Method and device for predicting diabetic nephropathy, electronic equipment and storage medium

Abstract

The invention provides a method and device for predicting diabetic nephropathy, electronic equipment and a storage medium. The method comprises the following steps: acquiring questionnaire data of a target diabetic patient, and a first fundus image and a second fundus image of the same type; constructing a target Bayesian network according to the questionnaire data; based on the target Bayesian network, calculating joint probability distribution of the target diabetic patient concurrent diabetic nephropathy to obtain a first probability value; performing difference feature extraction on the first fundus image and the second fundus image to obtain target difference feature data; performing concurrent diabetic nephropathy probability prediction on the target difference feature data to obtain a second probability value; and according to the first probability value and the second probability value, determining a target probability of complicating diabetic nephropathy of the target diabetic patient. The invention aims to construct a simple, economic and low-trauma early screening system for diabetic nephropathy.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a method, device, electronic device and storage medium for predicting diabetic nephropathy. Background Art

[0002] At present, the screening of diabetic nephropathy mainly relies on the determination of urine microalbumin / creatinine ratio. However, the level of urine microalbumin is affected by many factors, such as blood sugar control level, hypertension, kidney disease, body position changes, inflammatory response, pregnancy, high-protein diet and drugs. These factors may lead to the instability of urine microalbumin test results, thereby affecting the accuracy and reliability of screening. Therefore, there are certain limitations in relying solely on urine microalbumin / creatinine ratio for diabetic nephropathy screening.

[0003] In addition, the current diagnosis and follow-up of diabetic nephropathy usually requires patients to undergo a series of cumbersome and costly examinations, including blood and urine tests, renal ultrasound, CT / MRI imaging examinations, and renal puncture biopsy, etc. Although these examinations can provide important clinical information, their complicated procedures, high costs, and potential trauma often lead to poor patient compliance. Some patients even refuse invasive examinations such as renal puncture biopsy due to fear of pain or trauma.

[0004] In view of this, there is an urgent need for a convenient, economical, low-invasive and efficient early screening method for diabetic nephropathy that can perform accurate risk assessment based on the patient's clinical data and guide the necessity and direction of follow-up examinations, thereby significantly improving the early diagnosis rate of diabetic nephropathy and avoiding the high cost, complex procedures and patient compliance issues brought about by traditional diagnostic methods. Summary of the invention

[0005] In view of the above challenges and needs, the present invention is committed to providing a diabetic nephropathy prediction method based on advanced artificial intelligence technology and its supporting equipment, aiming to build a simple, economical and low-traumatic early screening system.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for predicting diabetic nephropathy, the method comprising: obtaining questionnaire data, a first fundus image and a second fundus image of the same type of a target diabetic patient; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data comprises answer values ​​corresponding to a plurality of diabetic nephropathy-related factors; constructing a target Bayesian network based on the questionnaire data; wherein the target Bayesian network is used to indicate the correlation between a plurality of diabetic nephropathy-related factors in the questionnaire data and the diagnosis of concurrent diabetic nephropathy; based on the target Bayesian network, calculating the joint probability distribution of the target diabetic patient's concurrent diabetic nephropathy to obtain a first probability value; performing difference feature extraction on the first fundus image and the second fundus image to obtain target difference feature data; performing probability prediction of concurrent diabetic nephropathy on the target difference feature data to obtain a second probability value; and determining the target probability of the target diabetic patient's concurrent diabetic nephropathy based on the first probability value and the second probability value.

[0007] In a second aspect, an embodiment of the present disclosure provides an apparatus for predicting diabetic nephropathy, the apparatus comprising: an acquisition module, for acquiring questionnaire data, a first fundus image and a second fundus image of the same type of a target diabetic patient; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data comprises answer sheet values ​​corresponding to a plurality of diabetic nephropathy-related factors; a construction module, for constructing a target Bayesian network based on the questionnaire data; wherein the target Bayesian network is used to indicate the correlation between a plurality of diabetic nephropathy-related factors in the questionnaire data and the diagnosis of concurrent diabetic nephropathy; a calculation module, for calculating the joint probability distribution of concurrent diabetic nephropathy in the target diabetic patient based on the target Bayesian network, to obtain a first probability value; an extraction module, for performing difference feature extraction on the first fundus image and the second fundus image, to obtain target difference feature data; a prediction module, for performing probability prediction of concurrent diabetic nephropathy on the target difference feature data, to obtain a second probability value; a determination module, for determining the target probability of concurrent diabetic nephropathy in the target diabetic patient based on the first probability value and the second probability value.

[0008] In a third aspect, an embodiment of the present disclosure provides an electronic device, including a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the above-mentioned method for predicting diabetic nephropathy.

[0009] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the above-mentioned method for predicting diabetic nephropathy.

[0010] The embodiments of the present disclosure bring the following beneficial effects: The above-mentioned method, device, electronic device and storage medium for predicting diabetic nephropathy compare and analyze fundus images collected from diabetic patients at different examination cycles, collect and analyze data from questionnaires of diabetic patients, and predict the probability of diabetic patients developing diabetic nephropathy, which can provide a simple, economical and low-invasive early screening system. This method aims to provide an unprecedented, convenient, economical and low-invasive early screening approach for diabetic nephropathy through efficient data mining and risk assessment strategies. This solution carefully integrates multi-dimensional data of diabetic patients, and uses advanced machine learning algorithms for deep learning and precise prediction, aiming to significantly improve the early identification rate of diabetic nephropathy, significantly reduce the diagnosis cost of patients, and provide a scientific basis for subsequent personalized medical decision-making, ultimately achieving optimal allocation of medical resources and precise disease management.

[0011] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or understood by practicing the present disclosure. The purpose and other advantages of the present disclosure are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0012] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1 This is a flow chart of an embodiment of a method for predicting diabetic nephropathy in an embodiment of the present disclosure; Figure 2 A schematic diagram of a device for predicting diabetic nephropathy provided in an embodiment of the present disclosure; Figure 3 A schematic diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0016] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] For ease of understanding, the specific process of the embodiment of the present disclosure is described below. Figure 1 , one embodiment of the method for predicting diabetic nephropathy in the embodiments of the present disclosure includes: Step S10, obtaining questionnaire data, a first fundus image and a second fundus image of the same type of the target diabetic patient; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data includes answer values ​​corresponding to a plurality of diabetic nephropathy-related factors; It can be understood that the target diabetic patients are patients who have been diagnosed with diabetes. According to the classification of diabetes, the embodiments of the present disclosure can predict diabetic nephropathy for patients with type 1 diabetes, type 2 diabetes, gestational diabetes and other types of diabetes. The questionnaires prepared for patients with different types of diabetes are also different, and the questionnaire data obtained are also correspondingly different.

[0018] It should be noted that the target diabetic patient can periodically perform the diabetic nephropathy prediction method provided by the embodiment of the present disclosure, and provide the questionnaire data collected in the current cycle and the second fundus image before the end of each examination cycle, so as to timely identify patients with high risk of concurrent diabetic nephropathy through regular tracking and follow-up, and timely intervene in treatment. For example, the target diabetic patient can undergo regular diabetic nephropathy prediction examinations every 3-6 months, and the specific period is not limited here.

[0019] In this embodiment, the questionnaire data is the data obtained after the target diabetic patient fills out the target questionnaire, which is the answer sheet of the target questionnaire, including the answer values ​​corresponding to a variety of diabetic nephropathy associated factors. Each question in the questionnaire can contain at least one diabetic nephropathy associated factor. The diabetic nephropathy associated factor is a factor that has a causal relationship with diabetic nephropathy, and can include physiological parameters, medical history parameters, family genetic history parameters, lifestyle parameters, clinical test parameters, case parameters, etc. The answer values ​​of some diabetic nephropathy associated factors of the target diabetic patient can be selectively provided. The more answer values ​​of diabetic nephropathy associated factors provided, the higher the accuracy of diabetic nephropathy prediction. Conversely, the higher the convenience provided to diabetic patients.

[0020] As an example but not limitation, the physiological parameters in the diabetic nephropathy-related factors may include the target diabetic patient's age, gender, fatigue, edema (such as swollen eyelids, swollen feet, etc.), frequent urination, foam in urine, itchy skin, loss of appetite or nausea and vomiting, high blood pressure, thirst and polydipsia, muscle cramps, and dyspnea, etc., which are physiological characteristics / symptoms that have a causal relationship with diabetic nephropathy. It should be noted that different types of diabetes may correspond to different parameters, which will not be described in detail here.

[0021] In one embodiment, the medical history parameters in the diabetic nephropathy associated factors may include the target diabetic patient's personal medical history parameters (such as whether he or she has hypertension, hyperlipidemia, heart disease, stroke, etc.), drug use history (such as whether he or she takes hypoglycemic drugs, antihypertensive drugs, lipid-lowering drugs, etc.) and other medical history parameters that have a causal relationship with diabetic nephropathy, and the specifics are not limited here.

[0022] As an example but not limitation, the family history parameters in the diabetic nephropathy-associated factors may include whether the target diabetic patient's immediate relatives have diabetes, hypertension, kidney disease, and other family history parameters that are causally related to diabetic nephropathy, but are not limited to the specifics here.

[0023] In one embodiment, the lifestyle habit parameters in the diabetic nephropathy associated factors may include the target diabetic patient's eating habit parameters, exercise habit parameters, work and rest habit parameters, whether or not the patient has a history of smoking, a history of long-term drinking, whether or not the patient has a habit of staying up late, and other lifestyle habit parameters that have a causal relationship with diabetic nephropathy, and the specifics are not limited here.

[0024] As an example but not limitation, the clinical test parameters in the diabetic nephropathy-related factors may include the target diabetic patient's blood pressure test results, urine protein test results, blood test results, blood lipid test results and other clinical indicator data that have a causal relationship with diabetic nephropathy, which are not specifically limited here. It should be noted that the clinical test parameters are historical inspection data, and may also be the inspection results of the current inspection cycle, or the inspection results of the historical inspection cycle and the current inspection cycle, which are not specifically limited here. In one embodiment, the clinical test parameters of the target diabetic patient can be automatically read from the case parameters of the target diabetic patient, or they can be filled in and uploaded by themselves, which is highly flexible.

[0025] Furthermore, if the clinical test parameters include the test results of the target diabetic patients in the historical examination cycle and the current examination cycle, then the differences in the various indicators in the test results of the target diabetic patients can be calculated based on the clinical test parameters, and the differences can be used as the basis for the subsequent construction of the target Bayesian network. Specifically, the target Bayesian network is constructed using the differences as the node values ​​of the corresponding indicators, which can more accurately predict diabetic nephropathy.

[0026] In one embodiment, the case parameters in the diabetic nephropathy associated factors may include case data uploaded by the target diabetic patient himself or herself, or may include case data of the target diabetic patient obtained through data mining technology. The case data may include examination and treatment records of the target diabetic patient in different hospitals, which are obtained with the authorization of the target diabetic patient to ensure the security and legality of the data.

[0027] In this embodiment, the first fundus image and the second fundus image are fundus images of the same type, and the types include but are not limited to: fundus images, fluorescent angiography images, and optical coherence tomography images, wherein the fluorescent angiography images include fluorescein fundus angiography (FFA) images and indocyanine green (ICG) angiography images, and the optical coherence tomography images include optical coherence tomography (OCT) images and optical coherence tomography vascular (OCTA) imaging images, which are not specifically limited here.

[0028] It should be noted that the first fundus image and the second fundus image are fundus images of the target diabetic patients at different periods, and the acquisition time of the first fundus image and the second fundus image differs by at least one inspection cycle. Specifically, the inspection cycle can be set by yourself, usually 3-6 months, and is not limited here.

[0029] Step S20, constructing a target Bayesian network based on the questionnaire data; wherein the target Bayesian network is used to indicate the correlation between multiple diabetic nephropathy-related factors in the questionnaire data and the diagnosis of concurrent diabetic nephropathy; It should be noted that Bayesian Network, also known as Bayesian Directed Acyclic Graph (BDAG) or Probabilistic Dependence Network, is a probabilistic graph model that represents the probabilistic dependency relationship between variables through a directed acyclic graph. In the medical field, for the prediction of diabetic nephropathy in diabetic patients, Bayesian Network can effectively model the complex relationship between various influencing factors and give more accurate prediction results.

[0030] It can be understood that the target Bayesian network is constructed through a preset Bayesian network model, which is generated based on sample data of a large number of diabetic patients, wherein the sample data covers the above-mentioned various diabetic nephropathy-related factors, thereby improving the prediction accuracy of diabetic nephropathy.

[0031] In one embodiment, when constructing a target Bayesian network based on questionnaire data, multiple diabetic nephropathy-associated factors are first extracted from the questionnaire data, and these diabetic nephropathy-associated factors and the diagnosis of diabetic nephropathy are used as nodes of the target Bayesian network. Based on prior knowledge, a directed edge connection structure of the target Bayesian network is constructed to obtain the target Bayesian network. Among them, two nodes with directed edge connection relationships are nodes with association / causal relationships, each node represents a diabetic nephropathy-associated factor, and the terminal node represents the diagnosis of diabetic nephropathy.

[0032] In one embodiment, before step S10, it also includes generating a target Bayesian network model, that is, the above-mentioned preset Bayesian network model. Specifically, the process of generating the target Bayesian network model includes: obtaining sample data of multiple diabetic patients, cleaning and preprocessing the sample data, including processing missing values ​​and outliers, and discretizing or standardizing continuous variables and other cleaning and preprocessing steps to obtain standard sample data.

[0033] Next, a basic Bayesian network is constructed using standard sample data. Specifically, the basic Bayesian network can be constructed using a pre-trained structural learning model / algorithm, or the opinions of medical experts can be combined to determine different diabetic nephropathy-associated factors, as well as the correlation / causal relationship between different diabetic nephropathy-associated factors and diabetic nephropathy. It can also be a combination of the two (structural learning model / algorithm and the opinions of medical experts) to obtain a basic Bayesian network.

[0034] Furthermore, the basic Bayesian network can be tuned. Specifically, preset scoring functions such as Bayesian information criterion (BIC), minimum description length (MDL), Akaike Information Criterion (AIC), cross-validation function, and scoring function based on likelihood ratio test can be used to extract or adjust the basic Bayesian network to obtain a second Bayesian network with a causal relationship structure that is better than the basic Bayesian network, thereby improving the prediction accuracy of diabetic nephropathy.

[0035] Next, according to the second Bayesian network, the conditional probability distribution of different diabetic nephropathy-related factors and confirmed diabetic nephropathy can be calculated to obtain conditional probability distribution data for the calculation of the first probability value. Specifically, the conditional probability distribution of different diabetic nephropathy-related factors and confirmed diabetic nephropathy can be calculated by Bayesian Estimation algorithm, Expectation Maximization (EM) algorithm, Maximum a Posteriori (MAP) probability estimation algorithm, Empirical Bayes algorithm, Variational Bayes Inference algorithm, etc., which are not limited here.

[0036] Step S30, calculating the joint probability distribution of the target diabetic patient complicated with diabetic nephropathy based on the target Bayesian network to obtain a first probability value; It is understandable that the core of the Bayesian network is to decompose the complex joint probability distribution into the product of a series of conditional probability distributions. For a target Bayesian network containing n nodes, its joint probability distribution can be expressed as: ; in, is the joint probability distribution of all variables, including all diabetic nephropathy-associated factors in the target Bayesian network, Is a node The conditional probability distribution of a node given its parent.

[0037] As an example but not a limitation, assume that a target Bayesian network contains nodes A, B, C, and D, where A, B, and C are nodes of diabetic nephropathy-related factors, and D is a node for confirming diabetic nephropathy. Assuming that A represents urine protein, B represents blood creatinine, and C represents hypertension, then the joint probability distribution of the target diabetic patient can be expressed as: P(A, B, C, D) = P(A) * P(B) * P(C) * P(D|A, B) Among them, the conditional probability distribution can be queried from the conditional probability distribution data. The following Table 1 is a schematic table of conditional probability distribution: Table 1 Urine protein (A) Serum creatinine (B) Hypertension (C) Probability of diabetic nephropathy (D=yes) Low Low no 0.05 Low Low yes 0.1 Low middle no 0.15 Low middle yes 0.25 Low high no 0.3 Low high yes 0.4 middle Low no 0.2 middle Low yes 0.35 middle middle no 0.4 middle middle yes 0.6 middle high no 0.5 middle high yes 0.7 high Low no 0.35 high Low yes 0.5 high middle no 0.5 high middle yes 0.75 high high no 0.6 high high yes 0.85 According to Table 1, the answer value corresponding to each diabetic nephropathy-related factor can be queried to obtain the final probability of diabetic nephropathy. For example, assuming that the target diabetic patient has "medium" urine protein, "high" blood creatinine, and "no" hypertension, then the first probability value of the target diabetic patient is 0.5, which is not limited here.

[0038] It should be noted that the conditional probability distribution in actual applications is more complex than that in Table 1. The conditional probability distribution is precise data obtained through modeling calculations. By calculating the joint probability distribution of target diabetic patients with diabetic nephropathy through the conditional probability distribution, an accurate first probability value can be obtained, thereby improving the prediction accuracy of diabetic nephropathy.

[0039] Step S40, extracting difference features from the first fundus image and the second fundus image to obtain target difference feature data; It is understandable that the microcirculation of the eye is one of the windows that reflects the health of the renal blood vessels. The anatomical structure of the eye has many similarities with the kidney. Therefore, based on the difference between the first fundus image and the second fundus image, the changes in the ocular microcirculation of the target diabetic patient in different examination cycles can be reflected, thereby reflecting whether the changing trend of the kidney has the risk of concurrent diabetic nephropathy. It can identify subtle differences that cannot be identified by the naked eye, making the accuracy of diabetic nephropathy prediction higher than manual naked eye identification.

[0040] In one embodiment, the difference feature extraction methods corresponding to different types of fundus images are also different. For different types of fundus images, the corresponding difference feature extraction models are pre-trained to make the difference features of fundus images collected in different examination cycles more accurate in advance. It should be noted that allowing diabetic patients to provide different types of fundus images for probability prediction of concurrent diabetic nephropathy brings great convenience to diabetic patients. Patients can comprehensively consider the cost and difficulty of the examination and choose the type of fundus image for examination. Regardless of the type of fundus image, it is non-invasive and the patient's compliance is better.

[0041] It should be noted that the target difference feature data can indicate the positive change, negative change, and non-directional change of different fundus image indicators. The fundus image indicators corresponding to different types of fundus images are also different. The fundus image indicators are the characteristics that can be reflected by the corresponding type of fundus images and have a strong correlation with diabetic nephropathy. As a rough example of fundus image indicators, for fundus images of fundus image type, blood vessels, optic discs, and macula are the characteristics that can be reflected by fundus images and have a strong correlation with diabetic nephropathy, that is, the fundus image indicators corresponding to the fundus image type. For fundus images of fluorescent angiography image type, the circulation properties of blood vessels and choroidal blood vessels are the characteristics that can be reflected by fundus images and have a strong correlation with diabetic nephropathy, that is, the fundus image indicators corresponding to the fluorescent angiography image type. For fundus images of optical coherence tomography image type, the structures of various layers of the retina and microvessels are the characteristics that can be reflected by fundus images and have a strong correlation with diabetic nephropathy, that is, the fundus image indicators corresponding to the optical coherence tomography image type.

[0042] In practical applications, fundus image indicators can be set more finely. For example, for fundus images of optical coherence tomography image type, retinal blood flow density, foveal avascular area, choroidal blood flow density, retinal thickness, choroidal thickness, etc. can be used as more detailed fundus image indicators, and the specifics are not limited here.

[0043] It can be understood that positive changes in fundus imaging indicators include but are not limited to positive changes such as addition, diffusion, increase, and enlargement, such as the addition and diffusion of blood vessels, increased fluidity of choroidal blood vessels, and increased retinal blood flow density; while negative changes include but are not limited to reverse changes such as decrease, reduction, and contraction, such as the disappearance and shrinkage of blood vessels, reduced fluidity of choroidal blood vessels, and reduced retinal blood flow density; non-directional changes may include structural changes such as optic disc and macula, changes in the structure of various layers of the retina, changes in microvascular structure, and color changes, etc., which are not limited here.

[0044] Step S50, predicting the probability of concurrent diabetic nephropathy on the target difference feature data to obtain a second probability value; It is understandable that after obtaining the target difference feature data, the extracted target difference feature data can be mapped to the probability of diabetic nephropathy through a classifier, thereby obtaining a second probability value. Specifically, the probability of diabetic nephropathy corresponding to the target difference feature data can be calculated by using a fully connected layer and a softmax function as a classifier to obtain a second probability value.

[0045] In one embodiment, the corresponding fourth confidence can be queried from the preset confidence relationship table according to the type of the first fundus image and the second fundus image, and the difference feature extraction model corresponding to the type of the first fundus image and the second fundus image, for calculation of the target probability. It should be noted that the larger the fourth confidence is, the higher the credibility of the second probability value is, and the closer it is to the target probability. The preset confidence relationship table is used to reflect the fourth confidence corresponding to different fundus image types and different difference feature extraction models, which can make the probability value with higher accuracy account for a higher proportion, so that the accuracy of diabetic nephropathy prediction is higher.

[0046] Step S60: Determine the target probability of the target diabetic patient having diabetic nephropathy according to the first probability value and the second probability value.

[0047] In one embodiment, the first probability value and the second probability value can be weighted and superimposed according to a first preset weight value corresponding to the first probability value and a second preset weight value corresponding to the second probability value, so as to obtain a target probability of diabetic nephropathy in a target diabetic patient, wherein the first preset weight value can be 0.35 and the second preset weight value can be 0.65, and the specific details are not limited here.

[0048] In one embodiment, the confidence levels corresponding to the first probability value and the second probability value can be calculated based on the target questionnaire and fundus images (including the first fundus image and the second fundus image), and then the target probability of the target diabetic patient developing diabetic nephropathy can be calculated based on the confidence level, weight, and the first probability value and the second probability value, so that the prediction of diabetic nephropathy is more accurate.

[0049] In one embodiment, when determining the target probability of a target diabetic patient suffering from diabetic nephropathy based on a first probability value and a second probability value, it includes: calculating a first confidence level of the first probability value based on the number of diabetic nephropathy-related factors in the questionnaire data; calculating a second confidence level of the first probability value based on the number of parameters of diabetic nephropathy-related factors of a specified type in the questionnaire data; determining a third confidence level of the first probability value based on the first confidence level and the second confidence level; determining a fourth confidence level of the second probability value based on the types of the first fundus image and the second fundus image; and weightedly superimposing a multiplication result of the third confidence level and the first probability value and a multiplication result of the second probability value and the fourth confidence level to obtain the target probability of the target diabetic patient suffering from diabetic nephropathy.

[0050] It can be understood that the greater the number of diabetic nephropathy associated factors in the questionnaire data, the higher the first confidence of the first probability value, and the greater the number of parameters of the specified type of diabetic nephropathy associated factors in the questionnaire data, the higher the second confidence of the first probability value. According to the weights corresponding to the number of diabetic nephropathy associated factors and the number of parameters, the first confidence and the second confidence are weightedly superimposed to obtain the third confidence of the first probability value, so that the first probability value can have different confidences according to the different degrees of importance corresponding to the number of diabetic nephropathy associated factors and the number of parameters in the questionnaire data, so that the prediction of diabetic nephropathy can be adjusted according to the comprehensiveness of the questionnaire filled out by the patient, so as to achieve a more accurate prediction effect.

[0051] In this embodiment, different types of fundus images have different degrees of influence on the confidence of the second probability value. The type of fundus influence and the confidence of the second probability can be set through a preset corresponding relationship, which is not specifically limited here.

[0052] Furthermore, the third confidence level is multiplied by the first probability value to obtain a first multiplication result, and then the fourth confidence level is multiplied by the second probability value to obtain a second multiplication result. Then, according to the preset weights corresponding to the first multiplication result and the second multiplication result, the first multiplication result and the second multiplication result are weightedly superimposed to obtain the target probability of diabetic nephropathy in the target diabetic patient.

[0053] The method for predicting diabetic nephropathy provided in the above-mentioned implementation method compares and analyzes the fundus images collected from diabetic patients at different examination cycles, collects and analyzes data from questionnaires of diabetic patients, and predicts the probability of diabetic patients developing diabetic nephropathy, which can provide a simple, economical and low-invasive early screening system. This method aims to provide an unprecedented, convenient, economical and low-invasive early screening approach for diabetic nephropathy through efficient data mining and risk assessment strategies. This solution carefully integrates the multi-dimensional data of diabetic patients, and uses advanced machine learning algorithms for deep learning and precise prediction, aiming to significantly improve the early identification rate of diabetic nephropathy, significantly reduce the diagnosis cost of patients, and provide a scientific basis for subsequent personalized medical decision-making, ultimately achieving optimal allocation of medical resources and precise disease management.

[0054] Next, the specific calculation method of the first probability value is described.

[0055] In one embodiment, when performing difference feature extraction on a first fundus image and a second fundus image to obtain target difference feature data, it includes: preprocessing the first fundus image and the second fundus image through a first convolutional neural network to obtain the first fundus image and the second fundus image that meet the feature extraction requirements; performing feature extraction on the first fundus image and the second fundus image that meet the feature extraction requirements according to the types of the first fundus image and the second fundus image to obtain first feature data of the first fundus image and second feature data of the second fundus image; performing differential calculation on the first feature data and the second feature data to obtain target difference feature data; and capturing significant features of the target difference feature data through an attention mechanism to obtain target difference feature data.

[0056] In the present embodiment, since the first fundus image and the second fundus image are fundus images collected in two different examination cycles, and the two fundus images may be collected in different hospitals using different types of medical equipment, there may be large differences in the properties of the two. By preprocessing the two through the first convolutional neural network of the present embodiment, the influencing properties of the two can be aligned and meet the requirements of subsequent artificial intelligence model processing. Compared with preprocessing the two through image processing technology, since the first convolutional neural network itself also belongs to the structure / module in the artificial intelligence model, the preprocessing of the first convolutional neural network can more accurately enable the two to meet the requirements of artificial intelligence model processing.

[0057] It should be noted that the first convolutional neural network is a pre-trained model structure that can process different fundus images into images that meet the same feature requirements in advance. In one embodiment, preprocessing the first fundus image and the second fundus image through the first convolutional neural network includes: size alignment, noise reduction, contrast enhancement, and distortion correction of the first fundus image and the second fundus image through the first convolutional neural network, so as to obtain the first fundus image and the second fundus image whose size, noise, contrast, and distortion meet the feature extraction requirements, so that subsequent artificial intelligence related processing can be more accurate and efficient.

[0058] In this embodiment, different feature extraction algorithms / models are used to extract features for different types of fundus images to obtain first feature data for the first fundus image and second feature data for the second fundus image, so that the fundus features are more consistent with the features that the corresponding type of fundus images are good at presenting, thereby improving the accuracy of diabetic nephropathy prediction.

[0059] In one embodiment, before performing differential calculation on the first feature data and the second feature data, the first feature data and the second feature data may be registered so that the feature points of the two correspond one to one, so that the subsequent differential calculation is more efficient and more accurate. In another embodiment, the first fundus image and the second fundus image may be registered when preprocessing the first fundus image and the second fundus image, or after obtaining the first fundus image and the second fundus image that meet the feature extraction requirements, so that the first feature data and the second feature data obtained subsequently are already registered feature data, and the specific registration timing is not limited here.

[0060] In this embodiment, the target differential feature data between the first feature data and the second feature data is obtained by differential calculation. The target differential feature data is used to indicate the difference between the first feature data and the second feature data. The differential calculation can effectively detect differences such as growth, atrophy or morphological changes in the fundus lesion area, so that subtle changes in the fundus can also be perceived. Compared with the traditional direct identification of the lesion area on a fundus image, it can better capture subtle changes in the patient's fundus, so that the budding of the lesion can also be discovered, and then effective preventive and therapeutic medical measures can be taken, or further examination measures can be taken, so that diabetic nephropathy can be more effectively controlled.

[0061] In one embodiment, when performing differential calculation on the first feature data and the second feature data, the first feature data and the second feature data can be first divided into feature points of the same size and the same number, and then the SURF (Speeded Up Robust Features) feature descriptors of each feature point are calculated. By calculating the distance between each feature descriptor, the corresponding feature points in the first feature data and the second feature data are matched. According to the matched feature points, the first feature data and the second feature data are registered so that the feature points are aligned. Finally, the first feature data and the second feature data after the registration are subtracted at the pixel level to obtain the target differential feature data. It should be noted that the above feature data is recorded through images, so the above feature data is actually also an image.

[0062] It is understandable that the target differential feature data objectively presents the difference between the first feature data and the second feature data, including differences that have reference value for the prediction of diabetic nephropathy, and also some differences that have no reference value for the prediction of diabetic nephropathy. Therefore, in this embodiment, by capturing significant features of the target differential feature data through the attention mechanism, the feature capture attention can be focused on the differences that have reference value for the prediction of diabetic nephropathy, and the target differential feature data is obtained, which plays a vital role in the prediction of diabetic nephropathy.

[0063] It should be noted that, depending on the type of different fundus images, the differences that need special attention, that is, the differences that have reference value are not the same. Therefore, in one embodiment, when the target differential feature data is captured by the attention mechanism, according to the types of the first fundus image and the second fundus image, the target differential feature data is captured by the corresponding attention mechanism to obtain the target differential feature data, so that the extraction of the differential features is more accurate and concise.

[0064] It can be understood that, according to the types of different fundus images, the feature extraction methods used and the dimensions of the features in advance are also different. In one embodiment, according to the types of the first fundus image and the second fundus image, feature extraction is performed on the first fundus image and the second fundus image that meet the feature extraction requirements to obtain first feature data of the first fundus image and second feature data of the second fundus image, including: if the types of the first fundus image and the second fundus image are fundus images, feature extraction of blood vessels, optic discs, and macula is performed on the first fundus image and the second fundus image respectively to obtain first feature data and second feature data of the first fundus image. the second characteristic data of the second fundus image; if the types of the first fundus image and the second fundus image are fluorescent angiography images, the flow properties of the blood vessels and / or choroidal blood vessels are extracted for the first fundus image and the second fundus image respectively to obtain the first characteristic data of the first fundus image and the second characteristic data of the second fundus image; if the types of the first fundus image and the second fundus image are optical coherence tomography images, the characteristics of the retinal layers and / or microvessels are extracted for the first fundus image and the second fundus image respectively to obtain the first characteristic data of the first fundus image and the second characteristic data of the second fundus image.

[0065] For fundus images, that is, fundus photos, the features of blood vessels, optic discs, and macula are all features with reference value for predicting diabetic nephropathy. Therefore, when training a model structure for extracting features of blood vessels, optic discs, and macula from fundus images, the first sample data with these feature tags is used for training. Specifically, the feature tags of the first sample data include: (1) Vascular features: vascular calcification (calcification of the vessel wall, manifested as bright white lines), vascular tortuosity (manifested as twisting and deformation of the vessel), and microaneurysm (manifested as small, round vascular dilation); (2) Characteristics of microvascular lesions: microaneurysms (manifested as small, round, red or orange lesions), hard exudates (manifested as yellow, waxy deposits, often appearing in the macular area), cotton-wool spots (manifested as white or grayish-white avascular areas, indicating local ischemia), neovascularization (manifested as new growth of abnormal blood vessels, often accompanied by bleeding), venous beading (manifested as focal venous dilatation, diffuse venous dilatation, and venous ring formation); (3) Thickness characteristics of the ganglion cell layer: The thickness of the ganglion cell layer reflects the number of retinal neurons and is correlated with diabetic nephropathy; (4) Characteristics of morphological changes in the macular area: The macula is the central area of ​​the retina, and its morphological changes can reflect the damage caused by diabetes to the retina.

[0066] It should be noted that the above-mentioned feature markers are all for fundus images, and are features that have reference value for predicting diabetic nephropathy. The feature extraction model is trained by using first sample data with at least one of the above-mentioned feature markers, thereby obtaining a model that can extract the above-mentioned features that have reference value for predicting diabetic nephropathy, and is used to extract features of blood vessels, optic disc, and macula for the first fundus image and the second fundus image, respectively.

[0067] In this embodiment, for the fluorescent angiography image, the characteristics of the flow properties of the blood vessels and / or choroidal blood vessels therein are all characteristics with reference value for predicting diabetic nephropathy. Therefore, when training a model structure for extracting the characteristics of the flow properties of the blood vessels and / or choroidal blood vessels in the fundus image, the training is performed using the second sample data with these feature tags. Specifically, the feature tags of the second sample data include: (1) Capillary leakage characteristics: Fluorescein dye leaks from blood vessels into surrounding tissues, appearing as fluorescent spots or plaques, indicating increased permeability of the vascular wall; (2) Vascular occlusion characteristics: complete or partial occlusion of a blood vessel, leading to local ischemia; (3) Fluorescein filling duration characteristics: The time it takes for the fluorescein dye to reach the choroid from the posterior ciliary artery reflects the speed of choroidal blood flow; (4) Fluorescein distribution characteristics: The uniformity of fluorescein distribution in the choroid reflects the permeability of the choroidal microvessels; (5) Fluorescein disappearance time: The time it takes for fluorescein to be cleared from the choroid reflects the rate of venous return.

[0068] It should be noted that the above-mentioned characteristic markers are all for fundus images of the fluorescent angiography image type, and are features that have reference value for predicting diabetic nephropathy. The feature extraction model is trained by using second sample data with at least one of the above-mentioned characteristic markers, thereby obtaining a model that can extract the above-mentioned features that have reference value for predicting diabetic nephropathy, and is used to extract features of the circulation properties of blood vessels and / or choroidal blood vessels for the first fundus image and the second fundus image, respectively.

[0069] In this embodiment, for the optical coherence tomography image, the features of the retinal layers and / or microvessels therein are all features with reference value for predicting diabetic nephropathy. Therefore, when training a model structure for extracting features of the retinal layers and / or microvessels from fundus images, training is performed using the third sample data with these feature markers. Specifically, the feature markers of the third sample data include: (1) Retinal thickness characteristics: including retinal nerve fiber layer thickness, inner nuclear layer thickness, outer nuclear layer thickness, etc.; (2) Characteristics of inner limiting membrane wrinkles: The inner limiting membrane is the innermost layer of the retina; (3) Retinal microvascular characteristics: OCT angiography (OCTA) can display information such as the density, distribution, and blood flow velocity of retinal microvessels; (4) Characteristics of the inner retinal cavity.

[0070] It should be noted that the above-mentioned characteristic markers are all for fundus images of the optical coherence tomography image type, and are features that have reference value for predicting diabetic nephropathy. The feature extraction model is trained by using third sample data with at least one of the above-mentioned characteristic markers, thereby obtaining a model that can extract the above-mentioned features that have reference value for predicting diabetic nephropathy, and is used to extract the features of the retinal layer structures and / or microvessels for the first fundus image and the second fundus image, respectively.

[0071] Specifically, when training the feature extraction model through the above-mentioned sample data (including the first sample data, the second sample data, and the third sample data), a model including structures such as the Visual Geometry Group (VGG) and the Residual Neural Network (ResNet) can be used as a basic model for training. The basic model is trained by inputting the above-mentioned sample data into the basic model. When the value of the loss function of the basic model decreases by less than a preset threshold, it can be determined that the model training is completed, and the basic model after training is determined as the feature extraction model.

[0072] In one embodiment, when the target differential feature data is captured by the attention mechanism to obtain the target differential feature data, it includes: determining the feature dimensions related to multiple concurrent diabetic nephropathy in the target differential feature data according to the types of the first fundus image and the second fundus image; for each feature dimension related to concurrent diabetic nephropathy, as an attention feature channel, performing the following processing: calculating the target differential feature data through three linear layers to obtain a query vector linear matrix, a key-value vector linear matrix, and a value vector linear matrix; performing dot product calculation on the query vector linear matrix and the key-value vector linear matrix to obtain a similarity matrix between the query vector linear matrix and the key-value vector linear matrix; normalizing the similarity matrix to obtain the attention weights corresponding to each attention feature channel; multiplying the attention weights corresponding to each attention feature channel by the value vector linear matrix to obtain the first difference feature data corresponding to each attention feature channel; based on the gating mechanism, fusing the first difference feature data corresponding to all attention feature channels to obtain the target difference feature data.

[0073] It should be noted that different types of fundus images have different feature dimensions associated with concurrent diabetic nephropathy. Specifically, the feature dimensions associated with concurrent diabetic nephropathy corresponding to different types of fundus images may be the same as the feature tags in the sample data of the corresponding types of fundus images. For example, the feature dimensions associated with concurrent diabetic nephropathy corresponding to fundus images of the fluorescent angiography image type may include: the feature tags in the above-mentioned second sample data. Similarly, the feature dimensions associated with concurrent diabetic nephropathy corresponding to fundus images of the optical coherence tomography image type may include: the feature tags in the above-mentioned third sample data. The specific details are not limited here.

[0074] In this embodiment, an attention feature channel is created for each feature dimension related to concurrent diabetic nephropathy, and the processing performed by each attention feature channel includes: calculating the target differential feature data through three linear layers to obtain a query vector linear matrix, a key-value vector linear matrix, and a value vector linear matrix; performing dot product calculation on the query vector linear matrix and the key-value vector linear matrix to obtain a similarity matrix between the query vector linear matrix and the key-value vector linear matrix; normalizing the similarity matrix to obtain the attention weights corresponding to each attention feature channel; multiplying the attention weights corresponding to each attention feature channel with the value vector linear matrix to obtain the first difference feature data corresponding to each attention feature channel.

[0075] Specifically, the three linear layers are query, key, and value linear layers. By calculating the target differential feature data through these three linear layers respectively, the query vector linear matrix, the key vector linear matrix, and the value vector linear matrix can be obtained, that is, the query vector linear matrix, the key-value vector linear matrix, and the value vector linear matrix.

[0076] Next, the similarity between the Query vector linear matrix and the Key vector linear matrix is ​​calculated through the dot product to obtain the similarity matrix. The obtained similarity matrix is ​​Softmax normalized to obtain the attention weight. The attention weight is multiplied by the Value vector linear matrix to obtain the weighted feature map, that is, the first difference feature data corresponding to each feature dimension.

[0077] It should be noted that after obtaining the first difference feature data corresponding to each feature dimension in each attention feature channel, it is necessary to fuse all the first difference feature data in all the attention feature channels. In this implementation, a gating mechanism is used to fuse the first difference feature data corresponding to all feature dimensions in all the attention feature channels, thereby obtaining the target difference feature data.

[0078] In one embodiment, when the first difference feature data corresponding to all feature dimensions in all attention feature channels are fused based on the gating mechanism, the first difference feature data corresponding to all feature dimensions in all attention feature channels can be fused through a preset gating formula, wherein the preset gating formula includes a target gating variable, and the calculation method of the target gating variable includes: concatenating the attention weights of each attention feature channel, inputting them into a fully connected layer to obtain an initial gating variable, and normalizing the initial gating variable through a Sigmoid function to obtain a target gating variable.

[0079] In one embodiment, when the first difference feature data corresponding to all feature dimensions in all attention feature channels are fused based on the gating mechanism, the first difference feature data corresponding to each attention feature channel is multiplied by the corresponding attention weight and the target gating variable, and the obtained second difference feature data are added to obtain the final fused feature, i.e., the target difference feature data.

[0080] It should be noted that this implementation is based on a gating mechanism and integrates different channel features in the differential of fundus feature maps of diabetic nephropathy. This approach has the advantages of strong flexibility and interpretability, which can effectively improve the performance of the model and make the results of diabetic nephropathy prediction more interpretable.

[0081] In one embodiment, when predicting the probability of concurrent diabetic nephropathy for target difference feature data to obtain a second probability value, it includes: performing feature dimension reduction on the target difference feature data through a fully connected layer including a nonlinear transformation function to obtain reduced dimension difference feature data; calculating the probability distribution of concurrent diabetic nephropathy and non-concurrent diabetic nephropathy on the reduced dimension difference feature data, and determining the obtained probability of concurrent diabetic nephropathy as the second probability value.

[0082] It can be understood that this embodiment maps the target difference feature data to a low-dimensional space, that is, the extracted high-dimensional difference feature vector is mapped to a low-dimensional feature space through a fully connected layer. The dimension of this low-dimensional space is usually smaller than the dimension of the original feature vector, which can reduce the complexity of the model and improve the computational efficiency. By introducing nonlinear transformation functions such as the Rectified Linear Unit (ReLU) function in the fully connected layer, the expressive power of the model can be enhanced, so that the reduced-dimensional difference feature data can better express the characteristics of concurrent diabetic nephropathy.

[0083] Next, this embodiment also calculates the probability distribution, that is, the output of the fully connected layer is passed through the softmax function to obtain a probability distribution. This probability distribution represents the probability that the sample belongs to each category, where the categories include concurrent diabetic nephropathy and non-concurrent diabetic nephropathy. Finally, the second probability value output is the probability of belonging to the concurrent diabetic nephropathy category.

[0084] Next, a specific method for calculating the first probability value is described.

[0085] In one embodiment, before the step of obtaining the questionnaire data of the target diabetic patient, the first fundus image and the second fundus image of the same type, it also includes: obtaining first sample data of multiple diabetic patients; wherein the first sample data includes patient parameters of multiple diabetic nephropathy-related factors; constructing causal relationships for the first sample data through a preset structural learning algorithm to obtain multiple first Bayesian networks; determining the second Bayesian network with the highest score from the multiple first Bayesian networks according to the Bayesian information criterion; performing conditional probability distribution calculation on the second Bayesian network by maximizing the likelihood function to obtain conditional probability distribution data of multiple diabetic nephropathy-related factors; based on the target Bayesian network, calculating the joint probability distribution of the target diabetic patient with diabetic nephropathy to obtain a first probability value, the step includes: obtaining the target conditional probability of each node in the target Bayesian network from the conditional probability distribution data; calculating the posterior probability of the target diabetic patient with diabetic nephropathy according to the target conditional probability of each node in the target Bayesian network to obtain a first probability value.

[0086] It can be understood that the first sample data includes patient parameters of multiple diabetic nephropathy associated factors of multiple diabetic patients who voluntarily provide data, which correspond to the answer values ​​of multiple diabetic nephropathy associated factors in the target questionnaire. The patient parameters are the parameter values ​​of the corresponding diabetic nephropathy associated factors of the diabetic patients. The number of diabetic nephropathy associated factors in the first sample data can be more than the number of diabetic nephropathy associated factors in the target questionnaire, so that the first sample data can cover a more comprehensive questionnaire and improve the prediction accuracy of diabetic nephropathy.

[0087] In this implementation, when determining the second Bayesian network, a structure learning algorithm such as a greedy search algorithm K2 or a Hill Climbing Algorithm can be used to construct a causal relationship for the first sample data, and the second Bayesian network can be determined by the BIC score. The following is an example of the K2 algorithm and the BIC score: It is understandable that the K2 algorithm, as a structural learning algorithm, plays an important role in constructing a Bayesian network model for predicting diabetic nephropathy. It can automatically learn the causal relationship between variables (i.e., diabetic nephropathy-associated factors) and confirmed diabetic nephropathy from the data, thereby helping us better understand the pathogenesis of diabetic nephropathy and establish a more accurate prediction model.

[0088] First, the variables are sorted according to domain knowledge (such as the pathogenesis of diabetic nephropathy) or the results of data analysis. Preferably, the diabetic nephropathy-related factors that have a greater impact on diabetic nephropathy are sorted first, such as blood sugar, blood pressure, etc. Then, each variable is taken as a node. Initially, there is no connection between nodes. Edges are added through the following iterative process: (1) For each node (e.g., urine protein), traverse all the previous nodes (e.g., blood sugar, blood pressure); (2) Calculate the BIC score increment brought by connecting the current node to each previous node; (3) Select the edge that brings the largest score increment and add it to the network. For example, if it is found that the connection between blood sugar and urine protein has the greatest improvement in the model score, add an edge between them; (4) Repeat steps 1-3 until the parent nodes of all nodes are determined, and the second Bayesian network can be obtained.

[0089] It should be noted that the K2 algorithm has the following advantages in predicting diabetic nephropathy: Automatic learning structure: There is no need to manually define the network structure in advance, and the relationship between variables can be automatically learned from the data; Strong interpretability: The learned network structure can intuitively show the causal relationship between variables, which helps us understand the pathogenesis of diabetic nephropathy; High prediction accuracy: The learned network structure can accurately predict the risk of kidney disease in diabetic patients.

[0090] It is understandable that maximizing the likelihood function is a commonly used parameter estimation method. Its core idea is: given a data set, find a set of parameter values ​​that maximizes the probability of the data set under the parameter. In other words, we believe that the parameter that is most likely to produce the observed data is the best parameter.

[0091] It should be noted that in the second Bayesian network, the maximum likelihood function is used to estimate the conditional probability table of each node. For a node X, its conditional probability table represents the probability of X taking various values ​​given the value of its parent node, that is, the conditional probability distribution data, where each node corresponds to a diabetic nephropathy-associated factor.

[0092] In this embodiment, after obtaining the conditional probability distribution data of multiple diabetic nephropathy-related factors, the joint probability distribution of the target diabetic patient complicated with diabetic nephropathy can be calculated according to the conditional probability distribution data to obtain a first probability value.

[0093] Specifically, the nodes in the target Bayesian network can be substituted into the preset Bayesian network model, and the target conditional probability of each node in the target Bayesian network can be obtained through the conditional probability distribution data in the Bayesian network model. The posterior probability of the endpoint node (i.e., confirmed diabetic nephropathy) is then calculated using the Bayesian formula to obtain the first probability value.

[0094] It should be noted that Bayesian networks have the following advantages in predicting diabetic nephropathy: (1) Causal relationship modeling: Bayesian networks can intuitively display the causal relationship between variables; (2) Uncertainty processing: Bayesian networks use probability to represent the relationship between variables, which can effectively handle the uncertainty in medical data and make the model more robust. (3) Multivariate analysis: Bayesian networks can consider the interactions between multiple variables at the same time, unlike traditional statistical methods which often only consider the relationship between two variables, and can comprehensively evaluate various influencing factors; (4) Incremental learning: When new data or knowledge emerges, the Bayesian network can be incrementally updated to improve the accuracy of the model and enable the model to adapt to new data; (5) Predictive ability: It can provide personalized predictions on the occurrence and development of diseases based on individual characteristics of patients, such as age, gender, and blood sugar control; (6) Strong interpretability: the structure and parameters of the Bayesian network have clear meanings, making the model highly interpretable and assisting clinical decision-making; (7) To process missing data, Bayesian networks can use existing information to perform probabilistic reasoning on missing data, thereby improving data utilization and improving data utilization.

[0095] Corresponding to the above method embodiment, see Figure 2A schematic diagram of a device for predicting diabetic nephropathy is shown, the device comprising: an acquisition module 21, used to acquire questionnaire data of a target diabetic patient, a first fundus image and a second fundus image of the same type; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data includes answer sheet values ​​corresponding to a plurality of diabetic nephropathy-related factors; a construction module 22, used to construct a target Bayesian network based on the questionnaire data; wherein the target Bayesian network is used to indicate the correlation between a plurality of diabetic nephropathy-related factors in the questionnaire data and the diagnosis of concurrent diabetic nephropathy; a calculation module 23, used to calculate the joint probability distribution of concurrent diabetic nephropathy in the target diabetic patient based on the target Bayesian network, and obtain a first probability value; an extraction module 24, used to perform difference feature extraction on the first fundus image and the second fundus image, and obtain target difference feature data; a prediction module 25, used to perform probability prediction of concurrent diabetic nephropathy on the target difference feature data, and obtain a second probability value; a determination module 26, used to determine the target probability of concurrent diabetic nephropathy in the target diabetic patient based on the first probability value and the second probability value.

[0096] The above-mentioned device for predicting diabetic nephropathy compares and analyzes fundus images collected from diabetic patients at different examination cycles, collects and analyzes data from questionnaires on diabetic patients, and predicts the probability of diabetic patients developing diabetic nephropathy, which can provide a simple, economical and low-invasive early screening system. This method aims to provide an unprecedented, convenient, economical and low-invasive early screening approach for diabetic nephropathy through efficient data mining and risk assessment strategies. This solution carefully integrates multi-dimensional data of diabetic patients, and uses advanced machine learning algorithms for deep learning and accurate prediction, aiming to significantly improve the early identification rate of diabetic nephropathy, significantly reduce the diagnosis cost of patients, and provide a scientific basis for subsequent personalized medical decision-making, ultimately achieving optimal allocation of medical resources and precise disease management.

[0097] Optionally, the extraction module 24 includes: a preprocessing unit, used to preprocess the first fundus image and the second fundus image through a first convolutional neural network to obtain the first fundus image and the second fundus image that meet the feature extraction requirements; an extraction unit, used to perform feature extraction on the first fundus image and the second fundus image that meet the feature extraction requirements according to the types of the first fundus image and the second fundus image, to obtain first feature data of the first fundus image and second feature data of the second fundus image; a calculation unit, used to perform differential calculation on the first feature data and the second feature data to obtain target differential feature data; a capture unit, used to capture significant features of the target differential feature data through an attention mechanism to obtain target difference feature data.

[0098] Optionally, the extraction unit is specifically used for: if the types of the first fundus image and the second fundus image are fundus images, then performing feature extraction of blood vessels, optic disc, and macula on the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image; if the types of the first fundus image and the second fundus image are fluorescent angiography images, then performing feature extraction of circulation properties of blood vessels and / or choroidal blood vessels on the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image; if the types of the first fundus image and the second fundus image are optical coherence tomography images, then performing feature extraction of retinal layer structures and / or microvessels on the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image.

[0099] Optionally, the capture unit is used to: determine feature dimensions related to multiple concurrent diabetic nephropathy in the target differential feature data according to the types of the first fundus image and the second fundus image; for each feature dimension related to concurrent diabetic nephropathy, take it as an attention feature channel and perform the following processing: calculate the target differential feature data through three linear layers to obtain a query vector linear matrix, a key-value vector linear matrix, and a value vector linear matrix; perform dot product calculation on the query vector linear matrix and the key-value vector linear matrix to obtain a similarity matrix between the query vector linear matrix and the key-value vector linear matrix; normalize the similarity matrix to obtain attention weights corresponding to each attention feature channel; multiply the attention weights corresponding to each attention feature channel by the value vector linear matrix to obtain first difference feature data corresponding to each attention feature channel; and fuse the first difference feature data corresponding to all attention feature channels based on the gating mechanism to obtain target difference feature data.

[0100] Optionally, the above-mentioned prediction module 25 is specifically used for: performing feature dimension reduction on the target difference feature data through a fully connected layer including a nonlinear transformation function to obtain reduced dimension difference feature data; performing probability distribution calculation of concurrent diabetic nephropathy and non-concurrent diabetic nephropathy on the reduced dimension difference feature data, and determining the obtained probability of concurrent diabetic nephropathy as the second probability value.

[0101] Optionally, the above-mentioned device also includes: an acquisition module, which is used to obtain first sample data of multiple diabetic patients; wherein the first sample data includes patient parameters of multiple diabetic nephropathy-related factors; a construction module, which is used to construct a causal relationship for the first sample data through a preset structural learning algorithm to obtain a first Bayesian network; a scoring module, which is used to determine the second Bayesian network with the highest score from the first Bayesian network according to the Bayesian information criterion; an algorithm module, which is used to perform conditional probability distribution calculation on the second Bayesian network by maximizing the likelihood function to obtain conditional probability distribution data of multiple diabetic nephropathy-related factors; the above-mentioned calculation module 23 is specifically used to: obtain the target conditional probability of each node in the target Bayesian network from the conditional probability distribution data; a posterior module, which is used to calculate the posterior probability of the target diabetic patient with diabetic nephropathy according to the target conditional probability of each node in the target Bayesian network to obtain a first probability value.

[0102] Optionally, the above-mentioned determination module 26 is specifically used to: calculate a first confidence level of a first probability value according to the number of diabetic nephropathy associated factors in the questionnaire data; calculate a second confidence level of the first probability value according to the number of parameters of diabetic nephropathy associated factors of a specified type in the questionnaire data; determine a third confidence level of the first probability value according to the first confidence level and the second confidence level; determine a fourth confidence level of the second probability value according to the types of the first fundus image and the second fundus image; and perform weighted superposition on the multiplication result of the third confidence level and the first probability value, and the multiplication result of the second probability value and the fourth confidence level to obtain a target probability of diabetic nephropathy in the target diabetic patient.

[0103] This embodiment also provides an electronic device, including a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the above method for predicting diabetic nephropathy. The electronic device can be a server or a terminal device.

[0104] See also Figure 3 As shown, the electronic device includes a processor 100 and a memory 101 , wherein the memory 101 stores machine executable instructions that can be executed by the processor 100 , and the processor 100 executes the machine executable instructions to implement the above-mentioned method for predicting diabetic nephropathy.

[0105] Further, Figure 3 The electronic device shown further includes a bus 102 and a communication interface 103 , and the processor 100 , the communication interface 103 and the memory 101 are connected via the bus 102 .

[0106] The memory 101 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 103 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used. The bus 102 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0107] The processor 100 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 100. The above processor 100 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present disclosure can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present disclosure can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 101, and the processor 100 reads the information in the memory 101 and completes the steps of the method of the above embodiment in combination with its hardware.

[0108] This embodiment also provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the above-mentioned method for predicting diabetic nephropathy.

[0109] The computer program product of the method, device, electronic device and storage medium for predicting diabetic nephropathy provided in the embodiments of the present disclosure includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be found in the method embodiments, which will not be repeated here.

[0110] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0111] In addition, in the description of the embodiments of the present disclosure, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present disclosure can be understood according to specific circumstances.

[0112] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0113] In the description of the present disclosure, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present disclosure and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present disclosure. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0114] Finally, it should be noted that the above embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above embodiments, those skilled in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for predicting diabetic nephropathy, characterized in that: The method comprises: Obtaining questionnaire data, a first fundus image and a second fundus image of the same type from a target diabetic patient; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data includes answer sheet values ​​corresponding to a plurality of diabetic nephropathy-related factors; According to the questionnaire data, a target Bayesian network is constructed; wherein the target Bayesian network is used to indicate the correlation between multiple diabetic nephropathy-related factors in the questionnaire data and the diagnosis of concurrent diabetic nephropathy; Based on the target Bayesian network, calculating the joint probability distribution of the target diabetic patient complicated with diabetic nephropathy to obtain a first probability value; Extracting difference features from the first fundus image and the second fundus image to obtain target difference feature data; Predicting the probability of concurrent diabetic nephropathy on the target difference feature data to obtain a second probability value; A target probability of the target diabetic patient having diabetic nephropathy is determined according to the first probability value and the second probability value.

2. The method according to claim 1, characterized in that The step of extracting difference features from the first fundus image and the second fundus image to obtain target difference feature data includes: Preprocessing the first fundus image and the second fundus image by a first convolutional neural network to obtain the first fundus image and the second fundus image that meet the feature extraction requirements; According to the types of the first fundus image and the second fundus image, feature extraction is performed on the first fundus image and the second fundus image that meet the feature extraction requirements to obtain first feature data of the first fundus image and second feature data of the second fundus image; Performing differential calculation on the first feature data and the second feature data to obtain target differential feature data; The target differential feature data is captured with significant features through an attention mechanism to obtain target differential feature data.

3. The method according to claim 2, characterized in that The step of performing feature extraction on the first fundus image and the second fundus image that meet the feature extraction requirements according to the types of the first fundus image and the second fundus image to obtain first feature data of the first fundus image and second feature data of the second fundus image comprises: If the types of the first fundus image and the second fundus image are fundus images, extracting features of blood vessels, optic discs, and macula from the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image; If the types of the first fundus image and the second fundus image are fluorescent angiography images, feature extraction of the flow properties of blood vessels and / or choroidal blood vessels is performed on the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image; If the types of the first fundus image and the second fundus image are optical coherence tomography images, the features of the retinal layer structures and / or microvessels are extracted from the first fundus image and the second fundus image respectively to obtain first feature data of the first fundus image and second feature data of the second fundus image.

4. The method according to claim 2, characterized in that: The step of capturing significant features of the target differential feature data through an attention mechanism to obtain the target differential feature data includes: Determining, according to the types of the first fundus image and the second fundus image, feature dimensions associated with multiple concurrent diabetic nephropathy in the target differential feature data; For each feature dimension related to concurrent diabetic nephropathy, as an attention feature channel, the following processing is performed: Calculating the target differential feature data through three linear layers to obtain a query vector linear matrix, a key-value vector linear matrix, and a value vector linear matrix; Performing dot product calculation on the query vector linear matrix and the key value vector linear matrix to obtain a similarity matrix between the query vector linear matrix and the key value vector linear matrix; Normalizing the similarity matrix to obtain the attention weight corresponding to each attention feature channel; Multiply the attention weights corresponding to the respective attention feature channels by the linear matrix of the value vector to obtain first difference feature data corresponding to each attention feature channel; Based on the gating mechanism, the first difference feature data corresponding to all the attention feature channels are fused to obtain the target difference feature data.

5. The method according to claim 1, characterized in that The step of performing probability prediction of concurrent diabetic nephropathy on the target difference feature data to obtain a second probability value comprises: Performing feature dimensionality reduction on the target difference feature data through a fully connected layer including a nonlinear transformation function to obtain reduced-dimensionality difference feature data; The probability distribution of concurrent diabetic nephropathy and non-concurrent diabetic nephropathy is calculated for the dimension-reduced difference feature data, and the obtained probability of concurrent diabetic nephropathy is determined as a second probability value.

6. The method according to claim 1, characterized in that Before the step of obtaining the questionnaire data of the target diabetic patient, the first fundus image and the second fundus image of the same type, the method further includes: Acquire first sample data of a plurality of diabetic patients; wherein the first sample data includes patient parameters of a plurality of diabetic nephropathy-related factors; Constructing a causal relationship for the first sample data using a preset structural learning algorithm to obtain a first Bayesian network; Determine a second Bayesian network with the highest score from the first Bayesian network according to the Bayesian Information Criterion; Performing conditional probability distribution calculation on the second Bayesian network by maximizing the likelihood function to obtain conditional probability distribution data of the multiple diabetic nephropathy-associated factors; The step of calculating the joint probability distribution of the target diabetic patient complicated with diabetic nephropathy based on the target Bayesian network to obtain a first probability value includes: Obtaining a target conditional probability of each node in the target Bayesian network from the conditional probability distribution data; According to the target conditional probability of each node in the target Bayesian network, the posterior probability of the target diabetic patient having diabetic nephropathy is calculated to obtain a first probability value.

7. The method according to claim 1, characterized in that The step of determining the target probability of the target diabetic patient having diabetic nephropathy according to the first probability value and the second probability value comprises: Calculating a first confidence level of the first probability value according to the number of diabetic nephropathy-associated factors in the questionnaire data; Calculating a second confidence level of the first probability value according to the number of parameters of diabetic nephropathy-associated factors of a specified type in the questionnaire data; Determining a third confidence level of the first probability value according to the first confidence level and the second confidence level; determining a fourth confidence level of the second probability value according to types of the first fundus image and the second fundus image; The result of multiplying the third confidence level by the first probability value and the result of multiplying the second probability value by the fourth confidence level are weighted and superimposed to obtain a target probability of the target diabetic patient suffering from diabetic nephropathy.

8. A device for predicting diabetic nephropathy, characterized in that: The device comprises: An acquisition module, used to acquire questionnaire data of a target diabetic patient, a first fundus image and a second fundus image of the same type; wherein the first fundus image is collected earlier than the second fundus image, and the questionnaire data includes answer sheet values ​​corresponding to a plurality of diabetic nephropathy-related factors; A construction module, used to construct a target Bayesian network according to the questionnaire data; wherein the target Bayesian network is used to indicate the correlation between multiple diabetic nephropathy-related factors and the diagnosis of concurrent diabetic nephropathy in the questionnaire data; A calculation module, configured to calculate the joint probability distribution of the target diabetic patient complicated with diabetic nephropathy based on the target Bayesian network, and obtain a first probability value; An extraction module, used for extracting difference features between the first fundus image and the second fundus image to obtain target difference feature data; A prediction module, used for predicting the probability of concurrent diabetic nephropathy for the target difference feature data to obtain a second probability value; A determination module is used to determine a target probability of the target diabetic patient having diabetic nephropathy according to the first probability value and the second probability value.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the method for predicting diabetic nephropathy according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when called and executed by a processor, prompt the processor to implement the method for predicting diabetic nephropathy according to any one of claims 1 to 7.

Images