IgA nephropathy prediction model and construction method thereof

Abstract

The invention discloses an IgA nephropathy prediction model and a construction method thereof, relates to the technical field of medical treatment, and aims to evaluate the potential risk of IgA nephropathy by constructing a feature set and measuring whether a kidney mesangial region is in a pathological enhancement state or not. Thirdly, based on the measurement result in the step S1, the hidden risk of the lesion is analyzed, and whether the patient has the high disease progress risk or not is further evaluated by calculating a hidden progress factor. The calculation of hidden progress factors helps identify high risk individuals in mild patients. And if the hidden progress factor exceeds a set threshold value, the system can automatically trigger a prediction mechanism to provide guidance for a subsequent treatment strategy. On the basis, through training of historical physical examination data and a prediction model of a patient, an IgA comprehensive prediction value is further output, a risk level is generated, comprehensive analysis from the historical data to a current physical examination result is achieved in the process, and the accuracy of the prediction result is ensured.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to an IgA nephropathy prediction model and a construction method thereof. Background Art

[0002] As a common primary glomerular disease, IgA nephropathy has become a focus of daily diagnosis and treatment by nephrologists. Early diagnosis and monitoring of IgA nephropathy are crucial to improving patient prognosis and optimizing treatment plans. Research on disease prediction and analysis methods for IgA nephropathy is a means of quantifying the development trend of the disease based on clinical and experimental data, thereby providing doctors with more accurate decision-making basis.

[0003] Although there are some clinical diagnostic methods for IgA nephropathy in the current medical field, such as urine testing, blood immune indicators and renal biopsy, these traditional diagnostic methods still have significant deficiencies in predicting disease progression and early intervention. First, urine and blood analysis are difficult to fully reflect the dynamic changes of pathological deposits, and most methods rely on the appearance of clinical symptoms, which can easily miss the early stages of the disease. These deficiencies are mainly due to the limitations of traditional diagnostic methods, which often rely on relatively basic and static clinical indicators. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides an IgA nephropathy prediction model and a construction method thereof, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented by the following technical solutions: a prediction model for IgA nephropathy and a method for constructing the same, comprising the following steps: S1: Obtain relevant symptoms of the patient to be tested through physical examination, and obtain feature sets after preprocessing. Based on the feature sets, measure whether the deposition is in a pathological enhancement state, and obtain historical data through the physical examination results of the patient to be tested in historical periods; S2: Based on the measurement results in S1, the lesion concealment risk of the patient to be tested is analyzed again to calculate the hidden progression factor LPI. If the hidden progression factor LPI exceeds the set threshold, the prediction mechanism is triggered, and the prediction model is trained in combination with the content of S1; S3: Start the prediction mechanism, output the IgA comprehensive prediction value IGX from the output end of the prediction model, and generate the risk level; S4: Infer intervention recommendations based on risk level to assess the improvement response of the patient's condition.

[0006] Preferably, the specific steps of S1 include: S11: Obtain various physical examination results through various physical examinations, and obtain the feature set of the patient to be tested by cleaning the data and eliminating duplicate values ​​of the physical examination results. The feature set includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD and mesangial deposition grayscale identification factor MES of the patient to be tested in this physical examination. Among them, various physical examinations include urine testing, blood testing and renal puncture biopsy.

[0007] Preferably, the specific step S1 further includes: S12: The specific method for obtaining the urine red blood cell morphology homogeneity index URR is as follows: Fresh urine samples were collected from the patients to be tested, and then centrifuged to prepare urine sediment slides. The red blood cell contours were extracted through microscopic image acquisition to obtain the red blood cell morphological characteristics. Each red blood cell was represented as a morphological feature vector to cluster and evaluate the consistency of the red blood cell morphological characteristics in the entire sample, and finally the urine red blood cell morphology homogeneity index (URR) was constructed.

[0008] Preferably, the specific step S1 further includes: S13: Based on the feature set and after dimensionless processing, the deposition response coefficient RSD of the physical examination of the patient to be tested is calculated to measure whether the deposition is in a pathological enhancement state. The specific calculation method is as follows: ; Among them, URR represents the urine red blood cell morphology homogeneity index, PRF represents the proteinuria fluctuation frequency, IGD represents the serum Gd-IgA1 concentration density value, MES represents the mesangial deposition grayscale identification factor, and e represents the Euler number, which is approximately 2.71828; S14: Retrieving the last physical examination result in the historical data to obtain the deposition response coefficient in the last physical examination result in the historical data , and the deposition response coefficient in the last physical examination result in the historical data Compare the value with the sedimentation response coefficient RSD of the current physical examination of the patient to be tested. If the sedimentation response coefficient RSD of the current physical examination of the patient to be tested exceeds the sedimentation response coefficient in the last physical examination result in the historical data, , it is judged that the current patient to be tested is in a pathological enhancement state, and a disease latent analysis instruction will be issued at this time; if the deposition response coefficient RSD of the current physical examination of the patient to be tested does not exceed the deposition response coefficient in the last physical examination result in the historical data , it is judged that the current patient to be detected is not in a pathological enhancement state.

[0009] Preferably, the specific steps of S2 include: S21: The prediction model is trained by dividing the historical data into a training set and a validation set, and the prediction model is validated by measuring the abnormal fluctuation degree of the training stability. After receiving the disease latent analysis instruction, the lesion latent risk of the patient to be tested is analyzed again to obtain the latent progression factor LPI of the patient to be tested for this physical examination, which is specifically obtained by the following formula: ; in, is to increase the weight of deposition intensity on progress, It is to nonlinearly amplify the abnormal behavior of red blood cells while ensuring that the value range is limited.

[0010] Preferably, the specific step S2 further includes: S22: Preset a threshold value, and compare the latent progression factor LPI with the pre-set threshold value to determine the concealment of the lesion of the patient to be detected. The specific contents are as follows: When the latent progression factor LPI exceeds the preset threshold, it indicates that the lesion of the patient to be tested is latent, and the prediction mechanism will be triggered at this time; When the latent progression factor LPI does not exceed the preset threshold, it indicates that the lesion of the patient to be tested is not yet latent, and the prediction mechanism is not triggered at this time.

[0011] Preferably, the specific steps of S3 include: S31: Pre-collect the physical examination results of the patient to be tested in the historical period to obtain historical data, wherein the historical data includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD, mesangial deposition grayscale identification factor MES, and the deposition response coefficient RSD and hidden progression factor LPI obtained in each physical examination of the patient to be tested in the historical period, and combine the deposition response coefficient RSD and hidden progression factor LPI of the physical examination of the patient to be tested to obtain an input feature set, and input the input feature set into the convolutional neural network model, and after training, use the trained convolutional neural network model as a prediction model to output the IgA comprehensive prediction value IGX from the output end of the prediction model, and the IgA comprehensive prediction value IGX is calculated and obtained according to the following formula: ; in, , is the Sigmoid activation function, used for probability normalization, where x in this formula refers to , represents the mean of all deposition response coefficients in the input feature set, represents the mean of all hidden progress factors in the input feature set, and Represents the mean of all deposition response coefficients in the input feature set and the mean of all hidden progress factors in the input feature set The weight of .

[0012] Preferably, the specific step S3 further includes: S32: By inputting a feature set, setting a risk range, and numerically comparing the risk range with the IgA comprehensive prediction value IGX to generate a corresponding risk level, the specific contents are as follows: If the IgA comprehensive prediction value IGX exceeds the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a high risk level, and a No. 1 intervention instruction will be issued at this time; If the IgA comprehensive prediction value IGX falls into the risk range, it means that the patient to be tested currently has IgA nephropathy at a medium risk level, and a second intervention instruction will be issued; If the IgA comprehensive predicted value IGX does not exceed the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a low risk level, and a third intervention instruction will be issued at this time.

[0013] Preferably, the specific step of S4 includes: S41: When the first intervention instruction and the second intervention instruction are received, the intervention suggestion will be reversed according to the risk level to construct the reversible risk assessment factor RVI; S42: When the reversible risk assessment factor RVI exceeds the pre-set assessment threshold, it means that the deposits and immune response are in the early stages. The patient will be treated by taking immunosuppressive drugs, regular follow-up, and timely adjustment of drugs. If the reversible risk assessment factor RVI does not exceed the pre-set assessment threshold, it means that the deposits have begun to appear. At this time, dialysis treatment will be performed, combined with renal puncture biopsy, to further clarify the pathological type, evaluate the degree of fibrosis and residual renal function. S43: According to the intervention recommendations in S42, the patient is given therapeutic intervention, and after the therapeutic intervention, S1 to S3 are repeated to determine the corresponding risk level generated again, until the third intervention instruction is issued, and the patient is given therapeutic intervention by taking immunosuppressive drugs and adjusting the medication in time.

[0014] Preferably, an IgA nephropathy prediction model is constructed by the aforementioned IgA nephropathy prediction construction method.

[0015] The present invention provides an IgA nephropathy prediction model and a construction method thereof, which have the following beneficial effects: (1) First, the patient's relevant symptoms are obtained through physical examination, and the data is preprocessed on this basis to construct a feature set. Based on these features, the renal mesangial area is measured to determine whether it is in a pathologically enhanced state, thereby evaluating the potential risk of IgA nephropathy. This process uses multiple key indicators, such as urine red blood cell morphology homogeneity index, proteinuria fluctuation frequency, serum Gd-IgA1 concentration density value, and mesangial deposition grayscale identification factor, to achieve comprehensive quantification of deposition activity. Compared with the traditional single indicator method, this fusion of multi-dimensional data provides a more accurate and reliable assessment of deposition reaction. Then, based on the measurement results in S1, the latent risk of the lesion is analyzed, and the latent progression factor LPI is calculated to further evaluate whether the patient has a high risk of disease progression. The calculation of the latent progression factor LPI value helps identify high-risk individuals among mild patients, which is of great significance for early intervention. If the latent progression factor LPI exceeds the set threshold, the system will automatically trigger the prediction mechanism to provide guidance for subsequent treatment strategies. On this basis, through the patient's historical physical examination data and the training of the prediction model, the IgA comprehensive prediction value IGX is further output, and the risk level is generated. This process realizes the comprehensive analysis from historical data to the current physical examination results, ensuring the accuracy of the prediction results. According to the generated risk level, the system can automatically reverse the intervention suggestions, so as to realize the formulation of personalized treatment plans, especially in the case where the patient's condition is mild but the potential risk of progression is large, timely treatment intervention suggestions can be provided to reduce the risk of further deterioration of the condition as much as possible. In general, the IgA nephropathy prediction method of the present invention significantly improves the early diagnosis, treatment predictability and individualized treatment level of IgA nephropathy through the fusion of multi-dimensional data, the accurate assessment of hidden progression risk and the recommendation of personalized intervention strategies, which has important clinical significance for reducing renal function damage and improving the quality of life of patients.

[0016] (2) In step S21, the prediction model is repeatedly trained and verified by dividing the historical data into a training set and a validation set. By measuring the stability of the model and the degree of abnormal fluctuation during the training process, the generalization ability of the model in different data scenarios is ensured. This process enables the model to adapt to the clinical manifestations and physical examination data of different patients, and enhances its ability to judge latent progression. Especially in patients with mild symptoms, those high-risk individuals with mild surface symptoms but active deposition can be effectively identified, so as to carry out early intervention. By calculating the latent progression factor LPI, combined with factors such as deposition intensity, abnormal red blood cell behavior and proteinuria fluctuations, the model can quantify the potential progression risk of the patient's condition.

[0017] (3) Step S4 automatically generates intervention recommendations based on the patient's risk level, and then adjusts the treatment strategy based on the RVI value, providing strong decision support for clinical treatment. In step S41, the reversible risk assessment factor RVI is further constructed by generating intervention instructions based on the patient's risk level. This factor combines the IgA comprehensive predictive value IGX and provides a treatment window by evaluating the patient's deposition status and immune response. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example 1 See also Figure 1 The present invention provides an IgA nephropathy prediction model and a construction method thereof, comprising the following steps: S1: Obtain relevant symptoms of the patient to be tested through physical examination, and obtain feature sets after preprocessing. Based on the feature sets, measure whether the deposition is in a pathological enhancement state, and obtain historical data through the physical examination results of the patient to be tested in historical periods; S2: Based on the measurement results in S1, the lesion concealment risk of the patient to be tested is analyzed again to calculate the hidden progression factor LPI. If the hidden progression factor LPI exceeds the set threshold, the prediction mechanism is triggered, and the prediction model is trained in combination with the content of S1; S3: Start the prediction mechanism, output the IgA comprehensive prediction value IGX from the output end of the prediction model, and generate the risk level; S4: Infer intervention recommendations based on risk level to assess the improvement response of the patient's condition.

[0021] In the embodiment of the present invention, the method not only solves the limitations of traditional detection methods, but also integrates multi-dimensional data (such as urine, serum and imaging, etc.) to achieve the prediction of disease progression and the automatic generation of intervention suggestions. Specifically, step S1 collects relevant symptoms of patients through physical examination, such as urine red blood cell morphology, proteinuria fluctuation frequency and serum Gd-IgA1 concentration, and forms a feature set after preprocessing, and then measures whether the deposition is in a pathologically enhanced state, providing an important basis for subsequent analysis.

[0022] Step S2 analyzes the latent progression factor LPI. If the LPI exceeds the set threshold, the prediction mechanism is triggered to indicate potential risks. Step S3 combines historical data with physical examination results to train the prediction model and generate the IgA comprehensive prediction value IGX and risk level, thereby providing early disease warning.

[0023] For example, if a patient shows mild proteinuria and occult hematuria during a physical examination, this method can accurately measure the pathological state of his renal deposits. If the latent progression factor LPI value is high and the IgA comprehensive prediction value IGX shows that the patient is at medium-to-high risk, the system will prompt that the patient may have a strong risk of latent lesion progression, and recommend further clinical examinations and early interventions, such as immunosuppressive therapy or renal puncture biopsy. Ultimately, through this intelligent prediction method, patients can receive timely intervention before their condition has significantly deteriorated, avoid missing the best time for treatment, reduce the occurrence of end-stage renal disease, and improve patients' quality of life and treatment effects.

[0024] In addition, doctors can also formulate personalized treatment plans for patients at different risk levels based on the risk levels generated by this method, thereby improving the accuracy and effectiveness of treatment. In short, the present invention can provide strong support for early screening and treatment decision-making of IgA nephropathy, reduce the risk of misdiagnosis and missed diagnosis, and improve the survival rate and treatment effect of patients.

[0025] Example 2 Please refer to Figure 1 , specifically: S1 specific steps include: S11: Obtain various physical examination results through various physical examinations, and obtain the feature set of the patient to be tested by cleaning the data and eliminating duplicate values ​​of the physical examination results. The feature set includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD and mesangial deposition grayscale identification factor MES of the patient to be tested in this physical examination. Among them, various physical examinations include urine testing, blood testing and renal puncture biopsy.

[0026] The specific steps of S1 also include: S12: The specific method for obtaining the urine red blood cell morphology homogeneity index URR is as follows: Collect fresh urine samples from the patients to be tested, centrifuge them, prepare urine sediment slides, and extract red blood cell contours through microscopic image acquisition to obtain red blood cell morphological characteristics. Each red blood cell is represented as a morphological feature vector to cluster and evaluate the consistency of red blood cell morphological characteristics in the entire sample. Finally, the urine red blood cell morphology homogeneity index URR is constructed, which is used to characterize the characteristics of pathological hematuria. Among them, proteinuria fluctuation frequency (PRF) represents the number of cycles of abnormal fluctuations in urine protein level per unit time, and is used to reflect the recurrence of the disease course, because IgA nephropathy often has "intermittent mild to moderate proteinuria", which can fluctuate with infection, pressure, etc. The serum Gd-IgA1 concentration density value IGD indicates its aggregation concentration and density distribution in the serum, reflecting the accumulation intensity of immune abnormalities in the body. The Gd-IgA1 concentration is detected by collecting blood samples and using the ELISA specific antibody method; The mesangial deposition grayscale identification factor MES reflects the image grayscale characteristics of immune deposits in the mesangial area in the pathological sections of renal puncture tissue. The higher the grayscale, the greater the deposition density and the more concentrated the IgA complex. The pathological sections (IgA immunohistochemical staining) were obtained through renal puncture, and the mesangial area positioning (segmentation) and the average grayscale of the deposition were extracted to output the image grayscale normalization parameters, which are as follows: ;in, is the grayscale mean, and are the lower and upper limits of the grayscale value respectively.

[0027] The specific steps of S1 also include: S13: Based on the feature set and after dimensionless processing, the deposition response coefficient RSD of the physical examination of the patient to be tested is calculated to measure whether the deposition is in a pathological enhancement state. The specific calculation method is as follows: ; Among them, URR represents the urinary red blood cell morphology homogeneity index, PRF represents the proteinuria fluctuation frequency, IGD represents the serum Gd-IgA1 concentration density value, MES represents the mesangial deposition grayscale identification factor; e represents the Euler number, and the value is approximately 2.71828; The outer root sign in the formula is used to standardize the sedimentary activity to the positive continuous space for subsequent processing; The first combines morphological hematuria abnormalities with proteinuria volatility; Emphasize the dominant role of severe pathological hematuria in the sedimentation reaction, Logarithmic compression transformation of proteinuria frequency was performed to suppress the amplification effect of high-frequency values; The second term combines the IgA immune abnormality concentration with the image deposition grayscale, and the denominator It is a typical Sigmoid transformation, which is used to normalize the mesangial deposition grayscale identification factor MES to (0, 1), emphasizing that the larger the grayscale (the heavier the deposition), the closer the Sigmoid value is to 1; this part plays a role in enhancing the deposition image as a whole.

[0028] S14: Retrieving the last physical examination result in the historical data to obtain the deposition response coefficient in the last physical examination result in the historical data , and the deposition response coefficient in the last physical examination result in the historical data Compare the value with the sedimentation response coefficient RSD of the current physical examination of the patient to be tested. If the sedimentation response coefficient RSD of the current physical examination of the patient to be tested exceeds the sedimentation response coefficient in the last physical examination result in the historical data, , it is judged that the current patient to be tested is in a pathological enhancement state, and a disease latent analysis instruction will be issued at this time; if the deposition response coefficient RSD of the current physical examination of the patient to be tested does not exceed the deposition response coefficient in the last physical examination result in the historical data , it is judged that the current patient to be detected is not in a pathological enhancement state.

[0029] In an embodiment of the present invention, in step S1, relevant symptoms of the patient to be tested are obtained through physical examination, and through data cleaning and deduplication, a feature set including urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD and mesangial deposition grayscale identification factor MES is constructed. This feature set covers urine testing, blood testing and renal puncture biopsy data, comprehensively considers the patient's multi-dimensional clinical manifestations, helps to more comprehensively evaluate the pathological enhancement status, and avoids the risk of misdiagnosis caused by a single indicator.

[0030] For example, suppose a patient shows mild proteinuria during a physical examination, but the morphology of urine red blood cells is consistently abnormal and the concentration of Gd-IgA1 in serum is high. By analyzing these characteristics, the system can quantify the risk of deposition and detect potential IgA nephropathy in a timely manner.

[0031] In step S13, the deposition response coefficient RSD is calculated based on the feature set to measure whether the deposition is in a pathologically enhanced state. The calculation of RSD takes into account the grayscale information of hematuria, proteinuria, immune abnormalities and mesangial deposition. Through dimensionless processing, RSD can effectively measure the activity of deposition, providing a scientific basis for disease progression risk assessment. In short, the IgA nephropathy prediction method of the present invention accurately assesses the patient's disease risk by comprehensively collecting patient physical examination data and combining historical information, and makes timely treatment intervention recommendations, which helps to improve the early diagnosis and personalized treatment of IgA nephropathy and significantly improve the patient's prognosis.

[0032] Example 3 Please refer to Figure 1 , specifically: S2 specific steps include: S21: The prediction model is trained by dividing the historical data into a training set and a validation set, and the prediction model is validated by measuring the abnormal fluctuation degree of the training stability. After receiving the disease latent analysis instruction, the lesion concealment risk of the patient to be tested is analyzed again to enhance the prediction model's recognition ability for individuals with mild clinical manifestations but highly active deposition, so as to obtain the latent progression factor LPI of the patient to be tested in this physical examination, which is specifically obtained by the following formula: ; in, is to increase the weight of deposition intensity on progress, It is to nonlinearly amplify the abnormal behavior of red blood cells while ensuring that the range is limited (the radian value is capped at π / 2); The denominator is added with +0.5 to avoid the denominator being 0, and to compress and regulate the stability of low-frequency proteinuria; This is to avoid drastic changes in the predicted model value and to enhance the convergence of the results when the exponent is too high.

[0033] The specific steps of S2 also include: S22: Preset a threshold value, and compare the latent progression factor LPI with the pre-set threshold value to determine the concealment of the lesion of the patient to be detected. The specific contents are as follows: When the latent progression factor LPI exceeds the preset threshold, it indicates that the lesion of the patient to be tested is latent and has a strong latent progression trend. The higher the potential progression risk, the more likely the prediction mechanism will be triggered. When the latent progression factor LPI does not exceed the preset threshold, it indicates that the lesion of the patient to be tested is not yet latent, and the prediction mechanism is not triggered at this time.

[0034] In an embodiment of the present invention, the IgA nephropathy prediction method provided by the present invention significantly improves the ability to identify the early stages of the disease through the calculation and analysis of the latent progression factor LPI, especially in patients with mild clinical manifestations but active deposition, and can accurately predict their risk of disease progression.

[0035] Specifically, step S2 divides historical data into training sets and validation sets, trains the prediction model based on them, and combines the measurement of abnormal fluctuations during the training process to ensure the stability and accuracy of the model. This process can effectively eliminate data bias and uncertainty, so that the model can adapt to the clinical characteristics of different patients in practical applications. By calculating the latent progression factor LPI, step S2 further enhances the prediction model's ability to identify latent progression, especially for patients whose clinical symptoms are not yet obvious but whose deposition is significant. The calculation formula of LPI combines factors such as deposition intensity, abnormal red blood cell behavior, and proteinuria volatility, so as to quantify the patient's lesion concealment risk and improve the accuracy of prediction. Especially in the stage of pathological enhancement, the increase of the latent progression factor LPI will trigger the prediction mechanism, providing a basis for subsequent treatment and intervention. This mechanism enables the prediction model to not only identify obvious lesions, but also accurately identify patients with active deposition but mild clinical manifestations, providing a scientific basis for early intervention of these high-risk individuals.

[0036] In addition, the threshold setting and risk judgment mechanism in step S2 further enhance the practical application value of the model. By comparing the latent progression factor LPI with the preset threshold, the degree of latent progression of the lesion can be accurately distinguished. When the latent progression factor LPI exceeds the set threshold, the system will automatically trigger the prediction mechanism to remind the doctor to take more active monitoring and intervention measures; conversely, when the latent progression factor LPI does not exceed the threshold, the patient's condition is relatively stable, and the system will not trigger unnecessary prediction instructions. This mechanism not only effectively avoids excessive intervention, but also ensures the rational use of clinical resources. In summary, the present invention provides a new solution for the early prediction and personalized treatment of IgA nephropathy through the accurate calculation and dynamic monitoring of latent progression factors, significantly improves the clinical application value of the prediction model, and can timely detect patients with high lesion potential and conduct precise intervention.

[0037] Example 4 Please refer to Figure 1 , specifically: S3 specific steps include: S31: Pre-collect the physical examination results of the patient to be tested in the historical period to obtain historical data, wherein the historical data includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD, mesangial deposition grayscale identification factor MES, and the deposition response coefficient RSD and hidden progression factor LPI obtained in each physical examination of the patient to be tested in the historical period, and combine the deposition response coefficient RSD and hidden progression factor LPI of the physical examination of the patient to be tested to obtain an input feature set, and input the input feature set into the convolutional neural network model, and after training, use the trained convolutional neural network model as a prediction model to output the IgA comprehensive prediction value IGX from the output end of the prediction model, and the IgA comprehensive prediction value IGX is calculated and obtained according to the following formula: ; in, , is the Sigmoid activation function, used for probability normalization, where x in this formula refers to , represents the mean of all deposition response coefficients in the input feature set, represents the mean of all hidden progress factors in the input feature set, and Represents the mean of all deposition response coefficients in the input feature set and the mean of all hidden progress factors in the input feature set The weight of (which can be obtained through model training); Specifically, the training process includes feature extraction, recognition error evaluation and training error minimization.

[0038] Among them, the trained convolutional neural network model is used as the prediction model. The prediction model mainly includes the following structural components: the convolution layer is used to extract features, the activation layer is used to increase the nonlinear expression ability, the pooling layer is used to reduce the dimension and enhance the position invariance, the flattening layer is used to convert into a vector input to the fully connected layer, the fully connected layer is used to combine high-dimensional features, and the output layer is used to output the final result.

[0039] Using weight fusion and nonlinear activation methods, the output structure is stable and probabilistic. The Sigmoid activation function makes the numerical range of the IgA comprehensive prediction value IGX be between 0 and 1, which has medical probability interpretation.

[0040] The specific steps of S3 also include: S32: By inputting a feature set, setting a risk range, and numerically comparing the risk range with the IgA comprehensive prediction value IGX to generate a corresponding risk level, the specific contents are as follows: If the IgA comprehensive prediction value IGX exceeds the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a high risk level, and a No. 1 intervention instruction will be issued at this time; If the IgA comprehensive prediction value IGX falls into the risk range, it means that the patient to be tested currently has IgA nephropathy at a medium risk level, and a second intervention instruction will be issued; If the IgA comprehensive predicted value IGX does not exceed the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a low risk level, and a third intervention instruction will be issued at this time.

[0041] In an embodiment of the present invention, by combining the patient's historical data with the current physical examination results, a convolutional neural network (CNN) model is used for accurate prediction, which provides effective support for the early diagnosis and personalized treatment of IgA nephropathy. In step S31, first, by pre-collecting the patient's historical physical examination data (such as urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD, mesangial deposition grayscale identification factor MES, deposition response coefficient RSD and hidden progression factor LPI), combined with the results of this physical examination, a comprehensive feature set is constructed. By inputting these features into the convolutional neural network model, the model conducts deep learning of the data during the training process and generates an IgA comprehensive prediction value IGX, which ranges from 0 to 1 and is mapped to a probability value through the Sigmoid activation function, reflecting the probability that the patient has IgA nephropathy.

[0042] In step S32, according to the result of the IgA comprehensive prediction value IGX, the system sets different risk levels. When the IgA comprehensive prediction value IGX value exceeds the high-risk threshold, the system will issue a No. 1 intervention instruction, indicating that the patient is in a high-risk state and needs to take immediate treatment intervention; when the IGX value falls into the medium-risk range, a No. 2 intervention instruction is issued, prompting regular monitoring and adjustment of treatment; and when the IGX value is lower than the low-risk threshold, the system issues a No. 3 intervention instruction, indicating that the patient is in a low-risk state and no major intervention is required. Through this multi-dimensional data fusion and intelligent prediction method, the present invention greatly improves the early prediction accuracy of IgA nephropathy and the personalized judgment of treatment response. It not only provides doctors with accurate risk assessment, but also helps to formulate more accurate intervention strategies, thereby improving the treatment effect and the quality of life of patients.

[0043] Example 5 Please refer to Figure 1 , specifically: S4 specific steps include: S41: When the first and second intervention instructions are received, the intervention suggestions will be inferred according to the risk level. The specific contents include: Construct the reversible risk assessment factor RVI, specifically: ; in, represents the comprehensive prediction value of IgA; This means that the high-risk threshold has not been reached and there is an intervention window; The higher the reversibility risk assessment factor RVI is, the stronger the reversibility is, and the more likely it is to be treated with drugs such as RAS blockers. The reversible risk assessment factor RVI is used to assess whether an individual has an intervention window or potential reversibility; The larger the overall value of , the higher the reversibility (non-deposition-dominated individuals); S42: When the reversible risk assessment factor RVI exceeds the pre-set assessment threshold, it means that the deposits and immune response are in the early stages. The patient should be treated by taking immunosuppressive drugs, such as glucocorticoids (such as prednisone), cyclophosphamide, etc., and regular follow-up, assessing changes in renal function and urine indicators every 3-6 months, and adjusting the drugs in time. If the reversible risk assessment factor RVI does not exceed the pre-set assessment threshold, it means that the deposits have begun to appear and the damage may have extended to the irreversible stage. At this time, dialysis treatment will be performed, and alternative treatments such as hemodialysis or peritoneal dialysis can be considered, combined with renal puncture biopsy to further clarify the pathological type, assess the degree of fibrosis and residual renal function. S43: According to the intervention recommendations in S42, the patient is given therapeutic intervention, and after the therapeutic intervention, S1 to S3 are repeated to determine the corresponding risk level generated again, until the third intervention instruction is issued, and the patient is given therapeutic intervention by taking immunosuppressive drugs and adjusting the medication in time.

[0044] In the embodiment of the present invention, the IgA nephropathy prediction method of the present invention significantly improves the accuracy of the treatment intervention decision of IgA nephropathy by constructing the reversible risk assessment factor RVI and combining the risk level generated by the prediction model in step S4. Through this step, timely monitoring of the patient's condition and formulation of personalized treatment plans can be achieved, thereby effectively intervening in the early stage of the disease and reducing the risk of further deterioration of the disease.

[0045] In step S41, when the No. 1 and No. 2 intervention instructions are received, the system will infer the reversible risk assessment factor RVI based on the patient's IgA comprehensive prediction value IGX and risk level. This factor is adjusted according to the IgA comprehensive prediction value and the set high-risk threshold. If the patient's risk is in the intervention stage, the reversible risk assessment factor RVI value is high, indicating that the patient's condition is still in the reversible stage and can be treated with drugs such as RAS blockers. The improvement of the reversible risk assessment factor RVI helps the system accurately determine whether the patient can slow down the progression of the disease through drug treatment and achieve the intervention effect. In step S42, when the reversible risk assessment factor RVI exceeds the preset threshold, it indicates that the deposits and immune response are in the early stages and are suitable for immunosuppressive drug treatment. For example, drugs such as glucocorticoids (such as prednisone) and cyclophosphamide can effectively slow the progression of the disease. At the same time, the patient's renal function and urine index changes are evaluated every 3-6 months, and the drug regimen is adjusted in time to achieve the best treatment effect.

[0046] When the reversible risk assessment factor RVI does not exceed the threshold, it means that the patient's deposition has begun to appear and the damage is more serious. At this time, dialysis treatment is required, and renal puncture biopsy is combined to clarify the pathological type and degree of fibrosis to ensure timely treatment. In step S43, the system re-executes S1 to S3 according to the follow-up results after treatment, and reviews and re-evaluates the patient's condition to determine whether further adjustment of the treatment plan is needed to ensure that the patient receives the most appropriate intervention. For example: For example, during the first physical examination, patient A had a high IGX value and a high RVI, indicating that his condition was in an early reversible state. The system recommended the use of RAS blockers and regular monitoring of renal function. In the follow-up 3 months later, patient A's renal function remained stable, urine indicators were normal, and the RVI value was still high, indicating that drug treatment was effective and continued maintenance treatment. For patient B, although the RVI value was low in the initial diagnosis, deposition was already apparent, and the system determined that his condition was irreversible. Dialysis treatment was recommended instead, combined with renal puncture biopsy for further evaluation to ensure the accuracy of treatment. Through this precise treatment intervention mechanism, the present invention further enhances the early diagnosis and personalized treatment capabilities of IgA nephropathy patients, which can effectively improve treatment outcomes and reduce renal function damage.

[0047] Specifically, all the aforementioned parameters are dimensionally processed to achieve unit elimination.

[0048] Specifically: an IgA nephropathy prediction model is constructed by the IgA nephropathy prediction construction method.

[0049] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting IgA nephropathy, characterized in that: The following steps are included: S1: Obtain relevant symptoms of the patient to be tested through physical examination, and obtain feature sets after preprocessing. Based on the feature sets, measure whether the deposition is in a pathological enhancement state, and obtain historical data through the physical examination results of the patient to be tested in historical periods; S2: Based on the measurement results in S1, the lesion concealment risk of the patient to be tested is analyzed again to calculate the hidden progression factor LPI. If the hidden progression factor LPI exceeds the set threshold, the prediction mechanism is triggered, and the prediction model is trained in combination with the content of S1; S3: Start the prediction mechanism, output the IgA comprehensive prediction value IGX from the output end of the prediction model, and generate the risk level; S4: Infer intervention recommendations based on risk level to assess the improvement response of the patient's condition.

2. A method for predicting and constructing IgA nephropathy according to claim 1, characterized in that: S1 specific steps include: S11: Obtain various physical examination results through various physical examinations, and obtain the feature set of the patient to be tested by cleaning the data and eliminating duplicate values ​​of the physical examination results. The feature set includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD and mesangial deposition grayscale identification factor MES of the patient to be tested in this physical examination. Among them, various physical examinations include urine testing, blood testing and renal puncture biopsy.

3. A method for predicting and constructing IgA nephropathy according to claim 2, characterized in that: The specific steps of S1 also include: S12: The specific method for obtaining the urine red blood cell morphology homogeneity index URR is as follows: Collect fresh urine samples from the patients to be tested, centrifuge them, prepare urine sediment slides, and extract red blood cell contours through microscopic image acquisition to obtain red blood cell morphological characteristics. Each red blood cell is represented as a morphological feature vector to cluster and evaluate the consistency of red blood cell morphological characteristics in the entire sample, and finally construct the urine red blood cell morphology homogeneity index URR; the specific method of obtaining the proteinuria fluctuation frequency PRF is as follows: ; Wherein, Bs is the number of fluctuations, T is the monitoring period; the serum Gd-IgA1 concentration density value IGD is obtained by using the ELISA specific antibody method; the mesangial deposition grayscale recognition factor MES is obtained by renal puncture to obtain pathological sections, extract the mesangial area positioning and the average deposition grayscale, and output the image grayscale normalization parameters, specifically: ;in, is the grayscale mean, and are the lower and upper limits of the grayscale value respectively.

4. A method for predicting and constructing IgA nephropathy according to claim 3, characterized in that: The specific steps of S1 also include: S13: Based on the feature set and after dimensionless processing, the deposition response coefficient RSD of the physical examination of the patient to be tested is calculated to measure whether the deposition is in a pathological enhancement state. The specific calculation method is as follows: ； Among them, URR represents the urine red blood cell morphology homogeneity index, PRF represents the proteinuria fluctuation frequency, IGD represents the serum Gd-IgA1 concentration density value, MES represents the mesangial deposition grayscale identification factor, and e represents the Euler number, which is approximately 2.71828; S14: Retrieving the last physical examination result in the historical data to obtain the deposition response coefficient in the last physical examination result in the historical data , and the deposition response coefficient in the last physical examination result in the historical data Compare the value with the sedimentation response coefficient RSD of the current physical examination of the patient to be tested. If the sedimentation response coefficient RSD of the current physical examination of the patient to be tested exceeds the sedimentation response coefficient in the last physical examination result in the historical data, , it is judged that the current patient to be tested is in a pathological enhancement state, and a disease latent analysis instruction will be issued at this time; if the deposition response coefficient RSD of the current physical examination of the patient to be tested does not exceed the deposition response coefficient in the last physical examination result in the historical data , it is judged that the current patient to be detected is not in a pathological enhancement state.

5. A method for constructing prediction of IgA nephropathy according to claim 4, characterized in that: The specific steps of S2 include: S21: The prediction model is trained by dividing the historical data into a training set and a validation set, and the prediction model is validated by measuring the abnormal fluctuation degree of the training stability. After receiving the disease latent analysis instruction, the lesion latent risk of the patient to be tested is analyzed again to obtain the latent progression factor LPI of the patient to be tested for this physical examination, which is specifically obtained by the following formula: ； in, is to increase the weight of deposition intensity on progress, It is to nonlinearly amplify the abnormal behavior of red blood cells while ensuring that the value range is limited.

6. A method for constructing prediction of IgA nephropathy according to claim 5, characterized in that: The specific steps of S2 also include: S22: Preset a threshold value, and compare the latent progression factor LPI with the pre-set threshold value to determine the concealment of the lesion of the patient to be detected. The specific contents are as follows: When the latent progression factor LPI exceeds the preset threshold, it indicates that the lesion of the patient to be tested is latent, and the prediction mechanism will be triggered at this time; When the latent progression factor LPI does not exceed the preset threshold, it indicates that the lesion of the patient to be tested is not yet latent, and the prediction mechanism is not triggered at this time.

7. A method for predicting and constructing IgA nephropathy according to claim 6, characterized in that: The specific steps of S3 include: S31: Pre-collect the physical examination results of the patient to be tested in the historical period to obtain historical data, wherein the historical data includes the urine red blood cell morphology homogeneity index URR, proteinuria fluctuation frequency PRF, serum Gd-IgA1 concentration density value IGD, mesangial deposition grayscale identification factor MES, and the deposition response coefficient RSD and hidden progression factor LPI obtained in each physical examination of the patient to be tested in the historical period, and combine the deposition response coefficient RSD and hidden progression factor LPI of the physical examination of the patient to be tested to obtain an input feature set, and input the input feature set into the convolutional neural network model, and after training, use the trained convolutional neural network model as a prediction model to output the IgA comprehensive prediction value IGX from the output end of the prediction model, and the IgA comprehensive prediction value IGX is calculated and obtained according to the following formula: ； in, , is the Sigmoid activation function, used for probability normalization, where x in this formula refers to , represents the mean of all deposition response coefficients in the input feature set, represents the mean of all hidden progress factors in the input feature set, and Represents the mean of all deposition response coefficients in the input feature set and the mean of all hidden progress factors in the input feature set The weight of .

8. A method for constructing prediction of IgA nephropathy according to claim 7, characterized in that: The specific steps of S3 also include: S32: By inputting a feature set, setting a risk range, and numerically comparing the risk range with the IgA comprehensive prediction value IGX to generate a corresponding risk level, the specific contents are as follows: If the IgA comprehensive prediction value IGX exceeds the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a high risk level, and a No. 1 intervention instruction will be issued at this time; If the IgA comprehensive prediction value IGX falls into the risk range, it means that the patient to be tested currently has IgA nephropathy at a medium risk level, and a second intervention instruction will be issued; If the IgA comprehensive predicted value IGX does not exceed the risk range, it means that the patient to be tested currently suffers from IgA nephropathy and is at a low risk level, and a third intervention instruction will be issued at this time.

9. A method for predicting and constructing IgA nephropathy according to claim 8, characterized in that: The specific steps of S4 include: S41: When the No. 1 intervention instruction and the No. 2 intervention instruction are received, the intervention suggestion will be reversed according to the risk level to construct the reversible risk assessment factor RVI, which is as follows: ;in, represents the comprehensive prediction value of IgA; S42: When the reversible risk assessment factor RVI exceeds the pre-set assessment threshold, it means that the deposits and immune response are in the early stages. The patient will be treated by taking immunosuppressive drugs, regular follow-up, and timely adjustment of drugs. If the reversible risk assessment factor RVI does not exceed the pre-set assessment threshold, it means that the deposits have begun to appear. At this time, dialysis treatment will be performed, combined with renal puncture biopsy, to further clarify the pathological type, evaluate the degree of fibrosis and residual renal function. S43: According to the intervention recommendations in S42, the patient is given therapeutic intervention, and after the therapeutic intervention, S1 to S3 are repeated to determine the corresponding risk level generated again, until the third intervention instruction is issued, and the patient is given therapeutic intervention by taking immunosuppressive drugs and adjusting the medication in time.

10. A prediction model for IgA nephropathy, characterized in that: The method is constructed by a method for predicting IgA nephropathy according to any one of claims 1 to 9.

Images