combining the correlation between different clinical indicator types and clinical diagnosis of kidney disease, as well as the empirical coefficients and bias terms of different clinical indicator types, the importance rates of different clinical indicator types were calculated using the following formula:

[0022]

[0023] Among them, IR j is the importance rate of the jth clinical indicator type, r j is the correlation between the jth clinical indicator type and the clinical diagnosis of kidney disease, x j is the index value of the jth clinical index type, w j is the empirical coefficient of the jth clinical indicator type, b is the bias term, α and β are weight coefficients, and α + β = 1.

[0024] Preferably, the establishment of a kidney disease diagnosis model based on the importance rate of different kidney disease types and the objective evaluation algorithm of grey theory specifically includes:

[0025] The clinical indicator data of healthy people were selected as the benchmark, and the mean or median of each indicator was used to form a reference sequence;

[0026] The actual test data of the patient to be diagnosed are used to form a comparison sequence;

[0027] The test data of all patients within the statistical period and the actual test data of the patients to be diagnosed are obtained to form global data;

[0028] Calculate the absolute value of the difference between each clinical indicator data in the comparison sequence and the reference sequence, which is the indicator difference;

[0029] Find the maximum and minimum values of all indicator differences in the global data and calculate the grey correlation coefficient;

[0030] Based on the objective evaluation algorithm of grey theory, a kidney disease diagnosis model is constructed. The kidney disease diagnosis model is as follows:

[0031] G i =ΣIR j ×γ ij

[0032] Among them, G i is the comprehensive relevance of kidney disease of the i-th patient, γ ij is the grey correlation coefficient of the jth clinical indicator type of the i-th patient.

[0033] Furthermore, an artificial intelligence-based auxiliary diagnosis and decision-making system for kidney disease is proposed, including:

[0034] Data acquisition module: The data acquisition module is used to collect patients' clinical test data from multiple medical devices and information systems in real time, and perform standardized preprocessing on the data to generate a kidney disease feature data set;

[0035] A feature analysis module, connected to the data acquisition module, configured to calculate the correlation between different clinical indicators and kidney disease diagnosis results based on a Pearson correlation coefficient algorithm, and output an importance rate for each clinical indicator type;

[0036] a diagnostic model construction module, the diagnostic model construction module being connected to the feature analysis module, the diagnostic model construction module being configured to use a grey theory objective evaluation algorithm, combined with the importance rate, to perform a correlation analysis on a healthy population reference sequence and a patient test sequence, to generate a kidney disease diagnostic model;

[0037] A model validation module, connected to the diagnostic model construction module, configured to evaluate the accuracy, AUC value, and clinical rationality of the kidney disease diagnostic model through cross-validation, ROC curve analysis, and SHAP interpretability framework, and output a credibility verification result;

[0038] a dynamic diagnosis module, the dynamic diagnosis module being connected to the data acquisition module and the diagnosis model building module and configured to input a real-time updated kidney disease detection data set into a validated kidney disease diagnosis model, calculate a comprehensive kidney disease diagnosis result for the patient, and generate a visual diagnosis sheet including a risk level and treatment recommendations;

[0039] The human-computer interaction platform includes a data storage unit, a communication interface and a visualization terminal. The data storage unit is used to encrypt and store historical diagnostic records and model parameters. The communication interface supports two-way data interaction with hospital information systems, mobile terminals and detection equipment. The visualization terminal is used to display diagnostic results, risk trend charts and explainable analysis reports based on model decisions.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] This paper proposes to integrate multidimensional data such as blood biochemical indicators (such as creatinine and urea nitrogen), urine test indicators (such as proteinuria and urine specific gravity), imaging features (such as kidney volume and cortical thickness), and patient medical history, and combines principal component analysis (PCA) dimensionality reduction with Pearson correlation coefficient screening to effectively extract key diagnostic features, overcoming the limitations of traditional methods that rely on a single indicator.

[0042] Grey theory correlation analysis is used to quantify the degree of deviation between patient data and health benchmarks, and the SHAP interpretability framework is combined to visualize the contribution of key indicators, enabling doctors to intuitively understand the model decision logic. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of the kidney disease auxiliary diagnosis and decision-making method proposed in Example 1 of this solution;

[0044] Figure 2 This is a flow chart of the method for establishing a kidney disease feature dataset proposed in Example 2 of this solution;

[0045] Figure 3 This is a flow chart of the method for determining the importance rates of different clinical indicator types proposed in Example 3 of this solution;

[0046] Figure 4 This is a flow chart of the method for establishing a kidney disease diagnosis model proposed in Example 4 of this solution;

[0047] Figure 5 Provide an architectural diagram of the electronic equipment proposed in this solution;

[0048] Figure 6 Schematic diagram of the computer-readable storage medium structure provided for this solution. DETAILED DESCRIPTION

[0049] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0050] Example 1:

[0051] Reference Figure 1 As shown, an artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease is characterized by comprising:

[0052] According to the type of kidney disease detected, the multidimensional features of kidney disease are extracted and a kidney disease feature dataset is established;

[0053] Based on the renal disease characteristic dataset, the importance rates of different clinical indicator types were determined based on the Pearson correlation coefficient;

[0054] According to the importance rate of different clinical indicator types, a kidney disease diagnosis model was established based on the objective evaluation algorithm of grey theory;

[0055] Establish evaluation indicators for the accuracy of kidney disease detection models and comprehensively evaluate whether the kidney disease detection models are credible and reliable;

[0056] Collect the patient's clinical test data and establish a kidney disease test data set according to the clinical test data type of kidney disease;

[0057] Substitute the kidney disease test data set into the kidney disease diagnosis model to determine the patient's comprehensive kidney disease diagnosis results and generate a kidney disease diagnosis form;

[0058] Establish a human-computer interaction platform to receive, store and analyze patients' clinical test data, and display kidney disease diagnosis results through a visual interface.

[0059] In this embodiment, it is proposed to integrate multidimensional data such as blood biochemical indicators (such as creatinine and urea nitrogen), urine test indicators (such as proteinuria and urine specific gravity), imaging features (such as kidney volume and cortical thickness) and patient medical history, combine principal component analysis (PCA) dimensionality reduction and Pearson correlation coefficient screening, and effectively extract key diagnostic features, thereby overcoming the limitation of traditional methods that rely on a single indicator. Grey theory correlation analysis is used to quantify the degree of deviation between patient data and health benchmarks, realizing the leap from "experience-driven" to "data intelligence-driven" in kidney disease diagnosis, combining high precision, high reliability and clinical practicality, and providing innovative solutions for chronic kidney disease management and acute kidney injury early warning.

[0060] Example 2

[0061] Reference Figure 2 As shown, in this embodiment, according to the kidney disease detection type, the multidimensional features of kidney disease are extracted, and the kidney disease feature data set is established, which specifically includes:

[0062] Fill missing values, remove outliers and normalize the original clinical data;

[0063] Extract multidimensional features from blood biochemical indicators, urine test indicators, imaging features and patient medical history;

[0064] Principal component analysis was used to reduce the dimensionality of multidimensional features, and the principal component features with a cumulative contribution rate greater than the analysis threshold were retained to form the kidney disease feature dataset;

[0065] The specific steps of principal component analysis are as follows:

[0066] The normalized original clinical data are used to construct a standardized data matrix Z, Z = [y ab ] A×B , where y ab is the bth feature in the ath sample, A is the number of samples, and B is the number of features;

[0067] Based on the standardized data matrix Z, construct the covariance matrix R;

[0068]

[0069] Perform eigendecomposition on the covariance matrix to obtain the eigenvalue λ b and the corresponding eigenvector v b ;

[0070] Arrange the eigenvalues in descending order and sort the corresponding eigenvectors simultaneously;

[0071] Combined with eigenvalue λ b and the corresponding eigenvector v b , calculate the cumulative contribution rate U k ;

[0072]

[0073] Determine the cumulative contribution rate U k The minimum k value when it is greater than the analysis threshold;

[0074] Map the standardized data matrix Z to the selected k principal components to obtain the reduced dimension dataset X P ;

[0075] X P =Z·V k

[0076] Among them, V k is a B×k matrix consisting of the first k eigenvectors.

[0077] Example 3

[0078] Reference Figure 3 As shown, in this embodiment, according to the kidney disease characteristic data set and based on the Pearson correlation coefficient, the importance rates of different clinical indicator types are determined specifically including:

[0079] The correlation between different clinical indicator types and clinical diagnosis of kidney disease was calculated based on Pearson correlation coefficient;

[0080] The calculation formula of Pearson correlation coefficient is:

[0081]

[0082] Where r j is the correlation between the jth clinical indicator type and the clinical diagnosis of kidney disease, x ij is the jth index value of the i-th patient, y i is the kidney disease diagnosis result of the i-th patient (e.g. binary label: 0 = healthy, 1 = sick), and x ij and y i The mean of , which is the number of pairs of data points analyzed;

[0083] Based on historical diagnostic data, determine the empirical coefficients and bias terms for different clinical indicator types. In actual applications, it is necessary to add different empirical coefficients to each clinical indicator type based on the impact of different indicators in historical experience. For example, if creatinine is more important than urea nitrogen, then the empirical coefficient of creatinine is higher than that of urea nitrogen.

[0084] Combining the correlation between different clinical indicator types and clinical diagnosis of kidney disease, as well as the empirical coefficients and bias terms of different clinical indicator types, the importance rates of different clinical indicator types were calculated using the following formula:

[0085]

[0086] Among them, IR j is the importance rate of the jth clinical indicator type, r j is the correlation between the jth clinical indicator type and the clinical diagnosis of kidney disease, x j is the index value of the jth clinical index type, w j is the empirical coefficient of the jth clinical indicator type, b is the bias term, α and β are both weight coefficients, and α+β=1, where It is the empirical term of logistic regression. The values of α and β are determined by the amount of data. For example, when the data is sufficient, it depends on statistical results, and α increases. When the data is sparse or noisy, it depends on expert experience correction, and β increases.

[0087] IR j The closer it is to 1, the more critical the indicator is for the diagnosis of kidney disease, which can prompt doctors to pay attention to IR first. j High indicators.

[0088] Example 4:

[0089] Reference Figure 4 As shown, in this embodiment, according to the importance rate of different kidney disease types, based on the objective evaluation algorithm of grey theory, a kidney disease diagnosis model is established, which specifically includes:

[0090] The clinical indicator data of healthy people were selected as the benchmark, and the mean or median of each indicator was used to form a reference sequence;

[0091] The actual test data of the patient to be diagnosed are used to form a comparison sequence;

[0092] The test data of all patients within the statistical period and the actual test data of the patients to be diagnosed are obtained to form global data;

[0093] Calculate the absolute value of the difference between each clinical indicator data in the comparison sequence and the reference sequence, which is the indicator difference;

[0094] Find the maximum and minimum values of all indicator differences in the global data and calculate the grey correlation coefficient;

[0095] The calculation formula of grey relational coefficient is:

[0096]

[0097] Among them, Δ maxis the maximum value of all indicator differences in the global data, Δ min is the minimum value of all indicator differences in the global data, Δ ij is the absolute difference of the jth clinical indicator type of the i-th patient, ρ is the discrimination coefficient, which is usually taken as 0.5 and is used to adjust the sensitivity of the correlation coefficient to the difference;

[0098] Based on the objective evaluation algorithm of grey theory, a kidney disease diagnosis model is constructed. The kidney disease diagnosis model is as follows:

[0099] G i =ΣIR j ×γ ij

[0100] Among them, G i is the comprehensive relevance of kidney disease of the i-th patient, γ ij is the grey correlation coefficient of the jth clinical indicator type of the i-th patient;

[0101] According to the comprehensive correlation G i The risk level is divided into:

[0102] Low risk: G i ∈[0,0.3) (close to the healthy reference sequence).

[0103] Medium risk: G i ∈[0.3,0.7)(some indicators are abnormal).

[0104] High risk: G i ∈[0.7,1] (significantly deviates from healthy data)

[0105] The above G i The risk classification intervals are merely illustrative examples. In some embodiments, the thresholds may be adjusted according to clinical needs. For example, the high-risk threshold may be set to 0.8 to improve diagnostic specificity.

[0106] Embodiment 5:

[0107] In this embodiment, SHAP interpretability analysis is specifically used to comprehensively judge whether the kidney disease detection model has credibility and reliability:

[0108] Calculate the SHAP value of each sample to quantify the contribution of the feature to the prediction result;

[0109] The importance of the overall global features is used to rank key clinical indicators. Based on expert experience, it is determined whether the ranking of key clinical indicators meets the requirements of expert review. For example, key indicators (eGFR, urine protein) must be at the top of the contribution ranking. If the key indicators do not meet the requirements, model optimization must be triggered and the values of α, β, and ρ must be readjusted.

[0110] Example 6:

[0111] In this embodiment, the kidney disease detection data set is substituted into the kidney disease diagnosis model to determine the patient's comprehensive kidney disease diagnosis result and generate a kidney disease diagnosis form as follows:

[0112] Set a window period, usually 7 days, and perform auxiliary diagnosis of the patient's kidney disease at set intervals within the window period;

[0113] Determine whether the patient's three consecutive auxiliary kidney disease diagnosis results meet the preset review conditions. If so, the expert review mechanism is triggered, and the patient's diagnostic file is sent to the doctor's backend. The doctor's backend makes a real diagnosis result and adjusts the kidney disease diagnosis model based on the real diagnosis result. If not, a diagnostic risk report for the patient is generated, which at least includes the patient's risk level for various kidney diseases and the contribution of various indicators to the risk assessment.

[0114] The review conditions include:

[0115] The risk level in any two kidney disease auxiliary diagnosis results spans more than two levels, such as the assessed risk changes from low to high;

[0116] The risk levels of any two of the three consecutive kidney disease auxiliary diagnosis results are different, such as the assessed risk changes from low to medium to low;

[0117] Any of the kidney disease diagnosis results contains clinical indicators that contribute abnormally to risk assessment, such as an excessively low eGFR contribution;

[0118] The adjustment of the kidney disease diagnosis model based on the actual diagnosis results is as follows:

[0119] Automatic parameter tuning based on Bayesian optimization:

[0120] Define the parameter search space:

[0121] α∈[0,1],β∈[0,1],ρ∈[0.1,0.9].

[0122] Constraint: α + β = 1.

[0123] The relationship between loss function and parameters is modeled through Gaussian process.

[0124] Iteratively select parameter combinations;

[0125] The new parameters were applied to 10% of the real-time diagnostic traffic and compared with the old model performance.

[0126] If the accuracy of the new model parameters is improved by ≥2%, the old parameters will be replaced.

[0127] At the same time, a log of each parameter change is recorded to support quick rollback.

[0128] In this embodiment, the diagnosis results are dynamically adjusted as the data is updated to avoid lags, and expert review reduces the risk of misdiagnosis, especially in scenarios with data noise or complex cases.

[0129] Embodiment seven:

[0130] This embodiment proposes an artificial intelligence-based kidney disease auxiliary diagnosis and decision-making system, which is used to implement the artificial intelligence-based kidney disease auxiliary diagnosis and decision-making method proposed in any one of the above-mentioned embodiments 1 to 6, including:

[0131] Data acquisition module: The data acquisition module is used to collect patients' clinical test data from multiple medical devices and information systems in real time, and perform standardized preprocessing on the data to generate a kidney disease feature data set;

[0132] A feature analysis module, connected to the data acquisition module, configured to calculate the correlation between different clinical indicators and kidney disease diagnosis results based on a Pearson correlation coefficient algorithm, and output an importance rate for each clinical indicator type;

[0133] a diagnostic model construction module, the diagnostic model construction module being connected to the feature analysis module, the diagnostic model construction module being configured to use a grey theory objective evaluation algorithm, combined with the importance rate, to perform a correlation analysis on a healthy population reference sequence and a patient test sequence, to generate a kidney disease diagnostic model;

[0134] A model validation module, connected to the diagnostic model construction module, configured to evaluate the accuracy, AUC value, and clinical rationality of the kidney disease diagnostic model through cross-validation, ROC curve analysis, and SHAP interpretability framework, and output a credibility verification result;

[0135] a dynamic diagnosis module, the dynamic diagnosis module being connected to the data acquisition module and the diagnosis model building module and configured to input a real-time updated kidney disease detection data set into a validated kidney disease diagnosis model, calculate a comprehensive kidney disease diagnosis result for the patient, and generate a visual diagnosis sheet including a risk level and treatment recommendations;

[0136] The human-computer interaction platform includes a data storage unit, a communication interface and a visualization terminal. The data storage unit is used to encrypt and store historical diagnostic records and model parameters. The communication interface supports two-way data interaction with hospital information systems, mobile terminals and detection equipment. The visualization terminal is used to display diagnostic results, risk trend charts and explainable analysis reports based on model decisions.

[0137] Furthermore, the method according to the embodiment of the present application can also be used with the aid of Figure 5The electronic device architecture shown in FIG. Figure 5 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the artificial intelligence-based kidney disease auxiliary diagnosis and decision-making method provided in this application. The electronic device 500 may also include a terminal interface 508. Of course, Figure 5 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 5 One or more components of an electronic device are shown.

[0138] Figure 6 This is a schematic diagram of the computer-readable storage medium structure provided by an embodiment of the present application. Figure 6 As shown, a computer-readable storage medium 600 according to one embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the artificial intelligence-based kidney disease auxiliary diagnosis decision method according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0139] In summary, the advantages of the present invention are: the present invention proposes to integrate multidimensional data such as blood biochemical indicators (such as creatinine, urea nitrogen), urine test indicators (such as proteinuria, urine specific gravity), imaging features (such as kidney volume, cortical thickness) and patient medical history, combined with principal component analysis (PCA) dimensionality reduction and Pearson correlation coefficient screening, to effectively extract key diagnostic features, thereby solving the limitation of traditional methods relying on a single indicator; gray theory correlation analysis is used to quantify the degree of deviation between patient data and health benchmarks, and the SHAP interpretability framework is combined to visualize the contribution of key indicators, so that doctors can intuitively understand the model decision logic.

[0140] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the invention as claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease, characterized in that: include: According to the type of kidney disease detected, the multidimensional features of kidney disease are extracted and a kidney disease feature dataset is established; Based on the renal disease characteristic dataset, the importance rates of different clinical indicator types were determined based on the Pearson correlation coefficient; According to the importance rate of different clinical indicator types, a kidney disease diagnosis model was established based on the objective evaluation algorithm of grey theory; Establish evaluation indicators for the accuracy of kidney disease detection models and comprehensively evaluate whether the kidney disease detection models are credible and reliable; Collect the patient's clinical test data and establish a kidney disease test data set according to the clinical test data type of kidney disease; Substitute the kidney disease test data set into the kidney disease diagnosis model to determine the patient's comprehensive kidney disease diagnosis results and generate a kidney disease diagnosis form; Establish a human-computer interaction platform to receive, store and analyze patients' clinical test data, and display kidney disease diagnosis results through a visual interface.

2. The artificial intelligence-based nephropathy auxiliary diagnosis and decision-making method according to claim 1, characterized in that: The extraction of multidimensional features of kidney disease according to the kidney disease detection type and the establishment of a kidney disease feature data set specifically include: Fill missing values, remove outliers and normalize the original clinical data; Extract multidimensional features from blood biochemical indicators, urine test indicators, imaging features and patient medical history; Principal component analysis was used to reduce the dimensionality of multidimensional features, and the principal component features with cumulative contribution rates greater than the analysis threshold were retained to form the kidney disease feature dataset.

3. The artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease according to claim 2, characterized in that: The determination of the importance rates of different clinical indicator types based on the renal disease characteristic dataset and the Pearson correlation coefficient specifically includes: The correlation between different clinical indicator types and clinical diagnosis of kidney disease was calculated based on Pearson correlation coefficient; Determine empirical coefficients and bias terms for different clinical indicator types based on historical diagnostic data; Combining the correlation between different clinical indicator types and clinical diagnosis of kidney disease, as well as the empirical coefficients and bias terms of different clinical indicator types, the importance rates of different clinical indicator types were calculated using the following formula: Among them, IR j is the importance rate of the jth clinical indicator type, r j is the correlation between the jth clinical indicator type and the clinical diagnosis of kidney disease, x j is the index value of the jth clinical index type, w j is the empirical coefficient of the jth clinical indicator type, b is the bias term, α and β are weight coefficients, and α + β = 1.

4. The artificial intelligence-based nephropathy auxiliary diagnosis and decision-making method according to claim 3, characterized in that: The establishment of a kidney disease diagnosis model based on the importance rate of different kidney disease types and the objective evaluation algorithm of grey theory specifically includes: The clinical indicator data of healthy people were selected as the benchmark, and the mean or median of each indicator was used to form a reference sequence; The actual test data of the patient to be diagnosed are used to form a comparison sequence; The test data of all patients within the statistical period and the actual test data of the patients to be diagnosed are obtained to form global data; Calculate the absolute value of the difference between each clinical indicator data in the comparison sequence and the reference sequence, which is the indicator difference; Find the maximum and minimum values of all indicator differences in the global data and calculate the grey correlation coefficient; Based on the objective evaluation algorithm of grey theory, a kidney disease diagnosis model is constructed. The kidney disease diagnosis model is as follows: G i =∑IR j ×γ ij Among them, G i is the comprehensive relevance of nephropathy of the i-th patient, γ ij is the grey correlation coefficient of the jth clinical indicator type of the i-th patient.

5. The artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease according to claim 4, characterized in that: Substituting the kidney disease detection data set into the kidney disease diagnosis model to determine the patient's comprehensive kidney disease diagnosis result and generate a kidney disease diagnosis form specifically includes: Set a window period, and perform auxiliary diagnosis of the patient's kidney disease at set intervals within the window period; Determine whether the patient's three consecutive auxiliary kidney disease diagnosis results meet the preset review conditions. If so, the expert review mechanism is triggered, and the patient's diagnostic file is sent to the doctor's backend. The doctor's backend makes a real diagnosis result and adjusts the kidney disease diagnosis model based on the real diagnosis result. If not, a diagnostic risk report for the patient is generated, which at least includes the patient's risk level for various kidney diseases and the contribution of various indicators to the risk assessment.

6. The artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease according to claim 5, characterized in that: The review conditions include: The risk level span of any two kidney disease auxiliary diagnosis results exceeds two levels; The risk levels of any two of the three consecutive kidney disease auxiliary diagnosis results are different; Any of the kidney disease diagnosis results contains clinical indicators that have an abnormal contribution to risk assessment.

7. An artificial intelligence-based assisted diagnosis and decision-making system for kidney disease, for implementing the artificial intelligence-based assisted diagnosis and decision-making method for kidney disease according to any one of claims 1 to 6, comprising: Data acquisition module: The data acquisition module is used to collect patients' clinical test data from multiple medical devices and information systems in real time, and perform standardized preprocessing on the data to generate a kidney disease feature data set; A feature analysis module, connected to the data acquisition module, configured to calculate the correlation between different clinical indicators and kidney disease diagnosis results based on a Pearson correlation coefficient algorithm, and output an importance rate for each clinical indicator type; a diagnostic model construction module, the diagnostic model construction module being connected to the feature analysis module, the diagnostic model construction module being configured to use a grey theory objective evaluation algorithm, combined with the importance rate, to perform a correlation analysis on a healthy population reference sequence and a patient test sequence, to generate a kidney disease diagnostic model; A model validation module, connected to the diagnostic model construction module, configured to evaluate the accuracy, AUC value, and clinical rationality of the kidney disease diagnostic model through cross-validation, ROC curve analysis, and SHAP interpretability framework, and output a credibility verification result; a dynamic diagnosis module, the dynamic diagnosis module being connected to the data acquisition module and the diagnosis model building module and configured to input a real-time updated kidney disease detection data set into a validated kidney disease diagnosis model, calculate a comprehensive kidney disease diagnosis result for the patient, and generate a visual diagnosis sheet including a risk level and treatment recommendations; The human-computer interaction platform includes a data storage unit, a communication interface and a visualization terminal. The data storage unit is used to encrypt and store historical diagnostic records and model parameters. The communication interface supports two-way data interaction with hospital information systems, mobile terminals and detection equipment. The visualization terminal is used to display diagnostic results, risk trend charts and explainable analysis reports based on model decisions.

8. An electronic device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute an artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease as described in any one of claims 1-6.

9. A computer-readable storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by a processor, an artificial intelligence-based auxiliary diagnosis and decision-making method for kidney disease according to any one of claims 1 to 6 is implemented.

Images