Eye health risk assessment method and system based on big data analysis

Abstract

The invention discloses an eye health risk assessment method and system based on big data analysis, and relates to the technical field of eye health risk assessment, multi-source personalized parameters of a user are analyzed based on induction factors of various existing eye diseases, and the eye diseases are divided into low-probability diseases and high-probability diseases; for the high-probability diseases and the low-probability diseases, evaluation is carried out based on monitoring parameters, and whether the occurrence risk of the eye diseases exists or not is judged according to evaluation results; the assessment results of the high-probability diseases and the low-probability diseases are collected, an eye risk assessment report is generated, and the eye risk assessment report comprises the eye diseases, with risks, of the user and corresponding prevention suggestions. Multi-dimensional and personalized eye health risk assessment based on big data analysis is realized, the accuracy and efficiency of assessment are improved, and powerful support is provided for early discovery and prevention of eye diseases.

Description

Technical Field

[0001] The present invention relates to the technical field of eye health risk assessment, and particularly to an eye health risk assessment method and system based on big data analysis. Background Art

[0002] With the changes in modern lifestyles, such as long-term use of electronic devices and increased environmental pollution, the incidence of eye diseases has shown an upward trend, and the need for eye health risk assessment has become increasingly urgent. The advent of the big data era has made it possible to obtain and integrate multi-source data, providing rich data support for eye health risk assessment, such as medical records, environmental monitoring data, wearable device data, etc.

[0003] However, the existing technologies only rely on single or a few data sources for eye health assessment, ignoring the impact of multi-source data such as living environment and work and rest on eye health, resulting in insufficient and inaccurate assessment; traditional methods mostly adopt general assessment criteria and do not fully consider individual differences such as age, family medical history, and living habits, making it difficult to meet the personalized needs of different users; at the same time, the existing technologies do not pay enough attention to the monitoring of low-probability diseases or lack effective monitoring methods, and cannot detect potential risks in time; moreover, some existing assessment models do not fully combine big data analysis and machine learning technologies, resulting in limited accuracy and reliability of assessment results.

[0004] Therefore, in view of the above problems, there is an urgent need for an eye health risk assessment method and system based on big data analysis. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technologies, the present invention provides an eye health risk assessment method and system based on big data analysis, which solves the problems of single traditional eye health assessment method, lack of personalization, and difficulty in accurately assessing the eye disease risks of different individuals; at the same time, it overcomes the problems of inability to effectively utilize multi-source data, insufficient monitoring of low-probability diseases, and inability to provide comprehensive prevention suggestions for users in time.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: An eye health risk assessment method based on big data analysis, comprising the following steps: obtaining multi-source personalized parameters of a user based on multiple data sources, where the multi-source personalized parameters of the user include living environment parameters, basic parameters, and work and rest parameters; analyzing the multi-source personalized parameters of the user based on the inducing factors of existing various eye diseases, outputting the occurrence probabilities of various eye diseases of the user, and then classifying the eye diseases into low-probability diseases and high-probability diseases according to the occurrence probabilities of various eye diseases of the user; for the high-probability diseases, determining the monitoring parameters of the high-probability diseases, and then inputting the obtained monitoring parameters into the evaluation model of the corresponding eye disease, and analyzing the output result to determine whether there is a risk of occurrence of the eye disease; for the low-probability diseases, determining the monitoring parameters and evaluation periods of the low-probability diseases, and respectively performing two rounds of evaluations based on the monitoring parameters at each time node within each evaluation period of each low-probability disease, and then respectively determining whether there is a risk of occurrence of each low-probability disease based on the results of the two rounds of evaluations; collecting the evaluation results of the high-probability diseases and the low-probability diseases, and generating an eye risk assessment report, where the eye risk assessment report includes the eye diseases for which the user has a risk of occurrence and corresponding prevention suggestions.

[0007] Further, the specific analysis of analyzing the multi-source personalized parameters of the user based on the inducing factors of existing various eye diseases and outputting the occurrence probabilities of various eye diseases of the user is as follows: The inducing factors are the multi-source personalized parameters, that is, the inducing factors include living environment parameters, basic parameters, and work and rest parameters; the living environment parameters include the ultraviolet intensity of the place of residence, the concentration of air pollutants, the altitude of the place of residence, and the relative humidity of the place of residence, the basic parameters include age, gender, basic eye condition, and family medical history, and the work and rest parameters specifically include sleep duration, daily intake of eye health-related nutritional elements, and eye use duration; collecting the diagnosis cases of existing various eye diseases, extracting the inducing factors of various eye diseases, and then obtaining the preset indicators of various eye diseases corresponding to various inducing factors; respectively taking the average of the absolute differences between the multi-source personalized parameters of the user and the preset indicators of various eye diseases corresponding to various inducing factors, and then marking the reciprocal of the average of the absolute differences as the occurrence probability of various eye diseases of the user.

[0008] Further, the specific analysis of classifying the eye diseases into low-probability diseases and high-probability diseases according to the occurrence probabilities of various eye diseases of the user is as follows: Obtaining a probability reference threshold based on the diagnosis cases of existing various eye diseases, and then respectively comparing the occurrence probabilities of various eye diseases of the user with the probability reference threshold; marking the eye diseases with occurrence probabilities greater than or equal to the probability reference threshold as the high-probability diseases of the user, and marking the eye diseases with occurrence probabilities less than the probability reference threshold as the low-probability diseases of the user.

[0009] Further, for the high - probability diseases, determine the monitoring parameters of the high - probability diseases, and then input the obtained monitoring parameters into the evaluation model corresponding to the eye diseases, and analyze the output results to judge whether there is a risk of occurrence of the eye diseases. The specific analysis is as follows: Set the monitoring parameters of the k - th high - probability disease of the user as: x k1 , x k2 , x k3 ,..., x kn , where k n represents the number of monitoring parameters of the k - th high - probability disease; Obtain the specific monitoring parameters of this high - probability disease of the user, and then input the specific monitoring parameters of this high - probability disease of the user into the evaluation model corresponding to the eye disease. The specific formula of the evaluation model is expressed as:

[0010] In the formula, P(y k = 1|x k ) represents the probability that the user has the k - th high - probability disease, and β k1 , β k2 , β k3 ,..., β kn represent the weight values of each monitoring parameter respectively, represents the bias term; Obtain the probability threshold T k of this evaluation model, and compare the output result of this evaluation model with the probability threshold. When P(y k = 1|x k ) > T k , it is determined that the user has a risk of occurrence of the k - th high - probability disease.

[0011] Further, for the low - probability diseases, determine the monitoring parameters and evaluation periods of the low - probability diseases, and conduct two rounds of evaluations respectively based on the monitoring parameters at each time node within each evaluation period of each low - probability disease. Then, based on the results of the two rounds of evaluations, judge whether there is a risk of occurrence of each low - probability disease. The specific analysis is as follows: Set the monitoring parameters of the h - th low - probability disease of the user as: x h1 , x h2 , x h3 ,..., x hm , where h mRepresents the number of monitoring parameters for the h-th low-probability disease; obtain the specific monitoring parameters of this low-probability disease for the user at each time node in the evaluation period, and then use the evaluation model to obtain the output results of the specific monitoring parameters of this low-probability disease for the user at each time node respectively; obtain the probability threshold of the evaluation model for this low-probability disease, compare the output results of the evaluation model at each time node with the probability threshold respectively, and obtain the comparison results of the output results at each time node. When there are more than half of the output results at the time nodes greater than the probability threshold, it is determined that the result of one round of evaluation is that there is a risk of occurrence of this low-probability disease; based on linear fitting, conduct a second-round evaluation on the specific monitoring parameters of this low-probability disease for the user at each time node in the evaluation period, and identify whether the result of the second-round evaluation is that there is a risk of occurrence of this low-probability disease; when both the result of the first-round evaluation and the result of the second-round evaluation are that there is a risk of occurrence of this low-probability disease, it is determined that the user has a risk of occurrence of the h-th low-probability disease.

[0012] Further, the specific analysis of the second-round evaluation of the specific monitoring parameters of this low-probability disease for the user at each time node in the evaluation period based on linear fitting is as follows: perform normalization processing on the monitoring parameters of the h-th low-probability disease for the user at each time node in the evaluation period, and then respectively take the sum average value of the monitoring parameters of the h-th low-probability disease for the user at each time node in the evaluation period as the unique eigenvalue of the h-th low-probability disease for the user at each time node in the evaluation period; set that the unique eigenvalue of the h-th low-probability disease for the user at each time node in the evaluation period satisfies a linear relationship with time, and the specific linear relationship expression is: η t = ε 0 + ε 1 *t + σ, where η t represents the unique eigenvalue, t represents time, ε 0 , ε 1 represent linear coefficients, and σ represents the error term; use the least squares method to estimate and obtain ε 1 , if ε 1 > 0, then it is determined that the unique eigenvalue of the h-th low-probability disease of the user has an increasing trend in time, and further it is determined that the result of the second-round evaluation is that there is a risk of occurrence of this low-probability disease.

[0013] An eye health risk assessment system based on big data analysis, which applies the above-mentioned eye health risk assessment method based on big data analysis, includes: a data acquisition module for obtaining multi-source personalized parameters of a user based on multiple data sources, where the multi-source personalized parameters of the user include living environment parameters, basic parameters, and work and rest parameters; an eye disease prediction module for analyzing the multi-source personalized parameters of the user based on the inducing factors of existing various eye diseases, outputting the occurrence probabilities of various eye diseases of the user, and then classifying the eye diseases into low-probability diseases and high-probability diseases according to the occurrence probabilities of various eye diseases of the user; a high-probability disease assessment module for determining the monitoring parameters of the high-probability diseases for the high-probability diseases, then inputting the obtained monitoring parameters into the evaluation model corresponding to the eye diseases, and analyzing the output results to determine whether there is a risk of occurrence of the eye diseases; a low-probability disease assessment module for determining the monitoring parameters and evaluation periods of the low-probability diseases for the low-probability diseases, and performing two rounds of evaluations respectively based on the monitoring parameters at each time node within each low-probability disease evaluation period, and then determining whether there is a risk of occurrence of each low-probability disease respectively based on the results of the two rounds of evaluations; a report generation module for collecting the evaluation results of the high-probability diseases and the low-probability diseases and generating an eye risk assessment report, where the eye risk assessment report includes the eye diseases for which the user has a risk of occurrence and corresponding prevention suggestions.

[0014] The present invention has the following beneficial effects:

[0015] The eye health risk assessment method and system based on big data analysis integrate multi-source personalized parameters, such as living environment, basic, and work and rest parameters, comprehensively and accurately analyze the occurrence probabilities of users' eye diseases, realize personalized assessment, and meet the health risk assessment needs of different users; distinguish high-probability and low-probability diseases, conduct targeted monitoring of high-probability diseases, and conduct two rounds of evaluations on low-probability diseases, improve the disease screening efficiency, and timely discover potential eye health risks; through the evaluation model and the two-round evaluation mechanism, provide scientific decision-making basis for doctors and users, help formulate reasonable prevention and treatment plans, and reduce the risk of occurrence of eye diseases; generate an eye risk assessment report containing prevention suggestions, help users understand their own eye health conditions, and improve users' attention and prevention awareness of eye health. Brief Description of the Drawings

[0016] Figure 1 It is a flowchart of an eye health risk assessment method based on big data analysis of the present invention.

[0017] Figure 2 It is a structural diagram of an eye health risk assessment system based on big data analysis of the present invention. Detailed Embodiments

[0018] The embodiments of the present application implement multi-dimensional and personalized eye health risk assessment based on big data analysis through an eye health risk assessment method and system, improve the accuracy and efficiency of the assessment, and provide strong support for the early detection and prevention of eye diseases.

[0019] The general idea of the embodiments of the present application is as follows: First, obtain the user's living environment parameters, basic parameters, and work and rest parameters from multiple data sources to comprehensively collect information related to the user's eye health; analyze the multi-source personalized parameters based on the inducing factors of eye diseases, output the probability of disease occurrence, and distinguish high-probability and low-probability diseases; for high-probability diseases, determine the monitoring parameters and use the evaluation model to judge the occurrence risk, for low-probability diseases, set the monitoring parameters and evaluation period, and judge the risk through two rounds of evaluation; finally, collect the evaluation results, generate an eye risk assessment report including preventive suggestions, and provide a comprehensive eye health risk assessment service for users.

[0020] Please refer to Figure 1 , the embodiments of the present invention provide a technical solution: An eye health risk assessment method based on big data analysis, including the following steps: obtaining multi-source personalized parameters of a user based on multiple data sources, where the multi-source personalized parameters of the user include living environment parameters, basic parameters, and work and rest parameters; analyzing the multi-source personalized parameters of the user based on the inducing factors of existing various eye diseases, outputting the occurrence probability of each eye disease of the user, and then classifying the eye diseases into low-probability diseases and high-probability diseases according to the occurrence probability of each eye disease of the user; for high-probability diseases, determining the monitoring parameters of the high-probability diseases, and then inputting the obtained monitoring parameters into the evaluation model corresponding to the eye disease, and analyzing the output result to judge whether there is a risk of occurrence of the eye disease; for low-probability diseases, determining the monitoring parameters and evaluation period of the low-probability diseases, and respectively performing two rounds of evaluation based on the monitoring parameters at each time node within each evaluation period of each low-probability disease, and then respectively judging whether there is a risk of occurrence of each low-probability disease based on the results of the two rounds of evaluation; collecting the evaluation results of high-probability diseases and low-probability diseases, and generating an eye risk assessment report, where the eye risk assessment report includes the eye diseases for which the user has a risk of occurrence and corresponding preventive suggestions.

[0021] Specifically, based on the inducing factors of various existing eye diseases, the multi-source personalized parameters of the user are analyzed, and the specific analysis of the occurrence probability of each eye disease of the user is as follows: The inducing factors are the multi-source personalized parameters, that is, the inducing factors include living environment parameters, basic parameters, and work and rest parameters; the living environment parameters include the ultraviolet intensity of the place of residence, the concentration of air pollutants, the altitude of the place of residence, and the relative humidity of the place of residence, the basic parameters include age, gender, basic eye condition, and family medical history, and the work and rest parameters specifically include sleep duration, daily intake of eye health-related nutritional elements, and eye use duration; collect the diagnosis cases of various existing eye diseases, extract the inducing factors of each eye disease, and then obtain the preset indicators of each eye disease corresponding to each inducing factor; respectively take the average of the absolute differences between the multi-source personalized parameters of the user and the preset indicators of each eye disease corresponding to each inducing factor, and then mark the reciprocal of the average of the absolute differences as the occurrence probability of each eye disease of the user.

[0022] In this implementation plan, the specific analysis of classifying eye diseases into low-probability diseases and high-probability diseases based on the occurrence probability of each eye disease of the user is as follows: Obtain the probability reference threshold based on the diagnosis cases of various existing eye diseases, and then compare the occurrence probability of each eye disease of the user with the probability reference threshold respectively; mark the eye diseases with an occurrence probability greater than or equal to the probability reference threshold as the high-probability diseases of the user, and mark the eye diseases with an occurrence probability less than the probability reference threshold as the low-probability diseases of the user.

[0023] Existing eye diseases include but are not limited to: myopia, hyperopia, astigmatism, cataract, glaucoma, macular degeneration, dry eye, conjunctivitis.

[0024] The ultraviolet intensity of the place of residence represents the ultraviolet radiation intensity received in the user's residential area. Excessive ultraviolet rays may increase the risk of eye diseases such as cataract and macular degeneration. It is obtained through data released by the local meteorological department and professional ultraviolet monitoring equipment (such as ultraviolet sensors), and is represented by the ultraviolet index (UVI), generally ranging from 0 to 11+, and the larger the value, the higher the ultraviolet intensity.

[0025] The concentration of air pollutants represents the content of various pollutants (such as PM2.5, PM10, sulfur dioxide, nitrogen oxides, etc.) in the air of the living environment. Air pollution may irritate the eyes and cause eye problems such as inflammation. It is obtained through the air quality report released by the local environmental protection department and a household air quality monitor, and the mass of pollutants per cubic meter of air (such as μg / m 3 ) represents the concentration of different pollutants.

[0026] The altitude of the place of residence represents the height of the user's place of residence relative to sea level. The low-oxygen environment in high-altitude areas may affect the eyes. It can be identified by querying through map software (such as Baidu Map, Amap) or using professional altitude measurement equipment (such as an altimeter), and is quantified in meters (m).

[0027] The relative humidity of the place of residence represents the percentage of the actual water vapor pressure in the air to the saturated water vapor pressure at the same temperature. Too high or too low humidity may affect eye health. It can be identified through data from local meteorological departments or household hygrometers, and is expressed as a percentage (%).

[0028] The age is the user's actual age. The probability of suffering from eye diseases varies among different age groups. It is provided by the user himself or obtained from valid documents such as an ID card based on the user's authorization, and is in years.

[0029] The gender is the user's biological gender. The incidence of certain eye diseases varies among different genders. It is provided by the user himself, and is coded with 0 representing female and 1 representing male.

[0030] The basic eye condition represents the current health status of the user's eyes, such as whether there is myopia, hyperopia, astigmatism, eye inflammation, etc. It is identified through examinations and diagnoses by ophthalmologists, such as vision tests, fundus examinations, intraocular pressure measurements, etc., and is in a hierarchical coding method, such as 0 representing no disease, 1 representing mild disease, 2 representing moderate disease, 3 representing severe disease, etc.

[0031] The family medical history represents whether there are relatives in the user's family who have eye diseases. Genetic factors may increase the risk of getting the disease. It is obtained by the user asking family members and providing relevant information, and 0 represents that there are no patients with relevant diseases in the family, and 1 represents that there are patients with relevant diseases.

[0032] The sleep duration represents the user's daily sleep time. Lack of sleep may affect the normal metabolism and repair function of the eyes. The user monitors it by recording or using smart wearable devices (such as fitness trackers, smartwatches), and is in hours.

[0033] The daily intake of eye health-related nutrient elements represents the amount of nutrient elements beneficial to eye health (such as vitamin A, C, E, lutein, etc.) that the user intakes every day. Lack of nutrient elements may affect eye health. It is estimated through dietary surveys and nutritional component analyses, or using professional nutritional monitoring equipment (relatively few at present), and the intakes of different nutrient elements are expressed in milligrams (mg).

[0034] The eye use duration represents the time the user spends using the eyes for close work, reading, looking at electronic screens, etc. every day. Prolonged eye use is likely to cause eye fatigue and myopia and other problems. It is recorded by the user himself or monitored using tools such as an eye tracker, mobile applications, etc., and is in hours.

[0035] The preset indicators for various eye diseases corresponding to various inducing factors are obtained through the analysis of a large number of existing eye disease diagnosis cases. The typical values or value ranges of each eye disease for each inducing factor are used as reference standards for judging the probability of a user getting the disease. Specifically, by collecting multi-source personalized parameter data of a large number of diagnosed eye disease patients and conducting statistical analysis, such as calculating the mean, median, standard deviation, etc., the typical indicators for each inducing factor are determined.

[0036] The probability reference threshold is the boundary value used to distinguish between high-probability and low-probability of a user getting an eye disease. Classify and statistically analyze the existing eye disease diagnosis cases according to multi-source personalized parameters, calculate the disease occurrence probability corresponding to each case, and then determine a suitable threshold according to the actual situation (such as clinical experience, disease prevention and control requirements, etc.) to make the division of high-probability and low-probability meaningful. Machine learning methods such as cross-validation can also be used to continuously adjust the threshold to find a threshold that can optimize the classification effect (such as optimal accuracy, recall rate, etc.).

[0037] Taking into account multi-source personalized parameters such as living environment, basic conditions, and work and rest, it is possible to provide each user with an assessment of the probability of developing eye diseases that conforms to their own characteristics, which is more accurate and targeted than the general assessment method; determining the preset indicators and probability reference threshold based on a large number of existing eye disease diagnosis cases makes the assessment process supported by actual data, improving the reliability and scientific nature of the assessment results; predicting in advance the probability of a user getting different eye diseases and distinguishing high-probability and low-probability diseases helps users and doctors take preventive and intervention measures in a timely manner to achieve early prevention and control of eye diseases; calculating the occurrence probability through simple average of absolute differences and reciprocals, and comparing with the probability reference threshold to divide the disease probability level, the method is intuitive and convenient for practical application and promotion.

[0038] Specifically, for high-probability diseases, determine the monitoring parameters of high-probability diseases, and then input the obtained monitoring parameters into the evaluation model of the corresponding eye disease. The specific analysis of judging whether there is a risk of occurrence of the eye disease by analyzing the output result is as follows: Set the monitoring parameters of the kth high-probability disease of the user to be: x k1 , x k2 , x k3 ,..., x kn , where kn represents the number of monitoring parameters of the kth high-probability disease; obtain the specific monitoring parameters of the user for this high-probability disease, and then input the specific monitoring parameters of the user for this high-probability disease into the evaluation model of the corresponding eye disease. The specific formula of the evaluation model is expressed as:

[0039] In the formula, P(y k =1|xk ) represents the probability that the user has the k-th high-probability disease, β k1 , β k2 , β k3 ,..., β kn respectively represent the weight values of each monitoring parameter, represents the bias term; obtain the probability threshold T of this evaluation model k , and compare the output result of this evaluation model with the probability threshold. When P(y k = 1|x k ) > T k , it is determined that the user has the risk of occurrence of the k-th high-probability disease.

[0040] In this implementation scheme, the specific monitoring parameters are determined by specific eye diseases. For example, for glaucoma, the monitoring parameters can be set as intraocular pressure, cup-to-disc ratio, visual field, and retinal nerve fiber layer thickness. The specific analysis is as follows: An increase in intraocular pressure is one of the important risk factors for glaucoma, and the normal intraocular pressure range is generally 10 - 21 mmHg; the cup-to-disc ratio refers to the ratio of the diameter of the optic disc depression (cup) to the overall diameter of the optic disc (disc). The cup-to-disc ratio of glaucoma patients usually increases. Generally, when the cup-to-disc ratio is greater than 0.6 or the difference in cup-to-disc ratio between the two eyes is greater than 0.2, glaucoma needs to be vigilant; glaucoma can cause visual field defects, and early visual field changes can be detected through visual field examinations. Common types of visual field defects include paracentral scotoma, nasal step, arcuate scotoma, etc.; the retinal nerve fiber layer of glaucoma patients will gradually become thinner, and the RNFLT is measured through technologies such as optical coherence tomography (OCT).

[0041] The evaluation model is obtained by training with existing historical data of eye diseases. The specific acquisition and analysis are as follows: Collect historical data of a large number of patients with and without the eye disease, including monitoring parameters of each patient (such as intraocular pressure, cup-to-disc ratio, etc.) and the final disease diagnosis result (diseased or not diseased). These data can come from hospital medical records, ophthalmology examination reports, etc.; Clean the collected historical data, remove missing values, outliers, etc., and standardize the data to convert monitoring parameters with different ranges and units into a unified scale for subsequent model training; Select a suitable evaluation model, such as a logistic regression model. The logistic regression model has the advantages of being simple, easy to understand, and highly interpretable, and is suitable for binary classification problems; Divide the preprocessed historical data into a training set and a test set, generally in a ratio of 7:3 or 8:2. Use the training set data to train the selected evaluation model. By continuously adjusting the parameters of the model (such as weight values and bias terms), the model can accurately predict whether a patient has the eye disease; Use the test set data to evaluate the trained model, calculate evaluation indicators such as the accuracy rate, recall rate, and F1 value of the model, and evaluate the performance of the model. If the model performance is not ideal, the model parameters can be adjusted or the model can be replaced, and training and evaluation can be carried out again.

[0042] In the evaluation model, the weight values and bias terms are usually obtained using maximum likelihood estimation and optimization algorithms (such as gradient descent method). The way to obtain the probability threshold of the evaluation model is as follows: Doctors determine a suitable probability threshold based on long-term clinical experience and the characteristics of the disease. The receiver operating characteristic (ROC) curve can also be used to determine the optimal probability threshold.

[0043] For diseases with high probabilities, through specialized monitoring and evaluation, combined with specific monitoring parameters and evaluation models, it is possible to more accurately judge whether a user has the risk of developing the disease, avoiding the errors caused by using a unified evaluation method for all diseases; Considering the specific monitoring parameters of different users, personalized disease risk assessment is realized. Since the monitoring parameters of each user are different, the calculated probability of getting the disease is also different, thus providing a risk judgment that is more in line with the actual situation of the user; After comparing the output result of the evaluation model with the probability threshold, a conclusion can be quickly obtained, providing efficient decision-making support for doctors and users. Doctors can take corresponding preventive or treatment measures in a timely manner according to the judgment result, and users can also understand their own health status; By setting different monitoring parameters and evaluation models for different high-probability diseases, it has good scalability and flexibility. With the in-depth development of medical research and the emergence of new monitoring technologies, the monitoring parameters and models can be continuously optimized.

[0044] Specifically, for low - probability diseases, determine the monitoring parameters and evaluation periods of low - probability diseases, and conduct two rounds of evaluations based on the monitoring parameters at each time node within each evaluation period of each low - probability disease. Then, based on the results of the two rounds of evaluations, respectively determine whether there is a risk of occurrence of each low - probability disease. The specific analysis is as follows: Set the monitoring parameters of the user's h - th low - probability disease as: x h1 , x h2 , x h3 ,..., x hm , where hm represents the number of monitoring parameters of the h - th low - probability disease; Obtain the specific monitoring parameters of the user for this low - probability disease at each time node of the evaluation period, and then use the evaluation model to obtain the output results of the specific monitoring parameters of the user for this low - probability disease at each time node respectively; Obtain the probability threshold of the evaluation model for this low - probability disease, compare the output results of the evaluation model at each time node with the probability threshold respectively, and obtain the comparison results of the output results at each time node. When there are more than half of the output results at the time nodes greater than the probability threshold, determine that the result of one - round evaluation is that there is a risk of occurrence of this low - probability disease; Conduct a second - round evaluation on the specific monitoring parameters of the user for this low - probability disease at each time node of the evaluation period based on linear fitting, and identify whether the result of the second - round evaluation is that there is a risk of occurrence of this low - probability disease; When both the result of the one - round evaluation and the result of the second - round evaluation are that there is a risk of occurrence of this low - probability disease, determine that the user has a risk of occurrence of the h - th low - probability disease.

[0045] In this implementation plan, the specific analysis of conducting a second - round evaluation on the specific monitoring parameters of the user for this low - probability disease at each time node of the evaluation period based on linear fitting is as follows: Normalize the monitoring parameters of the user's h - th low - probability disease at each time node of the evaluation period, and then respectively take the sum average value of the monitoring parameters of the user's h - th low - probability disease at each time node of the evaluation period as the unique eigenvalue of the user's h - th low - probability disease at each time node of the evaluation period; Set that the unique eigenvalue of the user's h - th low - probability disease at each time node of the evaluation period satisfies a linear relationship with time. The specific linear relationship expression is: η t = ε 0 + ε 1 *t + σ, where η t represents the unique eigenvalue, t represents time, ε 0 , ε 1 represent the linear coefficients, and σ represents the error term; Use the least - squares method to estimate ε 1 and σ. If ε 1 > 0, then determine that the unique eigenvalue of the user's h - th low - probability disease has an increasing trend in time, and then determine that the result of the second - round evaluation is that there is a risk of occurrence of this low - probability disease.

[0046] The evaluation model is obtained through training with historical data. The specific training process is as follows: Collect historical data of a large number of patients with and without this low-probability eye disease, covering the monitoring parameters (such as intraocular pressure, visual acuity, etc.) of each patient at different time nodes and the final disease diagnosis results (diseased or not diseased). These data can be obtained from channels such as the hospital's medical record system and ophthalmic examination records; for monitoring parameters with missing values, methods such as interpolation (such as linear interpolation, spline interpolation) or deleting samples with missing values can be used for processing; identify and process outliers in the data, and statistical methods (such as the z-score method) or machine learning methods (such as the isolation forest algorithm) can be used; standardize the monitoring parameters with different ranges and units to make the data comparable; select a suitable evaluation model according to the data characteristics and problem types. The specific evaluation model refers to the evaluation model for high-probability diseases. Logistic regression can be used to naturally output the probability of the event occurring; divide the preprocessed historical data into a training set and a test set (usually in a ratio of 7:3 or 8:2), use the training set data to train the selected evaluation model, and continuously adjust the model parameters so that the model can accurately predict whether a patient has this low-probability disease based on the monitoring parameters; use the test set data to evaluate the trained model, calculate evaluation indicators such as accuracy, recall rate, and F1 value. If the model performance is not ideal, the model parameters can be adjusted, the model can be replaced, or the data preprocessing steps can be further optimized until satisfactory performance is achieved.

[0047] The method for obtaining the probability threshold of the evaluation model refers to the method for obtaining the probability threshold of the evaluation model for high-probability diseases. Based on long-term clinical practice and understanding of this low-probability disease, doctors determine a suitable probability threshold in combination with factors such as the severity of the disease, treatment costs, and consequences. The receiver operating characteristic (ROC) curve can also be drawn. This curve takes the true positive rate (sensitivity) as the ordinate and the false positive rate (1 - specificity) as the abscissa, and is drawn by changing the probability threshold. Calculate the area under the ROC curve (AUC), and select the probability threshold that makes the point on the ROC curve closest to the upper left corner (true positive rate of 1, false positive rate of 0) as the optimal threshold, so as to achieve a better balance between sensitivity and specificity.

[0048] The acquisition of the unique eigenvalue helps to integrate multiple monitoring parameters into one value, simplifies the representation and processing of data, and reduces the complexity of subsequent analysis; there may be a certain correlation or interference between different monitoring parameters, and the unique eigenvalue can eliminate this interference to a certain extent and more comprehensively reflect the overall health status of the patient at each time node; after integrating multiple monitoring parameters into a single unique eigenvalue, it is more convenient for linear fitting and trend analysis, and can more clearly observe the trend of disease development.

[0049] Conducting two rounds of specialized and meticulous evaluations for low-probability diseases can more accurately determine the risk of disease occurrence, avoid misjudgments that may be caused by a single evaluation method, and improve the accuracy of risk assessment; setting an evaluation period and obtaining monitoring parameters at multiple time nodes enables dynamic monitoring of low-probability diseases. Considering that the development of diseases may be a gradual process, dynamic monitoring helps to timely capture the trend of disease development; through the comprehensive judgment of two rounds of evaluations, potential risks can be detected when the disease has not yet manifested significantly in the early stage, providing more time windows for early intervention and treatment of the disease; for low-probability diseases, due to their low incidence rate, they are easily overlooked. Such a strict two-round evaluation mechanism can effectively reduce the possibility of missed diagnoses and safeguard the eye health of users.

[0050] Please refer to Figure 2 , an eye health risk assessment system based on big data analysis, applying the above-mentioned eye health risk assessment method based on big data analysis, including: a data acquisition module for obtaining multi-source personalized parameters of a user based on multiple data sources, where the multi-source personalized parameters of the user include living environment parameters, basic parameters, and work and rest parameters; an eye disease prediction module for analyzing the multi-source personalized parameters of the user based on the inducing factors of existing various eye diseases, outputting the occurrence probabilities of various eye diseases of the user, and then classifying eye diseases into low-probability diseases and high-probability diseases according to the occurrence probabilities of various eye diseases of the user; a high-probability disease assessment module for determining the monitoring parameters of high-probability diseases for high-probability diseases, then inputting the obtained monitoring parameters into the evaluation model of the corresponding eye disease, and analyzing the output result to judge whether there is a risk of occurrence of the eye disease; a low-probability disease assessment module for determining the monitoring parameters and evaluation period of low-probability diseases for low-probability diseases, conducting two rounds of evaluations respectively based on the monitoring parameters at each time node within each low-probability disease evaluation period, and then respectively judging whether there is a risk of occurrence of each low-probability disease based on the results of the two rounds of evaluations; a report generation module for collecting the evaluation results of high-probability diseases and low-probability diseases and generating an eye risk assessment report, where the eye risk assessment report includes the eye diseases for which the user has a risk of occurrence and corresponding prevention suggestions.

[0051] In summary, the present application has at least the following effects:

[0052] By collecting multi-source personalized parameters such as the user's living environment parameters, basic parameters, and work and rest parameters, this method can provide customized eye health risk assessments for users. The personalized assessment is more accurate than the traditional one-size-fits-all assessment and can better reflect the user's actual eye health status. For high-probability eye diseases, this method can detect potential risks at an early stage and make judgments through corresponding assessment models. Early warning helps users take timely measures to prevent or delay the development of diseases. By classifying eye diseases into low-probability and high-probability categories and adopting different monitoring and assessment strategies respectively, this method can allocate medical resources more effectively. By evaluating low-probability diseases through a two-round assessment method, the possibility of false positives and false negatives can be further reduced, improving the accuracy and efficiency of the assessment. At the same time, this method also reduces unnecessary examinations and tests, reducing the burden on users. The generated eye risk assessment report not only lists the eye diseases for which the user is at risk but also provides corresponding prevention suggestions. The comprehensive report helps users better understand their eye health status and take corresponding preventive measures.

[0053] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] The present invention is described with reference to the flowcharts and structural diagrams of methods and systems according to the embodiments of the present invention. It should be understood that each process and module combination in the flowcharts and structural diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and structures Figure 1 one module or multiple modules.

[0055] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the specified functions in Figure 1 one process or multiple processes and structures Figure 1 one module or multiple modules.

[0056] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable apparatus to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable apparatus provide steps for implementing the specified functions in one process Figure 1 one process or multiple processes and architectures Figure 1 or multiple modules.

[0057] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0058] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for eye health risk assessment based on big data analysis, characterized in that: The following steps are involved: Obtaining multi-source personalized parameters of the user based on multiple data sources, wherein the multi-source personalized parameters of the user include living environment parameters, basic parameters, and work and rest parameters; Analyze the user's multi-source personalized parameters based on the existing inducing factors of various eye diseases, output the probability of occurrence of each eye disease of the user, and then classify the eye diseases into low-probability diseases and high-probability diseases according to the probability of occurrence of each eye disease of the user; For the high-probability disease, determine the monitoring parameters of the high-probability disease, and then use the acquired monitoring parameters to input the evaluation model of the corresponding eye disease, analyze the output results, and determine whether there is a risk of occurrence of the eye disease; For the low-probability diseases, determine the monitoring parameters and evaluation cycle of the low-probability diseases, conduct two rounds of evaluation based on the monitoring parameters at each time point in each low-probability disease evaluation cycle, and then determine whether there is a risk of occurrence of each low-probability disease based on the results of the two rounds of evaluations; The assessment results of high-probability diseases and low-probability diseases are collected to generate an eye risk assessment report, which includes eye diseases that the user is at risk of and corresponding prevention recommendations.

2. According to the method for eye health risk assessment based on big data analysis according to claim 1, it is characterized in that: Based on the existing inducing factors of various eye diseases, the user's multi-source personalized parameters are analyzed, and the specific analysis of the probability of occurrence of various eye diseases of the user is output as follows: The inducing factors are multi-source personalized parameters, that is, the inducing factors include living environment parameters, basic parameters and work and rest parameters; The living environment parameters include ultraviolet intensity, air pollutant concentration, altitude and relative humidity of the place of residence; the basic parameters include age, gender, basic eye condition and family medical history; the work and rest parameters specifically include sleep duration, daily intake of eye health-related nutrients and eye use duration; Collect existing diagnosis cases of various eye diseases, extract the predisposing factors of various eye diseases, and then obtain the preset indicators of various eye diseases corresponding to various predisposing factors; The absolute difference averages of the user's multi-source personalized parameters and the preset indicators of the inducing factors corresponding to each eye disease are taken respectively, and then the reciprocal of the absolute difference average is marked as the occurrence probability of each eye disease of the user.

3. The eye health risk assessment method based on big data analysis according to claim 1, characterized in that: The specific analysis of dividing eye diseases into low-probability diseases and high-probability diseases based on the probability of occurrence of each eye disease of the user is as follows: Obtaining a probability reference threshold based on existing diagnosis cases of various eye diseases, and then comparing the probability of occurrence of each eye disease of the user with the probability reference threshold; Eye diseases with an occurrence probability greater than or equal to a probability reference threshold are marked as high-probability diseases of the user, and eye diseases with an occurrence probability less than the probability reference threshold are marked as low-probability diseases of the user.

4. The eye health risk assessment method based on big data analysis according to claim 1, characterized in that: For the high-probability disease, the monitoring parameters of the high-probability disease are determined, and then the acquired monitoring parameters are input into the evaluation model of the corresponding eye disease, and the output results are analyzed to determine whether there is a risk of occurrence of the eye disease. The specific analysis is as follows: Set the monitoring parameters of the user's kth high probability disease to be: k1 ,x k2 ,x k3 ,...,x kn , where k n represents the number of monitoring parameters for the kth high probability disease; The specific monitoring parameters of the high-probability disease of the user are obtained, and then the specific monitoring parameters of the high-probability disease of the user are input into the evaluation model of the corresponding eye disease. The specific formula of the evaluation model is expressed as: Where P(y k =1|x k ) represents the probability that the user has the kth high-probability disease, β k1 ,β k2 ,β k3 ,...,β kn Respectively represent the weight value of each monitoring parameter, represents the bias term; Get the probability threshold T of the evaluation model k , compare the output of the evaluation model with the probability threshold, when P(y k =1|x k )>T k , it is determined that the user has the risk of developing the kth high-probability disease.

5. The eye health risk assessment method based on big data analysis according to claim 4 is characterized in that: For the low-probability diseases, the monitoring parameters and evaluation cycle of the low-probability diseases are determined, and two rounds of evaluations are performed based on the monitoring parameters of each time node in each low-probability disease evaluation cycle. Then, based on the results of the two rounds of evaluations, the specific analysis of whether there is a risk of occurrence of each low-probability disease is as follows: Set the monitoring parameters of the user's hth low-probability disease to be: h1 ,x h2 ,x h3 ,...,x hm , where h m represents the number of monitoring parameters for the h-th low-probability disease; Obtain the specific monitoring parameters of the low-probability disease for the user at each time node in the evaluation period, and then use the evaluation model to obtain the output results of the specific monitoring parameters of the low-probability disease for the user at each time node; Obtain the probability threshold of the low-probability disease assessment model, compare the output results of the assessment model at each time node with the probability threshold, and obtain the comparison results of the output results at each time node. When more than half of the output results at the time nodes are greater than the probability threshold, determine that the result of one round of assessment is that there is a risk of occurrence of the low-probability disease; Based on linear fitting, a second round of evaluation is performed on the specific monitoring parameters of the low-probability disease of the user at each time point in the evaluation period to identify whether the second round of evaluation results indicate the risk of occurrence of the low-probability disease; When the results of the first round of evaluation and the second round of evaluation both indicate that there is a risk of occurrence of the low-probability disease, it is determined that the user has a risk of occurrence of the h-th low-probability disease.

6. The eye health risk assessment method based on big data analysis according to claim 5, characterized in that: The specific analysis of the second round of evaluation of the specific monitoring parameters of the low-probability disease of the user at each time node of the evaluation period based on linear fitting is as follows: Normalize the monitoring parameters of the h-th low-probability disease of the user at each time node in the evaluation period, and then take the summed average of the monitoring parameters of the h-th low-probability disease of the user at each time node in the evaluation period and mark them as the unique characteristic value of the h-th low-probability disease of the user at each time node in the evaluation period; It is assumed that the unique characteristic value of the user's h-th low-probability disease at each time node in the evaluation period satisfies the linear relationship with time. The specific linear relationship expression is: η t =ε0+ε1*t+σ, where η t represents the unique eigenvalue, t represents time, ε0 and ε1 represent linear coefficients, and σ represents the error term; ε1 is estimated using the least squares method. If ε1>0, it is determined that the unique characteristic value of the user's h-th low-probability disease has an increasing trend over time, and then the second-round evaluation results are determined to be a risk of occurrence of the low-probability disease.

7. An eye health risk assessment system based on big data analysis, applying an eye health risk assessment method based on big data analysis as described in any one of claims 1 to 6, characterized in that: include: A data acquisition module, used to obtain multi-source personalized parameters of a user based on multiple data sources, wherein the multi-source personalized parameters of the user include living environment parameters, basic parameters and work and rest parameters; The eye disease prediction module is used to analyze the user's multi-source personalized parameters based on the existing predisposing factors of various eye diseases, output the probability of occurrence of each eye disease of the user, and then classify the eye diseases into low-probability diseases and high-probability diseases according to the probability of occurrence of each eye disease of the user; A high-probability disease assessment module is used to determine the monitoring parameters of the high-probability disease for the high-probability disease, and then use the acquired monitoring parameters to input the corresponding eye disease assessment model, analyze the output results, and determine whether there is a risk of occurrence of the eye disease; A low-probability disease assessment module is used to determine the monitoring parameters and assessment cycle of the low-probability disease for the low-probability disease, perform two rounds of assessments based on the monitoring parameters at each time node in each low-probability disease assessment cycle, and then determine whether there is a risk of occurrence of each low-probability disease based on the results of the two rounds of assessments; The report generation module is used to collect the assessment results of high-probability diseases and low-probability diseases and generate an eye risk assessment report, which includes the eye diseases that the user is at risk of and the corresponding prevention suggestions.

Images