Diabetes cardiovascular complication early warning method and system

Abstract

The invention relates to the technical field of medical health data processing, in particular to a diabetic cardiovascular complication early warning method and system, and the method comprises the steps: obtaining health data of a to-be-evaluated diabetic patient, encrypting the health data of the to-be-evaluated diabetic patient, and obtaining the encrypted health data; inputting the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications; and inputting the early warning features of the cardiovascular complications into a pre-set cardiovascular complication scoring model, obtaining a cardiovascular complication score of the to-be-evaluated diabetic patient, and performing early warning according to a pre-set early warning rule. The technical problems that in the prior art, data fusion precision is low, calculation efficiency is poor, and an effective mechanism for ensuring data consistency and safety is lacked are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical health data processing, and particularly relates to a diabetes cardiovascular complication early warning method and system. BACKGROUND

[0002] At present, the cardiovascular complication early warning of diabetes patients mainly relies on traditional statistical analysis methods and empirical models, which usually lack accurate modeling and analysis of individual health data. These methods are mostly based on a single data source or a small number of data features, lack cross-domain and multi-dimensional data fusion, and most of them fail to effectively combine encryption technology to protect patient privacy. With the increasing seriousness of data privacy and security issues, the existing technology has a big loophole in data sharing and privacy protection, making it difficult to effectively utilize multi-party data while ensuring data security.

[0003] At the same time, in the process of multi-party data sharing and collaborative computing, it is often necessary to decrypt data during fusion, which not only lacks effective privacy protection means, exposes users' sensitive health data, and leads to the risk of privacy leakage, but also may increase the computational complexity in the processing process. In addition, due to the heterogeneity of data sources, the existing technology often faces problems such as low fusion accuracy and poor computational efficiency during data fusion. Especially in cross-institutional and cross-regional data collaborative processing, there is a lack of effective mechanism to ensure the consistency and security of data.

[0004] Therefore, there is an urgent need for a diabetes cardiovascular complication early warning method and system. SUMMARY

[0005] (I) Technical problems to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides a diabetes cardiovascular complication early warning method and system, which solves the technical problems of low data fusion accuracy, poor computational efficiency and lack of effective mechanism to ensure the consistency and security of data in the prior art.

[0007] (II) Technical solutions

[0008] In order to achieve the above-mentioned purposes, the main technical solutions adopted by the present application include:

[0009] In a first aspect, the present application provides a diabetes cardiovascular complication early warning method, comprising:

[0010] S100, obtaining health data of a diabetes patient to be evaluated, encrypting the health data of the diabetes patient to be evaluated, and obtaining encrypted health data;

[0011] The health data of the to-be-evaluated diabetes patient includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urinary albumin data, creatinine data, and QT interval data of the to-be-evaluated diabetes patient obtained by a plurality of participants;

[0012] The plurality of participants include different hospitals and different departments of the hospitals;

[0013] S200, inputting the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications;

[0014] The early warning feature acquisition model includes: a blood glucose fluctuation rate acquisition sub-model, a blood pressure fluctuation rate acquisition sub-model, a heart rate variability acquisition sub-model, a urinary albumin / creatinine ratio acquisition sub-model, a QT interval feature acquisition sub-model, and a glycosylated hemoglobin acquisition sub-model;

[0015] S300, inputting the early warning features of cardiovascular complications into a pre-set cardiovascular complication scoring model to acquire a cardiovascular complication score of the to-be-evaluated diabetes patient, and performing early warning according to a pre-set early warning rule.

[0016] Optionally, the early warning features of cardiovascular complications include:

[0017] a blood glucose fluctuation rate, a blood pressure fluctuation rate, a heart rate variability, a urinary albumin / creatinine ratio, a QT interval feature, and a glycosylated hemoglobin level.

[0018] Optionally, the S100 specifically includes:

[0019] S110, acquiring health data of a to-be-evaluated diabetes patient, performing data preprocessing on the health data of the to-be-evaluated diabetes patient, and acquiring preprocessed health data;

[0020] S120, encrypting the preprocessed health data using a homomorphic encryption algorithm to acquire first encrypted data;

[0021] S130, according to a homomorphic encryption protocol, performing feature fusion on the first encrypted data to acquire a fused health data set as encrypted health data.

[0022] Optionally, the S130 specifically includes:

[0023] S131, classifying the first encrypted data into same-class data and cross-class data;

[0024] S132, according to a pre-set fusion rule, performing summation encryption operation on the same-class data to acquire a fusion result of each same-class data;

[0025] S133. Combine the fusion results of various similar data with cross-class data, and use homomorphic encryption protocol to fuse them to obtain the fused health data as the encrypted health data.

[0026] Optionally, S200 specifically includes:

[0027] S210. Input the encrypted blood glucose data into the pre-set blood glucose volatility sub-model and glycated hemoglobin acquisition sub-model to obtain blood glucose volatility and glycated hemoglobin levels respectively.

[0028] S220. Input the encrypted dynamic blood pressure data into the pre-set blood pressure volatility sub-model to obtain the blood pressure volatility.

[0029] S230. Input the encrypted heart rate data into the pre-set heart rate variability acquisition sub-model to obtain heart rate variability.

[0030] S240. Input the encrypted urine albumin and creatinine data into the pre-set urine albumin / creatinine ratio acquisition sub-model to obtain the urine albumin / creatinine ratio.

[0031] S250. Input the encrypted QT interval data into the pre-set QT interval feature acquisition sub-model to obtain the QT interval features.

[0032] Optionally, the process may further include the following steps before step S100:

[0033] S000. The cardiovascular complication scoring model is trained using the training dataset to obtain the trained cardiovascular complication scoring model.

[0034] The training dataset consists of glycemic variability, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, glycated hemoglobin level, and corresponding cardiovascular complication scores of other diabetic patients.

[0035] Optionally, S000 specifically includes:

[0036] Each data point in the training dataset is input into the cardiovascular complication scoring model for training until the pre-set regression loss function and classification loss function corresponding to the cardiovascular complication scoring model converge, thus obtaining the trained cardiovascular complication scoring model.

[0037] Optionally, S300 includes:

[0038] Input the early warning features of cardiovascular complications into the following formula to obtain the cardiovascular complication score of the diabetic patient to be evaluated:

[0039]

[0040] Where S is the cardiovascular complication score, B is the glycated blood glucose variability, H is the glycated hemoglobin level, P is the blood pressure variability, R is the heart rate variability, U is the urine albumin / creatinine ratio, Q is the QT interval characteristic, A is age, and Q... mean The mean of the QT interval characteristics, Q std denoted as , where is the standard deviation of the QT interval characteristics. w1, w2, w3, w4, w5, w6, and w7 represent the weights of blood glucose variability, glycated hemoglobin level, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, and age, respectively. c1, c2, and c3 are the coefficients of the nonlinear transformation. α is the weight of the linear combination component, β is the weight of the network structure component, k1 is the weight of node correlation, and k2 is the weight of node path information. To calculate the sum of the correlation degrees of all early warning feature nodes, It is the sum of path information for all warning feature nodes.

[0041] Optionally, in S400, the pre-set warning rule is:

[0042] When the cardiovascular complication score is less than or equal to 0.3, it is considered low risk and a Level 1 warning is issued.

[0043] When the cardiovascular complication score is greater than 0.3 and less than 0.7, it is considered a moderate risk and a level-two warning is issued.

[0044] When the cardiovascular complication score is greater than or equal to 0.7, it is considered high risk and a level 3 warning is issued.

[0045] Secondly, embodiments of the present invention provide a diabetic cardiovascular complication early warning system, comprising:

[0046] The first acquisition module is used to acquire the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and acquire the encrypted health data.

[0047] The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants.

[0048] The multiple participating parties include different hospitals and different departments within those hospitals;

[0049] The early warning feature acquisition module is used to input the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications.

[0050] The early warning feature acquisition model includes: a blood glucose fluctuation acquisition sub-model, a blood pressure fluctuation acquisition sub-model, a heart rate variability acquisition sub-model, a urine albumin / creatinine ratio acquisition sub-model, a QT interval feature acquisition sub-model, and a glycated hemoglobin acquisition sub-model.

[0051] The data analysis module is used to input the early warning features of the cardiovascular complications into a pre-set cardiovascular complication scoring model, obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

[0052] (III) Beneficial Effects

[0053] The beneficial effects of this invention are: the method for early warning of diabetic cardiovascular complications of this invention improves data security by encrypting data using a fully homomorphic encryption algorithm; and by using a cardiovascular complication scoring model, it solves the shortcomings of existing methods for early warning of diabetic cardiovascular complications in terms of data fusion accuracy and prediction accuracy, thereby improving the accuracy of early warning. Attached Figure Description

[0054] Figure 1 This is a schematic flowchart of a method for early warning of cardiovascular complications in diabetes according to an embodiment of the present invention;

[0055] Figure 2 This is a structural block diagram of a diabetic cardiovascular complication early warning system according to an embodiment of the present invention. Detailed Implementation

[0056] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] This invention proposes a method and system for early warning of diabetic cardiovascular complications. To address the significant vulnerabilities in existing diabetic cardiovascular complication prediction methods regarding data sharing and privacy protection, as well as the low accuracy of data fusion in cross-institutional and cross-regional collaborative data processing and the lack of effective mechanisms to ensure data consistency and security, this invention employs a fully homomorphic encryption algorithm to encrypt the data. This allows all participants to perform collaborative calculations without disclosing sensitive data, ensuring a high level of data privacy protection. Simultaneously, feature fusion based on encrypted data avoids the data silo problem in traditional methods, improving the accuracy and consistency of data fusion. Furthermore, a cardiovascular complication scoring model is used to conduct multi-dimensional analysis of diabetic cardiovascular complications, effectively improving the accuracy and adaptability of predictions.

[0058] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0059] Example 1

[0060] See Figure 1 An embodiment of the present invention provides a method for early warning of diabetic cardiovascular complications, comprising:

[0061] Step S100: Obtain the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and obtain the encrypted health data.

[0062] The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants;

[0063] The multiple participating parties include different hospitals and different departments within those hospitals;

[0064] Step S200: Input the encrypted health data into the pre-set early warning feature acquisition model to obtain early warning features of cardiovascular complications;

[0065] The early warning feature acquisition model includes: blood glucose fluctuation acquisition sub-model, blood pressure fluctuation acquisition sub-model, heart rate variability acquisition sub-model, urine albumin / creatinine ratio acquisition sub-model, QT interval feature acquisition sub-model and glycated hemoglobin acquisition sub-model;

[0066] Step S300: Input the early warning features of the cardiovascular complications into the pre-set cardiovascular complication scoring model to obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

[0067] The basic information of the diabetic patients to be evaluated includes their personal information, such as name, gender, and age.

[0068] This invention provides a method for early warning of diabetic cardiovascular complications. It employs a fully homomorphic encryption algorithm to encrypt data, thereby improving data security. By using a cardiovascular complication scoring model, it addresses the shortcomings of existing methods for early warning of diabetic cardiovascular complications in terms of data fusion accuracy and prediction accuracy, thus improving the accuracy of early warning.

[0069] Example 2

[0070] This invention provides a method for early warning of cardiovascular complications in diabetes, comprising:

[0071] Step S100: Obtain the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and obtain the encrypted health data.

[0072] The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants;

[0073] The multiple participating parties include different hospitals and different departments within those hospitals;

[0074] Blood glucose data is obtained through continuous glucose monitoring (CGM), which involves wearing a small sensor under the skin that automatically records the blood glucose level at set time intervals.

[0075] Ambulatory blood pressure monitoring involves wearing a portable blood pressure monitor that automatically measures blood pressure at preset time intervals (usually every 15-30 minutes), including data during sleep at night.

[0076] Furthermore, a professional electrocardiogram (ECG) examination can be performed to obtain detailed information on cardiac activity, including heart rate data and QT interval data.

[0077] Urinary albumin and creatinine data are obtained by collecting random or 24-hour urine samples and sending them to a medical laboratory for analysis. Laboratory technicians use specific chemical methods to determine the concentration of albumin and creatinine levels in the urine.

[0078] In practice, multiple participants can be multiple departments or units within a single medical institution, or a collaborative network between multiple medical institutions. The arrangement of these participants can be flexibly adjusted according to the actual situation.

[0079] For example, when the participants are multiple departments within a hospital, such as endocrinology, cardiology, and radiology, each department is responsible for collecting different types of health data. For instance, endocrinology might collect blood glucose data, cardiology might collect blood pressure data, and radiology might provide relevant imaging examination data.

[0080] Furthermore, multiple participants can also form a collaborative network among multiple hospitals. In this scenario, the participants may come from different hospitals, collaboratively providing early warnings of cardiovascular complications in diabetic patients. Each hospital is responsible for collecting and encrypting data from its own region or patient population, then transmitting the encrypted data to a unified computing node for further processing via a secure communication protocol. In this way, different hospitals can share patient health data without disclosing any patient's private information.

[0081] Specifically, health data for diabetic patients to be evaluated can be obtained in the following ways:

[0082] The physiological indicators of diabetic patients are monitored in real time using devices such as blood glucose meters, ambulatory blood pressure monitoring systems, and heart rate variability analyzers. These devices can automatically collect and upload data, ensuring the real-time nature and accuracy of the data collection process.

[0083] All participating parties can access the patient's health data through the Electronic Health Record (EHR) system during the treatment process. This includes data such as blood glucose levels, heart rate, blood pressure, urine albumin, and creatinine ratios. These records are typically provided by doctors or caregivers during the patient's treatment and serve as a basis for health management and disease early warning. It should be noted that all collected health data is encrypted using a fully homomorphic encryption algorithm by all participating parties. The application of fully homomorphic encryption ensures that necessary computational operations can still be performed on the encrypted data without decryption, thus effectively protecting patient privacy.

[0084] Specifically, when using fully homomorphic encryption, each participant generates a public key and a private key pair. The public key is used to encrypt data, and the private key is used to decrypt data. After data collection is complete, the participants use their public keys to encrypt the collected health data, obtaining encrypted data. The encrypted data is then transmitted to other participants or the central computing node via a secure communication protocol.

[0085] Furthermore, to ensure data privacy and security, all encrypted data can only be decrypted or used by the participant holding the corresponding private key. During this process, data from different participants is not exposed to other parties, guaranteeing the security of data transmission and the protection of patient privacy.

[0086] In the specific implementation process, step S100 specifically includes:

[0087] Step S110: Obtain the health data of the diabetic patients to be evaluated, perform data preprocessing on the health data of the diabetic patients to be evaluated, and obtain the preprocessed health data.

[0088] Specifically, interpolation and mean methods are used to fill in missing values; statistical methods (such as Z-scores and box plots) are applied to identify and process outliers, duplicate data are removed, and data from different scales are converted to the same scale.

[0089] Step S120: Encrypt the preprocessed health data using a fully homomorphic encryption algorithm to obtain the first encrypted data;

[0090] A key pair, including a public key and a private key, is generated using a fully homomorphic encryption algorithm; the preprocessed health data is then encrypted using the public key to obtain the first encrypted data.

[0091] Each participant transmits the encrypted data to the computing node via a secure communication protocol.

[0092] Fully homomorphic encryption algorithms allow computation on encrypted data without decryption. Anyone possessing the encrypted data and the corresponding public key can perform computations on the data without accessing the original data, while only the person with the corresponding private key can decrypt the result. Computation nodes refer to cloud service providers, data centers, or other types of servers. A computation node receives initial encrypted data from different participants and can perform computations on it without decryption. After computation, the result, still encrypted, can be sent back to the data owner, who can then decrypt the result using their own private key.

[0093] By using a fully homomorphic encryption algorithm, multiple parties (such as different hospitals or departments) encrypt their respective patient health data and then perform joint calculations. The data is processed in an encrypted state, thereby effectively avoiding the risk of data leakage.

[0094] Step S130: According to the homomorphic encryption protocol, the first encrypted data is feature fused to obtain the fused health dataset as the encrypted health data.

[0095] Step S130 specifically includes:

[0096] Step S131: Classify the first encrypted data into data of the same type and data across types;

[0097] Step S132: According to the pre-set fusion rules, perform summation and encryption operations on similar data to obtain the fusion results of each similar data.

[0098] Step S133: Combine the fusion results of various similar data with cross-class data, and use homomorphic encryption protocol to fuse them to obtain the fused health data as the encrypted health data.

[0099] Specifically, in this embodiment of the invention, the first encrypted data is divided into homogeneous data and cross-category data based on type. For homogeneous data, a homomorphic encryption protocol is used for encrypted fusion, merging the results of data of the same type through encrypted operations. During this process, all homogeneous data remains encrypted, and the fusion result obtained after the encrypted operations is still encrypted. The advantage of this design is that it avoids data leakage problems when fusing different types of data, ensuring privacy protection during the data fusion process. For cross-category data, i.e., the fusion of data of different types, a homomorphic encryption protocol is used for secondary fusion, further ensuring the confidentiality and integrity of the data. Through this encrypted fusion method, all data from different participants are ultimately aggregated into an encrypted dataset, providing a unified encrypted data source for subsequent feature extraction and health risk assessment.

[0100] In the above steps, by performing fusion and feature extraction operations on encrypted data, it is ensured that all participants can share data processing results without leaking their own data. Because homomorphic encryption algorithms have high computational efficiency, they not only guarantee data privacy but also enable subsequent processes such as model training and risk assessment to proceed smoothly, an effect that other traditional encryption methods cannot achieve.

[0101] Step S200: Input the encrypted health data into the pre-set early warning feature acquisition model to obtain early warning features of cardiovascular complications;

[0102] The early warning feature acquisition model includes: a sub-model for acquiring blood glucose volatility, a sub-model for acquiring blood pressure volatility, a sub-model for acquiring heart rate variability, a sub-model for acquiring urine albumin / creatinine ratio, a sub-model for acquiring QT interval features, and a sub-model for acquiring glycated hemoglobin.

[0103] Among them, the early warning characteristics of cardiovascular complications include:

[0104] Blood glucose variability, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, and glycated hemoglobin level.

[0105] The specific implementation process of step S200 is as follows:

[0106] Step S210: Input the encrypted blood glucose data into the pre-set blood glucose volatility sub-model and glycated hemoglobin acquisition sub-model to obtain blood glucose volatility and glycated hemoglobin levels respectively.

[0107] The encrypted blood glucose data will be input into the following formulas to obtain the blood glucose variability and glycated hemoglobin level:

[0108]

[0109] Where B represents blood glucose variability, H represents glycated hemoglobin level, and G represents blood glucose variability. i For the i-th blood glucose data point, Let be the mean blood glucose level, and n be the total number of data points.

[0110] Step S220: Input the encrypted dynamic blood pressure data into the pre-set blood pressure volatility sub-model to obtain the blood pressure volatility;

[0111] The blood pressure variability sub-model is as follows:

[0112]

[0113] Where P is the blood pressure variability, BP i For the i-th blood pressure data point, denoted as mean blood pressure, and m represents the total number of blood pressure data points.

[0114] Step S230: Input the encrypted heart rate data into the pre-set heart rate variability acquisition sub-model to obtain heart rate variability;

[0115] Heart rate variability submodel:

[0116]

[0117] Where R represents heart rate variability, RR i For the i-th heart rate interval data point, is the mean heart rate interval, and k is the total number of blood pressure data points.

[0118] Step S240: Input the encrypted urine albumin and creatinine data into the pre-set urine albumin / creatinine ratio acquisition sub-model to obtain the urine albumin / creatinine ratio;

[0119] Sub-model for obtaining urine albumin / creatinine ratio:

[0120]

[0121] Where U is the urine albumin / creatinine ratio, and A i For the i-th urinary albumin data point, C i Let p be the i-th creatinine data point, and p be the number of samples.

[0122] Step S250: Input the encrypted QT interval data into the pre-set QT interval feature acquisition sub-model to obtain the QT interval features.

[0123] QT interval feature acquisition sub-model:

[0124]

[0125] Where U represents the QT interval characteristic, QTi Let q be the i-th QT interval data point, and q be the total number of QT interval data points.

[0126] In the above formula, E(x) represents encrypting the value of x. All the above formulas operate in the encryption domain, and all addition, subtraction, multiplication and division refer to their corresponding homomorphic operations.

[0127] It should be noted that this invention, by extracting encrypted cardiovascular complication early warning features and employing a specific encrypted data processing method, ensures the privacy protection of the health data of all participating parties and achieves secure data fusion.

[0128] The above processing method not only protects patient privacy but also enables the confidential exchange and processing of health data among different medical institutions and participating parties. Using encrypted data to calculate differences, squared differences, and standard deviations effectively avoids directly exposing patients' personal information and allows complex mathematical calculations to be completed in an encrypted state during the extraction of each feature. The final result is an encrypted feature analysis, ensuring the security and operability of the entire process. This encryption method enables more accurate and comprehensive early warning of cardiovascular complications without compromising data privacy.

[0129] Furthermore, the cardiovascular complication early warning features selected in this invention are optimized based on the typical physiological and pathological characteristics of diabetic patients. The preferred features have significant cardiovascular risk prediction capabilities and can comprehensively reflect changes in the metabolism, cardiovascular health, and physiological signs of diabetic patients. These features are deep features extracted from existing health data through encrypted calculations, and have stronger predictive capabilities than the original data.

[0130] S300. Input the early warning features of the cardiovascular complications into the pre-set cardiovascular complication scoring model to obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

[0131] Specifically, this includes inputting early warning features of cardiovascular complications into the following formula to obtain a cardiovascular complication score for the diabetic patient to be evaluated:

[0132]

[0133] Where S is the cardiovascular complication score, B is the glycated blood glucose variability, H is the glycated hemoglobin level, P is the blood pressure variability, R is the heart rate variability, U is the urine albumin / creatinine ratio, Q is the QT interval characteristic, A is age, and Q... mean The mean of the QT interval characteristics, Q stddenoted as , where is the standard deviation of the QT interval characteristics. w1, w2, w3, w4, w5, w6, and w7 represent the weights of blood glucose variability, glycated hemoglobin level, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, and age, respectively. c1, c2, and c3 are the coefficients of the nonlinear transformation. α is the weight of the linear combination component, β is the weight of the network structure component, k1 is the weight of node correlation, and k2 is the weight of node path information. To calculate the sum of the correlation degrees of all early warning feature nodes, It is the sum of path information for all warning feature nodes.

[0134] In this formula, the use of a logarithmic transformation function to consider blood glucose volatility can reduce the impact of extreme blood glucose values ​​and make the data distribution more uniform. When considering glycated hemoglobin, the use of the sigmoid function can map the data to a finite interval and smoothly handle outliers. Furthermore, the use of a power function to consider blood pressure volatility can highlight changes within certain specific ranges. The use of a square root transformation to consider heart rate variability can ensure the positive direction of the data and reduce the impact of extreme values, etc.

[0135] Furthermore, the above formula also considers the correlation and path relationships between nodes in the network structure. By introducing correlation and path information, it can more intuitively show which features are strongly correlated and which features indirectly affect the final score through other features. This helps doctors understand the model's decision-making process and increases their trust in the model.

[0136] In this embodiment, the pre-set warning rule is as follows:

[0137] When the cardiovascular complication score is less than or equal to 0.3, it is considered low risk and a Level 1 warning is issued.

[0138] When the cardiovascular complication score is greater than 0.3 and less than 0.7, it is considered a moderate risk and a level-two warning is issued.

[0139] When the cardiovascular complication score is greater than or equal to 0.7, it is considered high risk and a level 3 warning is issued.

[0140] Furthermore, the procedure prior to step S100 also includes:

[0141] Step S000: Train the cardiovascular complication scoring model using the training dataset to obtain the trained cardiovascular complication scoring model;

[0142] The training dataset consists of other diabetic patients' blood glucose variability, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, glycated hemoglobin level, age, and corresponding cardiovascular complication scores.

[0143] Step S000 specifically includes:

[0144] Each data point in the training dataset is input into the cardiovascular complication scoring model for training until the pre-set regression loss function and classification loss function corresponding to the cardiovascular complication scoring model converge, thus obtaining the trained cardiovascular complication scoring model.

[0145] Furthermore, the method of this embodiment of the invention can also use a support vector machine model to further evaluate and determine whether the diabetic patient under evaluation has a risk of cardiovascular complications after obtaining the cardiovascular complication score.

[0146] Support Vector Machines (SVMs) classify data by finding the optimal hyperplane, minimizing the following objective function, expressed as:

[0147]

[0148] Where w is the weight vector of the support vector machine, b is the bias term, and ξ is the weight vector of the support vector machine. i is a slack variable representing the distance between the data point and the hyperplane, and C is a regularization parameter.

[0149] The score f(X) output by the support vector machine model SVM This determines the patient's risk category.

[0150] Nonlinear classification is performed using radial basis function (RBF) kernels, expressed as:

[0151]

[0152] Where x and x′ are input vectors, and 6 is the width parameter of the Gaussian kernel.

[0153] The support vector machine model outputs a score f(X). SVM Based on a set threshold, the patient's risk level is determined, and expressed as follows:

[0154]

[0155] Among them, Risk SVM =1 indicates that the patient is at high risk. SVM =0 indicates that the patient has no risk. SVM =0.5 indicates that the patient is at medium risk.

[0156] This invention further employs a Support Vector Machine (SVM) model. The SVM model, through its nonlinear mapping capability, can map data that was originally difficult to separate into a high-dimensional space, thereby enabling more accurate classification and prediction. In this structure, the cardiovascular complication scores of the diabetic patients to be evaluated serve as input features to the SVM model, allowing the model to further optimize the calculation of the probability of complication occurrence based on these preliminary risk assessments. This cascaded approach not only improves the model's expressive power but also enhances the accuracy and robustness of predictions. Especially for the complex, multifactorial disease of cardiovascular complications in diabetic patients, it ensures computational efficiency while fully mining the potential information in the data, providing more accurate early warnings.

[0157] This embodiment of a method for early warning of cardiovascular complications in diabetes employs a fully homomorphic encryption algorithm to encrypt data, enabling collaborative computation among participants without disclosing sensitive data and ensuring a high level of data privacy protection. Feature fusion based on encrypted data avoids the data silo problem in traditional methods, improving the accuracy and consistency of data fusion. A cardiovascular complication scoring model combined with a support vector machine model is used to perform multi-dimensional analysis of cardiovascular complication risk, effectively improving the accuracy and adaptability of predictions.

[0158] Example 3

[0159] This embodiment provides a method for early warning of cardiovascular complications in diabetes, as described in Embodiments 1 and 2. The method will not be described in detail here, but the beneficial effects of the method will be demonstrated through simulation experiments.

[0160] The experiment was conducted jointly by the endocrinology, cardiology, and metabolic departments of three hospitals in a certain city, collecting data from 300 diabetic patients aged 40-70. The health data used included patients' metabolic function (glucose levels, insulin resistance index), cardiovascular health (ambulatory blood pressure, heart rate), endocrine status (hormone levels), and physiological signs (urine albumin, creatinine ratio, and QT interval). Each hospital collected different types of health data, which were then encrypted using fully homomorphic encryption technology before being sent to the central computing node for analysis. All data collection and processing underwent standardization to ensure data consistency and integrity.

[0161] This invention involves multiple hospitals collaborating and employing a homomorphic encryption protocol to encrypt health data, ensuring data privacy protection. Each hospital uploads different types of health data to the computing node after fully homomorphically encrypting them. Next, all encrypted data undergoes data fusion processing, including the fusion of similar data and cross-class data. The fused encrypted health data is then analyzed using a cardiovascular complication scoring model. In this way, health data from multiple hospitals are combined to comprehensively extract key features, providing a more accurate basis for the final risk assessment. The assessment results are then transmitted to all participating parties using homomorphic encryption. Experimental results are shown in Tables 1 and 2.

[0162] Table 1 Comparison of Risk Assessment Results

[0163]

[0164] Table 2 Comparison of Risk Assessment Accuracy

[0165] Method Accuracy (%) Precision (%) Recall (%) F1 Score (%) Conventional method 82 80 79 84 Method of the present invention 94 95 92 93

[0166] This embodiment of the experiment demonstrates that the method of the present invention effectively solves the problems of isolated data from different hospitals, insufficient privacy protection, and inaccurate risk assessment in traditional methods through multi-party encrypted data fusion technology. In traditional methods, independent analysis of data by each hospital leads to information gaps and biases, particularly in core health indicators such as glycemic variability and ambulatory blood pressure variability, where prediction results show significant differences. The present invention uses a homomorphic encryption protocol for data fusion, ensuring data sharing among different hospitals while effectively protecting patient privacy, avoiding the risk of data leakage, and significantly improving the accuracy and consistency of the fused health data. Experimental results show that after data fusion using the method of the present invention, the prediction results for key indicators such as glycemic variability are significantly lower than those of traditional methods, demonstrating a more accurate risk assessment capability.

[0167] Furthermore, this invention employs a cardiovascular complication scoring model combined with a support vector machine model to accurately predict the health risks of diabetic patients. Through feature extraction and fusion, this invention significantly improves the predictive accuracy of cardiovascular complication early warning and exhibits better adaptability to data differences between different hospitals.

[0168] This invention significantly improves the early warning effect of diabetic cardiovascular complications by integrating and analyzing encrypted data while protecting privacy. It avoids the problems caused by data silos and privacy leaks in traditional methods and optimizes the accuracy of overall data processing and prediction.

[0169] Example 4

[0170] This embodiment of a diabetic cardiovascular complication early warning system includes:

[0171] The first acquisition module is used to acquire the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and acquire the encrypted health data.

[0172] The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants.

[0173] The early warning feature acquisition module is used to input the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications.

[0174] The early warning feature acquisition model includes: a blood glucose fluctuation acquisition sub-model, a blood pressure fluctuation acquisition sub-model, a heart rate variability acquisition sub-model, a urine albumin / creatinine ratio acquisition sub-model, a QT interval feature acquisition sub-model, and a glycated hemoglobin acquisition sub-model.

[0175] The data analysis module is used to input the early warning features of the cardiovascular complications into a pre-set cardiovascular complication scoring model, obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

[0176] The early warning system for diabetic cardiovascular complications in this embodiment significantly improves the early warning effect of diabetic cardiovascular complications while protecting privacy. It avoids the problems caused by data silos and privacy leaks in the prior art and optimizes the overall processing and prediction speed.

[0177] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0178] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0179] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that they are in indirect contact through an intermediate medium. Furthermore, "above," "over," or "on top" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," or "beneath" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0180] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0181] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for early warning of cardiovascular complications in diabetes, characterized in that, include: S100: Obtain the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and obtain the encrypted health data. The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants. The multiple participating parties include different hospitals and different departments within those hospitals; S200. Input the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications; The early warning feature acquisition model includes: a blood glucose fluctuation acquisition sub-model, a blood pressure fluctuation acquisition sub-model, a heart rate variability acquisition sub-model, a urine albumin / creatinine ratio acquisition sub-model, a QT interval feature acquisition sub-model, and a glycated hemoglobin acquisition sub-model. S300. Input the early warning features of the cardiovascular complications into the pre-set cardiovascular complication scoring model to obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

2. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, The early warning features of cardiovascular complications include: Blood glucose variability, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, and glycated hemoglobin level.

3. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, Specifically, S100 includes: S110. Obtain the health data of the diabetic patient to be evaluated, perform data preprocessing on the health data of the diabetic patient to be evaluated, and obtain the preprocessed health data. S120. Encrypt the preprocessed health data using a fully homomorphic encryption algorithm to obtain the first encrypted data; S130. According to the homomorphic encryption protocol, the first encrypted data is feature fused to obtain the fused health dataset as the encrypted health data.

4. The method for early warning of diabetic cardiovascular complications according to claim 3, characterized in that, S130 specifically includes: S131. Classify the first encrypted data into data of the same type and data across types; S132. According to the pre-set fusion rules, perform summation and encryption operations on similar data to obtain the fusion results of each similar data. S133. Combine the fusion results of various similar data with cross-class data, and use homomorphic encryption protocol to fuse them to obtain the fused health data as the encrypted health data.

5. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, Specifically, S200 includes: S210. Input the encrypted blood glucose data into the pre-set blood glucose volatility sub-model and glycated hemoglobin acquisition sub-model to obtain blood glucose volatility and glycated hemoglobin levels respectively. S220. Input the encrypted dynamic blood pressure data into the pre-set blood pressure volatility sub-model to obtain the blood pressure volatility. S230. Input the encrypted heart rate data into the pre-set heart rate variability acquisition sub-model to obtain heart rate variability. S240. Input the encrypted urine albumin and creatinine data into the pre-set urine albumin / creatinine ratio acquisition sub-model to obtain the urine albumin / creatinine ratio. S250. Input the encrypted QT interval data into the pre-set QT interval feature acquisition sub-model to obtain the QT interval features.

6. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, The process preceding S100 also includes: S000. The cardiovascular complication scoring model is trained using the training dataset to obtain the trained cardiovascular complication scoring model. The training dataset consists of glycemic variability, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, glycated hemoglobin level, and corresponding cardiovascular complication scores of other diabetic patients.

7. The method for early warning of diabetic cardiovascular complications according to claim 6, characterized in that, The S000 specifically includes: Each data point in the training dataset is input into the cardiovascular complication scoring model for training until the pre-set regression loss function and classification loss function corresponding to the cardiovascular complication scoring model converge, thus obtaining the trained cardiovascular complication scoring model.

8. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, The S300 includes: Input the early warning features of cardiovascular complications into the following formula to obtain the cardiovascular complication score of the diabetic patient to be evaluated: Where S is the cardiovascular complication score, B is the glycated blood glucose variability, H is the glycated hemoglobin level, P is the blood pressure variability, R is the heart rate variability, U is the urine albumin / creatinine ratio, Q is the QT interval characteristic, A is age, and Q... mean The mean of the QT interval characteristics, Q std denoted as , where is the standard deviation of the QT interval characteristics. w1, w2, w3, w4, w5, w6, and w7 represent the weights of blood glucose variability, glycated hemoglobin level, blood pressure variability, heart rate variability, urine albumin / creatinine ratio, QT interval characteristics, and age, respectively. c1, c2, and c3 are the coefficients of the nonlinear transformation. α is the weight of the linear combination component, β is the weight of the network structure component, k1 is the weight of node correlation, and k2 is the weight of node path information. To calculate the sum of the correlation degrees of all early warning feature nodes, It is the sum of path information for all warning feature nodes.

9. The method for early warning of diabetic cardiovascular complications according to claim 1, characterized in that, In S400, the pre-set warning rule is: When the cardiovascular complication score is less than or equal to 0.3, it is considered low risk and a Level 1 warning is issued. When the cardiovascular complication score is greater than 0.3 and less than 0.7, it is considered a moderate risk and a level-two warning is issued. When the cardiovascular complication score is greater than or equal to 0.7, it is considered high risk and a level 3 warning is issued.

10. A diabetic cardiovascular complication early warning system, characterized in that, include: The first acquisition module is used to acquire the health data of the diabetic patient to be evaluated, encrypt the health data of the diabetic patient to be evaluated, and acquire the encrypted health data. The health data of the diabetic patients to be evaluated includes: basic information, blood glucose data, ambulatory blood pressure data, heart rate data, urine albumin data, creatinine data, and QT interval data of the diabetic patients to be evaluated, obtained through multiple participants. The multiple participating parties include different hospitals and different departments within those hospitals; The early warning feature acquisition module is used to input the encrypted health data into a pre-set early warning feature acquisition model to acquire early warning features of cardiovascular complications. The early warning feature acquisition model includes: a blood glucose fluctuation acquisition sub-model, a blood pressure fluctuation acquisition sub-model, a heart rate variability acquisition sub-model, a urine albumin / creatinine ratio acquisition sub-model, a QT interval feature acquisition sub-model, and a glycated hemoglobin acquisition sub-model. The data analysis module is used to input the early warning features of the cardiovascular complications into a pre-set cardiovascular complication scoring model, obtain the cardiovascular complication score of the diabetic patient to be evaluated, and issue an early warning according to the pre-set early warning rules.

Images