Intelligent illness state evaluation and early warning system for emergency internal medicine department

Abstract

The invention, which relates to the technical field of the emergency internal medicine department, discloses an intelligent assessment and early warning system for the illness state of the emergency internal medicine department, and the system carries out quantitative assessment based on pathological characteristics of multiple dimensions through a comprehensive assessment and early warning module and a blood sugar fluctuation and fundus disease analysis module. The integration of the data can form a comprehensive fundus lesion risk factor Rrisk, and the fundus lesion risk of a diabetic patient is subjected to all-around quantitative evaluation. According to the system, fundus lesion risks of patients are divided through set risk levels, and intervention suggestions of different levels are provided in combination with a comprehensive fundus lesion risk factor Rrisk. When the patient is at a high risk level, the system can trigger related early warning in time to remind medical staff to take targeted treatment or further examination, and serious consequences such as disease development to blindness and the like are avoided. The multi-dimensional comprehensive evaluation not only improves the diagnosis accuracy of the fundus lesion of the patient, but also provides powerful decision support for clinical treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency internal medicine, and specifically to an intelligent evaluation and warning system for emergency internal medicine conditions. Background Art

[0002] Emergency internal medicine is a key department in the medical field, responsible for dealing with various acute conditions and sudden diseases, especially important in the treatment of acute complications of chronic diseases such as diabetes. With the aging of the population and the prevalence of chronic diseases, the number of diabetes patients is increasing year by year, and the acute complications caused by diabetes, especially diabetic retinopathy, in emergency internal medicine, the fundus lesions of diabetes are a key diagnosis and treatment point. Diabetic retinopathy poses a major threat to the eye health of patients, with a complex pathogenesis and being closely related to multiple factors such as blood glucose fluctuations, disease duration, and retinal microvascular damage in patients. Therefore, the early detection and accurate evaluation of diabetic retinopathy have become one of the core problems urgently to be solved in emergency internal medicine.

[0003] At present, the evaluation of diabetic retinopathy in emergency internal medicine mostly relies on traditional diagnostic methods such as fundus examination and blood glucose monitoring. However, these methods have some deficiencies. First, although fundus examination is an important means to judge retinal lesions, this method requires the professional experience of doctors, has a long examination cycle and poor real-time performance, and is difficult to meet the rapid treatment needs of emergencies. Second, the relationship between blood glucose fluctuations and fundus lesions is relatively complex, and traditional evaluation methods cannot accurately quantify the long-term impact of blood glucose fluctuations on fundus lesions. In addition, existing evaluation systems usually ignore the non-linear impact of diabetes duration on fundus lesions and lack a data-driven comprehensive evaluation tool. Therefore, existing evaluation methods have large defects in accuracy, real-time performance, and personalized management, and cannot effectively support the rapid and accurate decision-making needs of emergency internal medicine. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides an intelligent evaluation and warning system for emergency internal medicine conditions, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: including a disease duration extraction module, a feature extraction module, a blood glucose fluctuation and fundus disease analysis module, a comprehensive evaluation and warning module, and a deep evaluation and warning module;

[0006] The disease duration extraction module extracts the disease duration data in the electronic health record EHR of diabetes patients by constructing a disease condition evaluation and warning platform, and at the same time constructs a time series database, connects the time series database with the disease condition evaluation and warning platform, and stores the collected disease duration data into the time series database;

[0007] The feature extraction module preprocesses by extracting the disease course data from the time series database to obtain a pathological data set, and extracts features from the pathological data set to obtain a pathological feature vector;

[0008] The blood glucose fluctuation and fundus disease analysis module constructs a disease course algorithm model and a fundus disease risk algorithm model, inputs the obtained pathological feature vector into the disease course algorithm model and the fundus disease risk algorithm model, and calculates respectively to output the disease course impact factor Rdt, the risk contribution value Rwx of retinal microvascular damage, and the risk contribution value Rbv of blood glucose fluctuation;

[0009] The comprehensive evaluation and early warning module comprehensively calculates the output results of the disease course algorithm model and the fundus disease risk algorithm model to output the comprehensive fundus lesion risk factor Rrisk, and then preliminarily judges the risk of diabetic fundus lesions based on the output result of the comprehensive fundus lesion risk factor Rrisk, and divides the fundus lesions of diabetic patients into different risk levels based on the preliminary judgment result;

[0010] The risk levels include first-level risk, second-level risk, and third-level risk;

[0011] When the deep evaluation and early warning module identifies that a diabetic patient is at second-level risk and third-level risk, it triggers a deep analysis. The deep analysis performs a deep calculation based on the comprehensive fundus lesion risk factor Rrisk to output a deep risk factor Rdeep, and based on the output result of the deep risk factor Rdeep, it deeply determines the degree of fundus lesions, and generates relevant early warning measures based on the deep determination result.

[0012] Preferably, the disease course extraction module includes a disease course extraction unit and a data storage unit;

[0013] The disease course extraction unit accesses the electronic health record EHR of the hospital through the set API application program interface to extract the disease course data of diabetic patients;

[0014] The disease course data includes patient information, diabetes duration Dt, blood glucose measurement B, and fundus examination images;

[0015] The data storage unit constructs a time series database, sets a write port and a write port, stores the disease course data in the time series database according to the extraction timestamp through the write port, and at the same time generates a unique character ID based on the patient information through coding technology, and each unique character ID corresponds to the disease course data of the relevant diabetic patient.

[0016] Preferably, the feature extraction module includes a preprocessing unit and a feature extraction unit;

[0017] The preprocessing unit extracts the course data through the write port and preprocesses the course data. The preprocessing methods include image preprocessing and data preprocessing;

[0018] The image preprocessing uses AI image processing technology on the fundus examination images in the course data to intelligently identify the number of retinal microhemorrhages Nmic, the area of the retinal lesion region Aretina, the area of the retinal edema region Aedema, and the total area of the retina Azqymj in the fundus examination images;

[0019] The data preprocessing performs data cleaning and normalization on the diabetes duration Dt and blood glucose measurement B in the course data, combined with the obtained number of retinal microhemorrhages Nmic, the area of the retinal lesion region Aretina, the area of the retinal edema region Aedema, and the total area of the retina Azqymj, normalizing all the data in the course data to the range of [0, 1], removing outliers, missing data, and eliminating the influence of dimensions;

[0020] The obtained number of retinal microhemorrhages Nmic, the area of the retinal lesion region Aretina, the area of the retinal edema region Aedema, the total area of the retina Azqymj, the diabetes duration Dt, and the blood glucose measurement B after preprocessing are summarized to generate a pathological data set;

[0021] The feature extraction unit performs feature extraction based on the obtained pathological data set to obtain a pathological feature vector;

[0022] The pathological feature vector includes the diabetes duration Dt, the blood glucose measurement B, the blood glucose fluctuation index Bv, the retinal microvascular injury index Rmi, and the retinal edema index Rsw;

[0023] The blood glucose fluctuation index Bv is extracted by calculating the standard deviation based on each detected blood glucose measurement B. The specific extraction formula is: , where n represents the total number of measurements, Bi represents the i-th blood glucose measurement value, represents the average blood glucose measurement;

[0024] The retinal microvascular injury index Rmi is calculated and extracted based on the number of retinal microhemorrhages Nmic and the area of the retinal lesion region Aretina. The specific extraction formula is: ;

[0025] The retinal edema index Rsw is calculated and extracted based on the area of the retinal edema region Aedema and the total area of the retina Azqymj. The specific extraction formula is: .

[0026] Preferably, the blood glucose fluctuation and fundus disease analysis module includes a disease course lesion analysis unit, a microvascular injury analysis unit, and a blood glucose analysis unit;

[0027] The disease course lesion analysis unit extracts the diabetes duration Dt in the pathological feature vector and inputs it into the disease course algorithm model to calculate and output the disease course impact factor Rdt, and analyzes the impact of the disease course on fundus lesions;

[0028] The disease course impact factor Rdt is calculated and output through the following disease course algorithm model;

[0029] ;

[0030] In the formula, tanh represents the hyperbolic tangent function, represents the parameter that regulates the impact of the disease course on fundus lesions, e represents the exponential function, represents the exponential decay term, represents the sensitivity that regulates the impact of the disease course on risk changes, represents the exponential decay term adjustment coefficient, represents the speed control coefficient of the exponential decay term.

[0031] Preferably, the microvascular injury analysis unit constructs a fundus disease risk algorithm model, and the fundus disease risk algorithm model includes a microvascular injury risk algorithm model and a blood glucose fluctuation risk algorithm model;

[0032] At the same time, the retinal microvascular injury index Rmi in the pathological feature vector is extracted and input into the microvascular injury risk algorithm model for calculation to output the risk contribution value Rwx of retinal microvascular injury, and the degree of microvascular injury of the retina is quantified;

[0033] The risk contribution value Rwx of retinal microvascular injury is calculated and output through the following microvascular injury risk algorithm model;

[0034] ;

[0035] In the formula, ln represents the natural logarithm, represents the logarithmic transformation term, represents the logarithmic transformation term adjustment coefficient, Rmi 3 represents the cubic term of the retinal microvascular injury index, represents the cubic term adjustment coefficient.

[0036] Preferably, the blood glucose analysis unit extracts the blood glucose fluctuation index Bv in the pathological feature vector and inputs it into the blood glucose fluctuation risk algorithm model for calculation to output the blood glucose fluctuation risk contribution value Rbv, and quantifies the impact of blood glucose fluctuation on fundus lesions;

[0037] The blood glucose fluctuation risk contribution value Rbv is calculated and output through the following blood glucose fluctuation risk algorithm model;

[0038] ;

[0039] In the formula, e represents the exponential function, represents the exponential term of the blood glucose fluctuation index, represents the amplification control coefficient of the exponential term of the blood glucose fluctuation index, sin represents the sine function, represents the sensitivity control coefficient of the blood glucose fluctuation index, represents the periodic control coefficient of the blood glucose fluctuation index.

[0040] Preferably, the comprehensive evaluation and early warning module includes a comprehensive summary unit and a grading evaluation unit;

[0041] The comprehensive summary unit combines the blood glucose fluctuation, the course influence factor Rdt output by the fundus disease analysis module, the risk contribution value Rwx of retinal microvascular damage, and the blood glucose fluctuation risk contribution value Rbv, and combines the retinal edema index Rsw to perform comprehensive calculation and output the comprehensive fundus disease risk factor Rrisk, so as to comprehensively quantify the risk of fundus disease;

[0042] The comprehensive fundus disease risk factor Rrisk is calculated and output through the following algorithm formula;

[0043] ;

[0044] In the formula, represents the retinal edema index adjustment coefficient.

[0045] Preferably, the grading evaluation unit sets the fundus disease risk interval [0.3, 0.6] for diabetic patients based on the fundus disease risk by the user, and then makes a preliminary judgment on the fundus disease risk situation of the current diabetic patient based on the output result of the comprehensive fundus disease risk factor Rrisk, and classifies the lesion grade of the diabetic patient. The specific judgment content is as follows;

[0046] When the comprehensive fundus disease risk factor Rrisk < 0.3, it is marked as a first-level risk;

[0047] When 0.3 ≤ the comprehensive fundus disease risk factor Rrisk < 0.6, it is marked as a second-level risk;

[0048] When the comprehensive fundus disease risk factor Rrisk ≥ 0.6, it is marked as a third-level risk;

[0049] Among them, the lesion risk levels increase from low to high, with the first-level risk < the second-level risk < the third-level risk.

[0050] Preferably, the depth evaluation and early warning module includes a depth analysis unit and a depth evaluation unit;

[0051] The depth analysis unit triggers depth analysis after a preliminary judgment of secondary risk and tertiary risk. The depth analysis extracts the average glycated hemoglobin value HbA1c of diabetic patients within 3 months from the electronic health record EHR, and calculates the blood glucose control index Dci. The specific algorithm formula is: , where HbA1c represents the average glycated hemoglobin value, 6.5 represents the diabetes diagnosis standard, and 8.5 represents abnormal blood glucose control;

[0052] Then, based on the combination of the blood glucose control index Dci and the comprehensively determined risk factor Rrisk of diabetic retinopathy, the depth risk factor Rdeep is output to conduct a depth analysis of the risk of diabetic retinopathy;

[0053] The depth risk factor Rdeep is calculated and output through the following algorithm formula;

[0054] ;

[0055] In the formula, represents the blood glucose control index adjustment coefficient, represents the second adjustment coefficient of the retinal edema index.

[0056] Preferably, the depth evaluation unit extracts the historical data of diabetes from the electronic health record EHR of the hospital through the API application programming interface, combines the clinical examination results and treatment results, and sets the lesion degree interval [0.4, 0.7] by the user. Then, the output result of the depth risk factor Rdeep is used to conduct a depth determination of the risk of diabetic retinopathy in patients. The specific determination content is as follows;

[0057] When the depth risk factor Rdeep < 0.4, a regular examination is prompted at this time;

[0058] When 0.4 ≤ depth risk factor Rdeep < 0.7, a first warning message is generated at this time to prompt the medical staff to adjust the medication for diabetic patients;

[0059] When the depth risk factor Rdeep ≥ 0.7, a second warning message is generated at this time to prompt the medical staff to intervene in the treatment.

[0060] The present invention provides an intelligent evaluation and early warning system for emergency internal medicine conditions. It has the following beneficial effects:

[0061] (1) By constructing a disease course extraction module, a feature extraction module, and a blood glucose fluctuation and fundus disease analysis module, the system realizes the intelligent collection and analysis of the disease course data, blood glucose monitoring, and fundus examination images of diabetic patients. The disease course extraction module extracts the disease course data of diabetic patients by accessing the hospital's electronic health record EHR and stores it in a time series database, providing a basis for the subsequent extraction of pathological data sets and the generation of pathological feature vectors. Based on the pathological feature vectors, the fundus lesion risk of patients is comprehensively evaluated, high-risk patients are identified at an early stage, and timely warnings are issued. Through the comprehensive analysis of the disease course impact factor Rdt, the risk contribution value Rwx of retinal microvascular damage, and the risk contribution value Rbv of blood glucose fluctuation, the system can provide targeted warnings, providing a basis for the early intervention of diabetic retinopathy and significantly improving the early diagnosis and intervention effects of fundus lesions.

[0062] (2) Through the comprehensive evaluation warning module and the blood glucose fluctuation and fundus disease analysis module, the system conducts a quantitative evaluation based on pathological features in multiple dimensions, including diabetes duration, blood glucose fluctuation, retinal microvascular damage, and retinal edema. The fusion of these data can form a comprehensive fundus lesion risk factor Rrisk, providing a comprehensive quantitative evaluation of the fundus lesion risk of diabetic patients. The system divides the fundus lesion risk of patients according to the set risk levels and provides intervention suggestions at different levels in combination with the comprehensive fundus lesion risk factor Rrisk. When the patient is at a high-risk level, the system can promptly trigger relevant warnings, reminding medical staff to take targeted treatments or further examinations to avoid serious consequences such as blindness due to the progression of the disease. This multi-dimensional comprehensive evaluation not only improves the diagnostic accuracy of patients' fundus lesions but also provides strong decision-making support for clinical treatment, ensuring that patients can obtain the best treatment and management.

[0063] (3) Through the in-depth evaluation warning module, in combination with the blood glucose control situation of diabetic patients and the fundus lesion risk factor Rrisk, the system provides a personalized risk assessment mechanism. When the patient's comprehensive fundus lesion risk factor Rrisk exceeds the set risk threshold, the system can output an in-depth risk factor Rdeep through in-depth analysis and calculation and conduct corresponding interventions according to the patient's specific condition. Especially when the fundus lesion risk level is grade two or three, the in-depth evaluation module will conduct in-depth analysis in combination with the patient's glycated hemoglobin value HbA1c and blood glucose control situation, accurately evaluate the progression of diabetic retinopathy, and trigger necessary clinical intervention measures, such as increasing the frequency of fundus examinations or adjusting the treatment plan. Through this personalized intervention mechanism, the system can effectively reduce the aggravation of fundus lesions caused by the patient's failure to receive timely intervention, improving the treatment effect and the patient's quality of life. Description of the Drawings

[0064] Figure 1This is a schematic diagram of the process of an intelligent evaluation and warning system for emergency internal medicine conditions of the present invention. Specific implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] Embodiment 1

[0067] Please refer to Figure 1 , the present invention provides an intelligent evaluation and warning system for emergency internal medicine conditions. To achieve the above objectives, the present invention is realized through the following technical solutions: including a course extraction module, a feature extraction module, a blood glucose fluctuation and fundus disease analysis module, a comprehensive evaluation and warning module, and a depth evaluation and warning module;

[0068] The course extraction module extracts the course data in the electronic health record EHR of diabetic patients by constructing a disease condition evaluation and warning platform, and at the same time constructs a time series database. After connecting the time series database with the disease condition evaluation and warning platform, the collected course data is stored in the time series database;

[0069] The feature extraction module preprocesses the course data in the time series database to obtain a pathological data set, and performs feature extraction on the pathological data set to obtain a pathological feature vector;

[0070] The blood glucose fluctuation and fundus disease analysis module calculates the course impact factor Rdt, the risk contribution value Rwx of retinal microvascular damage, and the risk contribution value Rbv of blood glucose fluctuation by constructing a course algorithm model and a fundus disease risk algorithm model, and inputs the obtained pathological feature vector into the course algorithm model and the fundus disease risk algorithm model for separate calculations;

[0071] The comprehensive evaluation and warning module comprehensively calculates the output results of the course algorithm model and the fundus disease risk algorithm model to output the comprehensive fundus disease risk factor Rrisk, and then preliminarily judges the risk of diabetic fundus disease based on the output result of the comprehensive fundus disease risk factor Rrisk, and divides the fundus disease of diabetic patients into different risk levels based on the preliminary judgment result;

[0072] The risk levels include first-level risk, second-level risk, and third-level risk;

[0073] When the in-depth evaluation and warning module identifies a diabetic patient as a secondary risk or a tertiary risk, it triggers in-depth analysis. The in-depth analysis is based on the comprehensive fundus lesion risk factor Rrisk to calculate and output the in-depth risk factor Rdeep, and based on the output result of the in-depth risk factor Rdeep, it determines the degree of fundus lesions in depth, and generates relevant warning measures based on the in-depth determination result.

[0074] In this embodiment, the disease course extraction module of the system accesses the electronic health record EHR of diabetic patients and stores the disease course data in the time series database, ensuring the efficient management and access of data. The feature extraction module preprocesses the disease course data in the time series database, obtains the pathological data set, and extracts the pathological feature vector, providing detailed pathological information support and ensuring the accuracy and comprehensiveness of the data. The blood glucose fluctuation and fundus disease analysis module analyzes the pathological feature vector by constructing a disease course algorithm model and a fundus disease risk algorithm model, and calculates and outputs the disease course impact factor Rdt, the risk contribution value Rwx of retinal microvascular damage, and the risk contribution value Rbv of blood glucose fluctuation respectively, quantifying the risk factors of fundus lesions from multiple dimensions and providing accurate data support for comprehensive evaluation. The comprehensive evaluation and warning module comprehensively calculates each risk factor, outputs the comprehensive fundus lesion risk factor Rrisk, and makes a preliminary judgment on the fundus lesion risk of diabetic patients, and divides the risk level according to the comprehensive risk factor, thus realizing the effective warning of fundus lesion risk. When a diabetic patient is rated as a secondary risk or a tertiary risk, the in-depth evaluation and warning module starts in-depth analysis, further calculates the in-depth risk factor Rdeep, combines information such as the patient's glycated hemoglobin value, accurately evaluates the severity of fundus lesions, and generates corresponding warning measures based on the evaluation result, so as to provide more targeted intervention. Compared with the prior art, the present invention breaks through the limitations of the traditional method for evaluating the risk of diabetic retinopathy through real-time collection and in-depth analysis of multi-dimensional data, and can more comprehensively and dynamically monitor the changes in the patient's disease course and its impact on fundus lesions. The prior art mostly relies on a single data source or static diagnostic means, while this system not only can identify the risk of fundus lesions earlier through intelligent data processing and combined analysis of multiple models, but also can provide personalized warnings and interventions according to different risk levels, greatly improving the accuracy of warning and the timeliness of treatment, and promoting the intelligentization and precision of diabetic retinopathy management.

[0075] Embodiment 2

[0076] Specifically: The disease course extraction module includes a disease course extraction unit and a data storage unit;

[0077] The disease course extraction unit extracts the disease course data of diabetic patients by setting an API application program interface to connect the disease condition evaluation and warning platform to the hospital's electronic health record EHR.

[0078] The course data includes patient information, diabetes duration Dt, blood glucose measurement B, and fundus examination images;

[0079] The data storage unit constructs a time-series database, sets up a write-out port and a write-in port, stores the course data into the time-series database according to the extraction timestamp through the write-in port, and at the same time generates a unique character ID based on the patient information through coding technology. Each unique character ID corresponds to the course data of the relevant diabetic patient.

[0080] In this embodiment, the course extraction unit of the system connects the disease condition assessment and early warning platform with the hospital's electronic health record EHR system through an API interface, realizing the automatic extraction of the course data of diabetic patients. This data collection process avoids the errors and delays of manual entry, ensuring the real-time and accuracy of the data. The data storage unit constructs a time-series database and stores the course data in an orderly manner according to the timestamp, which not only improves the efficiency of data access and storage, but also generates a unique character ID for each patient through a unique coding technology, ensuring the precise matching and personalized management of the course data and patient information. This module has obvious advantages in the existing technology, breaking through the limitations of traditional manual input and static database management, and improving the automation level of data processing. Compared with the existing technology, the traditional methods of collecting and storing course data often have problems such as data redundancy, information asymmetry, and processing delays. The innovation of this system lies in the combination of a time-series database and a unique ID coding, which can not only extract and store course data in real time and accurately, but also achieve fast retrieval and multi-dimensional analysis through structured data storage, providing a solid data foundation for subsequent data processing and intelligent assessment.

[0081] Embodiment 3

[0082] Specifically: The feature extraction module includes a preprocessing unit and a feature extraction unit;

[0083] The preprocessing unit extracts the course data through the write-out port and preprocesses the course data. The preprocessing methods include image preprocessing and data preprocessing;

[0084] The image preprocessing uses AI image processing technology for the fundus examination images in the course data to intelligently identify the number of retinal microhemorrhages Nmic, the area of the retinal lesion region Aretina, the area of the retinal edema region Aedema, and the total area of the retina Azqymj in the fundus examination images;

[0085] Data preprocessing performs data cleaning and normalization on the diabetes duration Dt and blood glucose measurement B in the disease course data, in combination with the obtained number of retinal microhemorrhages Nmic, retinal lesion area Aretina, retinal edema area Aedema, and total retinal area Azqymj, normalizing all the data in the disease course data to the range of [0, 1], removing outliers, missing data, and eliminating the influence of dimensions;

[0086] The obtained number of retinal microhemorrhages Nmic, retinal lesion area Aretina, retinal edema area Aedema, total retinal area Azqymj, diabetes duration Dt, and blood glucose measurement B after preprocessing are summarized to generate a pathological data set;

[0087] The feature extraction unit performs feature extraction on the basis of the obtained pathological data set to obtain a pathological feature vector;

[0088] The pathological feature vector includes diabetes duration Dt, blood glucose measurement B, blood glucose fluctuation index Bv, retinal microvascular injury index Rmi, and retinal edema index Rsw;

[0089] The blood glucose fluctuation index Bv is extracted by calculating the standard deviation based on each detected blood glucose measurement B. The specific extraction formula is: , where n represents the total number of measurements, Bi represents the i-th blood glucose measurement value, represents the average blood glucose measurement;

[0090] The retinal microvascular injury index Rmi is calculated and extracted based on the number of retinal microhemorrhages Nmic and the retinal lesion area Aretina. The specific extraction formula is: ;

[0091] The retinal edema index Rsw is calculated and extracted based on the retinal edema area Aedema and the total retinal area Azqymj. The specific extraction formula is: .

[0092] In this embodiment, the system performs image preprocessing and data preprocessing on the course data through a preprocessing unit, and intelligently identifies key indicators such as the number of retinal microhemorrhages, the area of retinal lesion regions, and the area of retinal edema regions in fundus examination images through AI image processing technology. Data preprocessing further optimizes the course data. Through data cleaning and normalization, outliers are removed, missing data is filled, and the influence of dimensions is eliminated, thus ensuring the consistency and comparability of the data. This precise data processing method lays a solid foundation for subsequent analysis and feature extraction. The feature extraction unit extracts pathological feature vectors from the pathological dataset, providing high-quality input data for comprehensive evaluation. In particular, the extraction of features such as the blood glucose fluctuation index, the retinal microvascular damage index, and the retinal edema index further strengthens the quantitative evaluation of disease risks, making the risk prediction of diabetic retinopathy more scientific and accurate. Compared with the prior art, traditional pathological data processing often relies on manual analysis, with high errors and inefficiencies. Through the innovative image processing and data cleaning methods of this module, the system can automatically extract multi-dimensional pathological features from complex course data, greatly improving the processing efficiency and accuracy. At the same time, the standardized and quantitative manner of the feature extraction process makes the evaluation of the risk of diabetic retinopathy more objective, avoiding the interference of human factors.

[0093] Embodiment 4

[0094] Specifically: The blood glucose fluctuation and fundus disease analysis module includes a course lesion analysis unit, a microvascular damage analysis unit, and a blood glucose analysis unit;

[0095] The course lesion analysis unit extracts the diabetes duration Dt from the pathological feature vector, inputs it into the course algorithm model, calculates and outputs the course impact factor Rdt, and analyzes the impact of the course on fundus lesions;

[0096] The course impact factor Rdt is calculated and output through the following course algorithm model;

[0097] ;

[0098] In the formula, tanh represents the hyperbolic tangent function, which is used to simulate the characteristic that the lesion risk increases rapidly in the initial stage of the course, and then the growth rate gradually slows down. When the course is very small, the risk increases rapidly; when the course increases to a certain extent, the growth tends to be stable. represents the parameter that adjusts the impact of the course on fundus lesions, controls the amplitude of the hyperbolic tangent function part. The larger this value is, the stronger the impact of the course on fundus lesions. e represents the exponential function. represents the exponential decay term. represents the sensitivity that adjusts the change of risk by the course. represents the adjustment coefficient of the exponential decay term. The velocity control coefficient representing the exponential decay term;

[0099] The impact of the patient's diabetes duration on fundus lesions is not linear. Instead, the risk gradually increases as the duration prolongs. However, after a too-long duration, the impact may tend to level off. The hyperbolic tangent function tanh is used to model and analyze the impact of the duration.

[0100] The microvascular injury analysis unit constructs a fundus disease risk algorithm model, which includes a microvascular injury risk algorithm model and a blood glucose fluctuation risk algorithm model;

[0101] At the same time, the retinal microvascular injury index Rmi in the pathological feature vector is extracted and input into the microvascular injury risk algorithm model for calculation to output the risk contribution value Rwx of retinal microvascular injury, so as to quantify the degree of microvascular injury of the retina;

[0102] The risk contribution value Rwx of retinal microvascular injury is calculated and output through the following microvascular injury risk algorithm model;

[0103] ;

[0104] In the formula, ln represents the natural logarithm, represents the logarithmic transformation term, represents the logarithmic transformation term adjustment coefficient, controlling that the risk change of initial microvascular injury is relatively gentle, Rmi 3 represents the cubic term of the retinal microvascular injury index, indicating that as the severity of microvascular injury increases, the risk intensifies, represents the cubic term adjustment coefficient, controlling the degree of sharp increase in its risk during severe microvascular injury;

[0105] Retinal microvascular injury is an important indication of diabetic retinopathy, and its impact has strong non-linear characteristics. The formula combining logarithmic transformation and polynomial is used to model the impact of microvascular injury on fundus lesions.

[0106] The blood glucose analysis unit extracts the blood glucose fluctuation index Bv in the pathological feature vector, inputs it into the blood glucose fluctuation risk algorithm model for calculation to output the blood glucose fluctuation risk contribution value Rbv, so as to quantify the impact of blood glucose fluctuation on fundus lesions;

[0107] The blood glucose fluctuation risk contribution value Rbv is calculated and output through the following blood glucose fluctuation risk algorithm model;

[0108] ;

[0109] In the formula, e represents the exponential function, represents the exponential term of the blood glucose fluctuation index, The amplification control coefficient of the exponential term representing the blood glucose fluctuation index, and sin represents the sine function. The sensitivity control coefficient of the blood glucose fluctuation index. The periodic control coefficient of the blood glucose fluctuation index, which is used to control the impact of postprandial blood glucose fluctuations.

[0110] In this embodiment, the disease course lesion analysis unit of the system analyzes the impact of the disease course on fundus lesions through the disease course algorithm model. It accurately simulates the rapid growth characteristics of fundus lesions in the initial stage of the disease course using the hyperbolic tangent function tanh, and presents a risk growth pattern that tends to be stable after the disease course prolongs. This non-linear model not only reveals the progressive impact of the diabetes disease course on fundus lesions, but also can finely adjust the degree of influence of the disease course on risk changes, providing a more accurate disease course factor Rdt for risk assessment, breaking through the linear assumption in traditional methods and providing a more practical analysis of the impact of the disease course. The microvascular injury analysis unit deeply analyzes the impact of retinal microvascular injury through the fundus disease risk algorithm model, especially the microvascular injury risk algorithm model. Using the logarithmic transformation and polynomial combination formula, it successfully models the complex non-linear relationship between retinal microvascular injury and fundus lesions, thereby obtaining the risk contribution value Rwx of retinal microvascular injury. This method not only realizes the quantification of microvascular injury, but also reveals the different impacts of different degrees of injury on fundus lesions, providing more detailed and accurate injury assessment results. The blood glucose analysis unit focuses on the impact of blood glucose fluctuations on fundus lesions, and quantifies the contribution of the blood glucose fluctuation index Bv to fundus lesions through the blood glucose fluctuation risk algorithm model. This model uses a combination of exponential function and sine function to effectively simulate the periodic changes of blood glucose fluctuations and their long-term impact on fundus lesions. Especially in the sensitivity control of postprandial blood glucose fluctuations, it further improves the accurate analysis of blood glucose fluctuations. Compared with the prior art, this module provides a more comprehensive and accurate assessment of diabetic fundus lesions by combining non-linear modeling and multi-dimensional data analysis.

[0111] Embodiment 5

[0112] Specifically: The comprehensive evaluation and early warning module includes a comprehensive summary unit and a grading evaluation unit;

[0113] The comprehensive summary unit combines the disease course impact factor Rdt, the risk contribution value Rwx of retinal microvascular injury, and the blood glucose fluctuation risk contribution value Rbv output by the blood glucose fluctuation and fundus disease analysis module, and combines the retinal edema index Rsw to perform comprehensive calculation and output the comprehensive fundus lesion risk factor Rrisk to comprehensively quantify the fundus lesion risk.

[0114] The comprehensive fundus lesion risk factor Rrisk is calculated and output through the following algorithm formula;

[0115] ;

[0116] In the formula, represents the retinal edema index adjustment coefficient, which is used to control the final impact degree of retinal edema on the risk.

[0117] The grading evaluation unit is set by the user based on the risk of fundus lesions to determine the risk interval [0.3, 0.6] of diabetic patients' fundus lesions. Then, based on the output result of the comprehensive fundus lesion risk factor Rrisk, the current risk situation of the diabetic patients' fundus lesions is initially determined, and the lesion grades of the diabetic patients are divided. The specific determination content is as follows;

[0118] When the comprehensive fundus lesion risk factor Rrisk < 0.3, it is marked as a first-level risk. The fundus lesion risk of such patients is relatively low, and it is recommended to regularly control blood sugar and conduct fundus examinations;

[0119] When 0.3 ≤ the comprehensive fundus lesion risk factor Rrisk < 0.6, it is marked as a second-level risk. The fundus lesion risk of such patients is medium, and it is recommended to further increase the frequency of fundus examinations and pay attention to blood sugar fluctuations;

[0120] When the comprehensive fundus lesion risk factor Rrisk ≥ 0.6, it is marked as a third-level risk. The fundus lesion risk of such patients is relatively high, and the system recommends in-depth fundus examinations and possible intervention measures;

[0121] Among them, the lesion risk levels are from low to high, first-level risk < second-level risk < third-level risk.

[0122] In this embodiment, the comprehensive summarization unit of the system calculates the comprehensive fundus disease risk factor Rrisk by integrating the course influence factor Rdt, the risk contribution value Rwx of retinal microvascular damage, the risk contribution value Rbv of blood glucose fluctuation, and the retinal edema index Rsw output by the blood glucose fluctuation and fundus disease analysis module, and realizes the quantitative and refined analysis of the fundus disease risk. This algorithm effectively reflects the comprehensive risk of fundus diseases in diabetic patients through the weighted integration of various risk factors, significantly improving the accuracy and meticulousness of risk assessment. The grading evaluation unit then makes a preliminary determination of the comprehensive fundus disease risk factor Rrisk based on the set risk interval and grades the fundus disease risk of diabetic patients. By classifying patients into first-level risk, second-level risk, and third-level risk grades, the system can provide personalized medical advice according to different risk grades. For patients at first-level risk, regular blood glucose control and fundus examinations are recommended; for patients at second-level risk, it is recommended to increase the frequency of fundus examinations and pay attention to blood glucose fluctuations; while for patients at third-level risk, in-depth fundus examinations and possible intervention measures are prompted. This refined classification based on risk levels can provide accurate decision-making support for clinicians, ensuring the early detection and timely intervention of fundus disease risks.

[0123] Example 6

[0124] Specifically: The in-depth evaluation and warning module includes an in-depth analysis unit and an in-depth evaluation unit;

[0125] The in-depth analysis unit triggers in-depth analysis after initially judging as second-level risk and third-level risk. The in-depth analysis extracts the average glycated hemoglobin value HbA1c of diabetic patients within 3 months from the electronic health record EHR and calculates the blood glucose control index Dci. The specific algorithm formula is: , where HbA1c represents the average glycated hemoglobin value, the normalized value, 6.5 represents the diabetes diagnosis standard, and 8.5 represents abnormal blood glucose control;

[0126] Then, by combining the blood glucose control index Dci with the initially determined comprehensive fundus disease risk factor Rrisk, the in-depth risk factor Rdeep is output to conduct in-depth analysis of the fundus disease risk of diabetic patients;

[0127] The in-depth risk factor Rdeep is calculated and output through the following algorithm formula;

[0128] ;

[0129] In the formula, represents the blood glucose control index adjustment coefficient, represents the second adjustment coefficient of the retinal edema index.

[0130] The depth assessment unit extracts the historical data of diabetes from the hospital's electronic health record (EHR) through the API (Application Programming Interface), combines the clinical examination results and treatment outcomes, and sets the lesion severity range [0.4, 0.7] by the user. Then, it outputs the obtained deep risk factor Rdeep to determine the risk of fundus lesions in diabetic patients. The specific determination content is as follows;

[0131] When the deep risk factor Rdeep < 0.4, it indicates that the fundus lesions of diabetic patients are normal, and regular examinations are prompted at this time;

[0132] When 0.4 ≤ deep risk factor Rdeep < 0.7, it indicates a group of patients whose fundus lesions are gradually worsening but have not reached the level of urgent intervention. At this time, a first warning message is generated to prompt medical staff to adjust the medications for diabetic patients;

[0133] When the deep risk factor Rdeep ≥ 0.7, it indicates that the fundus lesions of diabetic patients are relatively severe. At this time, a second warning message is generated to prompt medical staff for intervention treatment, such as laser treatment and anti-VEGF drugs, etc.

[0134] In this embodiment, after the preliminary risk assessment is at level two and level three risks, the depth analysis unit triggers the depth analysis to deeply evaluate the blood glucose control situation of the patient. By extracting the glycated hemoglobin value HbA1c from the electronic health record (EHR) and calculating the blood glucose control index Dci, combined with the comprehensive fundus lesion risk factor Rrisk, the deep risk factor Rdeep is obtained. The calculation of this risk factor further considers the patient's blood glucose control level and the specific situation of fundus lesions, thus providing a more comprehensive analysis basis for the judgment of fundus lesion risk. The depth assessment unit then deeply determines the deep risk factor by combining the clinical examination results, historical data, and treatment effects. According to the calculation results, corresponding warning messages are automatically generated. This module can not only timely detect the progression of fundus lesions in diabetic patients but also accurately provide intervention suggestions for medical staff, such as medication adjustment or laser treatment, etc., significantly improving the pertinence and effectiveness of treatment. Compared with the prior art, the main innovation of the depth assessment and warning module lies in the dynamic monitoring of blood glucose control and the combined assessment of fundus lesion risk. Traditional assessment methods may only rely on a single indicator of blood glucose or fundus lesions, while this module greatly improves the accuracy and timeliness of assessment by integrating multi-dimensional data and conducting in-depth analysis. Through the detailed tracking and real-time warning of the patient's condition, this system can effectively reduce the missed diagnosis rate or misdiagnosis rate of fundus lesions, promote the formulation of personalized treatment plans, thereby improving the prevention and treatment effect of diabetic retinopathy and improving the treatment prognosis of patients.

[0135] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

Claims

1. An intelligent assessment and early warning system for emergency internal medicine conditions, characterized by: It includes disease course extraction module, feature extraction module, blood sugar fluctuation and fundus disease analysis module, comprehensive assessment and early warning module and in-depth assessment and early warning module; The disease course extraction module extracts the disease course data in the electronic health record (EHR) of the diabetic patient by constructing a disease assessment and early warning platform, and at the same time constructs a time series database, connects the time series database with the disease assessment and early warning platform, and stores the collected disease course data in the time series database; The feature extraction module extracts the disease course data from the time series database for preprocessing to obtain a pathology data set, and performs feature extraction on the pathology data set to obtain a pathology feature vector; The blood sugar fluctuation and fundus disease analysis module constructs a disease course algorithm model and a fundus disease risk algorithm model, inputs the acquired pathological feature vector into the disease course algorithm model and the fundus disease risk algorithm model, and respectively calculates and outputs the disease course influencing factor Rdt, the risk contribution value Rwx of retinal microvascular damage, and the blood sugar fluctuation risk contribution value Rbv; The comprehensive assessment and early warning module comprehensively calculates the output results of the disease course algorithm model and the fundus disease risk algorithm model to output a comprehensive fundus lesion risk factor Rrisk, and then preliminarily determines the risk of diabetic fundus lesions based on the output result of the comprehensive fundus lesion risk factor Rrisk, and divides the fundus lesions of diabetic patients into different risk levels based on the preliminary determination result; The risk levels include primary risk, secondary risk and tertiary risk; The depth assessment and early warning module triggers a depth analysis when it identifies a diabetic patient as being at level 2 or level 3 risk. The depth analysis performs a depth calculation based on the comprehensive fundus lesion risk factor Rrisk to output a depth risk factor Rdeep, and performs a depth determination of the extent of fundus lesions based on the output result of the depth risk factor Rdeep, and generates relevant early warning measures based on the depth determination result.

2. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 1 is characterized by: The disease course extraction module includes a disease course extraction unit and a data storage unit; The disease course extraction unit connects the disease assessment and early warning platform to the hospital's electronic health record (EHR) by setting an API application program interface to extract the disease course data of the diabetic patient; The disease course data includes patient information, diabetes duration Dt, blood sugar monitoring amount B and fundus examination images; The data storage unit constructs a time series database and sets a write port and a write port. The medical course data is stored in the time series database according to the extraction timestamp through the write port. At the same time, a unique character ID is generated based on the patient information through encoding technology. Each unique character ID corresponds to the medical course data of the relevant diabetic patient.

3. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 2 is characterized by: The feature extraction module includes a preprocessing unit and a feature extraction unit; The preprocessing unit extracts the medical course data through the write port and preprocesses the medical course data, wherein the preprocessing method includes image preprocessing and data preprocessing; The image preprocessing uses AI image processing technology on the fundus examination images in the course of disease data to intelligently identify the number of retinal microbleeds Nmic, the area of ​​retinal lesions Aretina, the area of ​​retinal edema Aedema and the total area of ​​retina Azqymj in the fundus examination images; The data preprocessing is performed by cleaning and normalizing the diabetes duration Dt and blood glucose monitoring value B in the course data, combining the obtained retinal microbleed number Nmic, retinal lesion area Aretina, retinal edema area Aedema and total retinal area Azqymj, normalizing all data in the course data to the range of [0,1], removing outliers, missing data and eliminating dimension effects; The number of retinal microbleeds Nmic, the area of ​​retinal lesions Aretina, the area of ​​retinal edema Aedema, the total area of ​​retina Azqymj, the duration of diabetes Dt and the blood glucose monitoring amount B obtained after preprocessing are summarized to generate a pathological data set; The feature extraction unit obtains a pathology feature vector by performing feature extraction based on the acquired pathology data set; The pathological feature vector includes diabetes duration Dt, blood sugar monitoring amount B, blood sugar fluctuation index Bv, retinal microvascular damage index Rmi and retinal edema index Rsw; The blood sugar fluctuation index Bv is extracted by calculating the standard deviation based on the blood sugar monitoring amount B detected each time. The specific extraction formula is: , where n represents the total number of measurements, Bi represents the i-th blood glucose measurement value, Indicates the average blood sugar measurement; The retinal microvascular injury index Rmi is calculated and extracted based on the number of retinal microbleeds Nmic and the area of ​​retinal lesions Aretina. The specific extraction formula is: ; The retinal edema index Rsw is calculated and extracted based on the retinal edema area Aedema and the total retinal area Azqymj. The specific extraction formula is: .

4. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 3 is characterized by: The blood sugar fluctuation and fundus disease analysis module includes a disease course and pathological changes analysis unit, a microvascular damage analysis unit and a blood sugar analysis unit; The disease course and lesion analysis unit extracts the duration of diabetes Dt in the pathological feature vector, inputs it into the disease course algorithm model, calculates and outputs the disease course influence factor Rdt, and analyzes the influence of the disease course on the fundus lesions; The disease course influencing factor Rdt is calculated and output by the following disease course algorithm model; ； In the formula, tanh represents the hyperbolic tangent function, represents the parameter that regulates the effect of the disease course on fundus lesions, e represents the exponential function, represents the exponential decay term, represents the sensitivity of the disease course to changes in risk, represents the exponential decay term adjustment coefficient, Speed ​​control coefficient representing the exponential decay term.

5. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 4 is characterized by: The microvascular damage analysis unit constructs an eye fundus disease risk algorithm model, wherein the eye fundus disease risk algorithm model includes a microvascular damage risk algorithm model and a blood sugar fluctuation risk algorithm model; At the same time, the retinal microvascular injury index Rmi in the pathological feature vector is extracted and input into the microvascular injury risk algorithm model to calculate and output the risk contribution value Rwx of retinal microvascular injury to quantify the degree of retinal microvascular injury; The risk contribution value Rwx of retinal microvascular damage is calculated and output by the following microvascular damage risk algorithm model; ； In the formula, ln represents the natural logarithm, represents the logarithmic transformation term, Represents the logarithmic transformation adjustment coefficient, Rmi 3 represents the cubic term of the retinal microvascular damage index, Represents the cubic adjustment coefficient.

6. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 5 is characterized by: The blood sugar analysis unit extracts the blood sugar fluctuation index Bv from the pathological feature vector, inputs it into the blood sugar fluctuation risk algorithm model, calculates and outputs the blood sugar fluctuation risk contribution value Rbv, and quantifies the impact of blood sugar fluctuation on fundus lesions; The blood sugar fluctuation risk contribution value Rbv is calculated and output by the following blood sugar fluctuation risk algorithm model; ； In the formula, e represents the exponential function, The index term representing the blood sugar fluctuation index, represents the increase control coefficient of the exponential term of the blood sugar fluctuation index, sin represents the sine function, Indicates the sensitivity control coefficient of blood sugar fluctuation index, Represents the periodic control coefficient of blood sugar fluctuation index.

7. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 3 is characterized by: The comprehensive assessment and early warning module includes a comprehensive summary unit and a grade classification assessment unit; The comprehensive summary unit performs comprehensive calculation and output of the comprehensive fundus lesion risk factor Rrisk by combining the disease course influencing factor Rdt, the risk contribution value Rwx of retinal microvascular damage and the risk contribution value Rbv of blood sugar fluctuation output by the blood sugar fluctuation and fundus disease analysis module with the retinal edema index Rsw, and performs comprehensive quantification of the fundus lesion risk; The comprehensive fundus lesion risk factor Rrisk is calculated and output by the following algorithm formula; ； In the formula, represents the retinal edema index adjustment factor.

8. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 7 is characterized by: The grading evaluation unit sets the fundus lesion risk interval [0.3, 0.6] of the diabetic patient based on the fundus lesion risk by the user, and then makes a preliminary determination of the fundus lesion risk of the current diabetic patient based on the output result of the comprehensive fundus lesion risk factor Rrisk, and grading the diabetic patient. The specific determination contents are as follows; When the comprehensive fundus lesion risk factor Rrisk is less than 0.3, it is marked as a first-level risk; When 0.3≤comprehensive fundus lesion risk factor Rrisk<0.6, it is marked as secondary risk; When the comprehensive fundus lesion risk factor Rrisk ≥ 0.6, it is marked as level 3 risk; Among them, the lesion risk level is from low to high, level one risk < level two risk < level three risk.

9. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 8, characterized in that: The depth assessment warning module includes a depth analysis unit and a depth assessment unit; The in-depth analysis unit triggers in-depth analysis after preliminary judgment of secondary risk and tertiary risk. The in-depth analysis extracts the average glycated hemoglobin value HbA1c of diabetic patients within 3 months from the electronic health record EHR to calculate the blood sugar control index Dci. The specific algorithm formula is: , where HbA1c represents the average glycated hemoglobin value, 6.5 represents the diagnostic standard for diabetes, and 8.5 represents abnormal blood sugar control; Then, based on the combined calculation of the blood sugar control index Dci and the preliminary determined comprehensive fundus lesion risk factor Rrisk, the deep risk factor Rdeep is output to conduct an in-depth analysis of the fundus lesion risk of diabetic patients; The depth risk factor Rdeep is calculated and output by the following algorithm formula; ； In the formula, represents the adjustment factor of the glycemic control index, Represents the second adjustment coefficient of retinal edema index.

10. The intelligent assessment and early warning system for emergency internal medicine conditions according to claim 9, characterized in that: The depth assessment unit extracts historical data of diabetes from the hospital's electronic health record EHR through an API application program interface, combines clinical examination results and treatment results, and allows the user to set the lesion severity interval [0.4, 0.7], and then outputs the obtained depth risk factor Rdeep to deeply determine the risk of fundus lesions in diabetic patients. The specific determination content is as follows; When the depth risk factor Rdeep is less than 0.4, regular inspection is prompted; When 0.4≤deep risk factor Rdeep<0.7, the first warning message is generated to prompt medical staff to adjust the medication of diabetic patients; When the deep risk factor Rdeep≥0.7, a second warning message is generated to prompt medical staff to intervene in treatment.

Images