Method and system for predicting and intervening myopia progress of teenagers

Abstract

The invention relates to the technical field of ophthalmology medicine, computer vision and artificial intelligence, in particular to a teenager myopia progress prediction and intervention method and a teenager myopia progress prediction and intervention system. A deep learning model is constructed by using the data, the myopia development trend is predicted, eye using behaviors are monitored in real time through a multi-modal fusion technology, the data are deeply analyzed, and a personalized myopia intervention scheme is formulated according to a model prediction result; according to the system, the prediction accuracy is improved through multi-source data fusion, the binocular independent observation and real-time monitoring technology is adopted, eye use details are accurately captured, data driving and technical innovation are integrated, accurate prediction, effective intervention and scientific evaluation of myopia are achieved, and the system is high in practicability and high in practicability. And a comprehensive and intelligent solution is provided for myopia prevention and control.

Description

Technical Field

[0001] The present invention relates to the fields of ophthalmology, computer vision and artificial intelligence technology, and in particular to a method and system for predicting and intervening in the progression of myopia in adolescents. Background Art

[0002] In recent years, with the prevalence of electronic products and increasing academic pressure, myopia rates among adolescents have continued to rise, becoming a global public health issue. Traditional myopia prevention and control methods rely primarily on regular vision tests and passive intervention, lacking scientific prediction mechanisms and personalized intervention strategies. Existing technologies for myopia prevention and control present the following major challenges: First, there is a lack of accurate predictions of myopia development trends, making it impossible to formulate targeted prevention and control strategies; second, insufficiently accurate monitoring of eye behavior makes it difficult to detect adverse eye habits in real time; third, intervention measures are limited in scope and lack personalization and intelligence; and finally, there is a lack of scientific evaluation mechanisms for intervention effectiveness.

[0003] In existing technologies, some studies use statistical methods to predict the progression of myopia, but these methods typically only consider a single factor, such as visual acuity data or environmental factors, and lack a comprehensive analysis of multi-dimensional data. Other studies focus on monitoring eye behavior, but they often use a single perspective to collect data and are unable to accurately capture the details of eye movements. Some studies have proposed intervention plans, but lack intelligent feedback mechanisms and effect evaluation systems. Therefore, there is an urgent need for a systematic solution that can integrate multi-source data, accurately predict the progression of myopia, provide personalized interventions, and scientifically evaluate the effects. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for predicting and intervening in the progression of myopia in adolescents. By collecting multi-dimensional data, constructing a deep learning prediction model, applying multimodal fusion analysis technology and reinforcement learning intervention mechanism, accurate prediction of the development trend of myopia in adolescents and personalized intervention can be achieved, thereby effectively controlling the progression of myopia.

[0005] The present invention proposes a method for predicting and intervening in the progression of myopia in adolescents, comprising:

[0006] Acquisition steps: Acquisition of genetic data, environmental data, lifestyle data, eye behavior data, and vision data;

[0007] Prediction step: building a deep learning model based on the data to predict the development trend of myopia;

[0008] Monitoring step: using multimodal fusion technology to monitor eye behavior in real time and analyze the eye behavior data;

[0009] Intervention step: generating a myopia intervention plan based on the prediction results of the deep learning model;

[0010] Evaluation step: Evaluate the effectiveness of the intervention and determine whether a stable state has been achieved.

[0011] Preferably, the acquisition of the eye behavior data specifically includes: collecting the visual angle information and blinking frequency within 60 seconds through independent observation of both eyes at a time interval of 0.5 seconds, and generating the eye behavior data using a multimodal fusion eye behavior analysis method.

[0012] Preferably, the multimodal fusion eye behavior analysis method comprises the following steps:

[0013] The subject's eyes are used as two independent observation perspectives, and two cameras are set up to capture the subject's gaze angle;

[0014] The data collected by the two cameras are uploaded in real time and fused into an animation in chronological order to determine whether the outdoor activity situation meets the standards;

[0015] Eye movement data analysis technology based on deep learning is used to process the collected eye movement data into a gaze angle curve to determine whether the time spent using eyes at close range meets the standard.

[0016] Preferably, the standard for determining the time spent using the eyes at close range is as follows: if it is detected that the subject is less than 40 cm away from the display screen, the timer is started; if it is detected that the subject is greater than or equal to 40 cm away from the display screen, the timer is stopped; and whether the time spent using the eyes at close range meets the standard is determined based on the timing data of the timer.

[0017] Preferably, the animation includes outdoor activity animation and close-up eye animation;

[0018] The output curve of the outdoor activity animation is: using the average outdoor eye behavior data of the subject's two eyes as a reference standard, the collected gaze angle curve data of the subject is normalized, and the normalized gaze angle curve animation is output;

[0019] The output curve of the close-up eye animation is: using the gaze angle curve data of the subject before myopia as a reference standard, standardizing the collected gaze angle curve data of the subject, and outputting the standardized gaze angle curve data as a reference standard.

[0020] Preferably, the intervention step utilizes a reinforcement learning algorithm to construct the close-up eye animation, specifically:

[0021] When eye use behavior deviates from healthy standards, it will be punished; when it meets the standards, it will be rewarded.

[0022] Preferably, the myopia progression intervention program comprises:

[0023] Remind subjects to engage in outdoor activities;

[0024] Analyze the subjects' close-up eye use time data;

[0025] Based on the deep learning model, the progression of myopia in the subjects is analyzed and future intervention plans are generated.

[0026] Preferably, in the prediction step, the deep learning model is used to predict the future development trend of the subject based on the eye behavior data, close-range eye use time data, myopia development time and myopia degree data, and generate a future intervention plan; if it is predicted that the subject's future myopia degree will be greater than 600 degrees, the subject is reminded to consider surgical treatment.

[0027] Preferably, the evaluation step is specifically as follows:

[0028] Based on the subjects' eye behavior data, eye usage time data, and myopia progression data during the intervention process, the intervention effect and appropriate intervention time point are predicted;

[0029] When the degree of myopia progression decreases by more than 0.90D (i.e. 0.90 diopters), the intervention is judged to be successful and the intensity of intervention can be gradually reduced; when the degree of myopia progression decreases by a smaller amount but more than 0D, the intervention is judged to be effective and the intervention should continue; when the degree of myopia progression does not change, the intervention is judged to be ineffective and a new intervention plan needs to be formulated.

[0030] The adolescent myopia progression prediction and intervention system implementing the method comprises:

[0031] Genetic module, used to obtain genetic data that influences the development of myopia;

[0032] Environmental module, used to obtain environmental data that affects the development of myopia;

[0033] Lifestyle habit module, used to obtain data on lifestyle habits that affect the development of myopia;

[0034] Eye behavior module, used to obtain eye behavior data that affects the development of myopia;

[0035] A prediction module, which is used to build a deep learning model based on the acquired data and predict the development trend of myopia;

[0036] a control module for controlling the eye behavior module to monitor eye behavior in real time using multimodal fusion technology and analyze eye behavior data when the subject wears glasses or orthokeratology lenses;

[0037] An intervention module, configured to generate a myopia intervention plan based on the prediction results of the deep learning model;

[0038] The evaluation module is used to evaluate the intervention effect and determine whether a stable state has been reached.

[0039] The beneficial effects of this invention are: first, it improves the accuracy of myopia progression prediction through multi-source data fusion; second, it uses binocular independent observation and real-time monitoring technology to accurately capture the details of eye behavior; third, it introduces a reinforcement learning mechanism to improve the targeted intervention and user compliance; and finally, it establishes a scientific intervention effect evaluation system to achieve dynamic optimization of intervention plans. These innovations together form a closed-loop myopia prevention and control ecosystem, providing a new technical path for the prevention and control of myopia in adolescents. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a flow chart of the method for predicting and intervening in the progression of myopia in adolescents according to the present invention;

[0041] Figure 2 Schematic diagram of eye behavior data collection in the present invention;

[0042] Figure 3 This is a flow chart of the multimodal fusion eye behavior analysis method of the present invention;

[0043] Figure 4 This is a flow chart for determining the standard of near-distance eye use time in the present invention;

[0044] Figure 5 Animation generation and analysis flow chart for the present invention;

[0045] Figure 6 This is a schematic diagram of the reinforcement learning intervention mechanism of the present invention;

[0046] Figure 7 A schematic diagram of the components of the myopia progression intervention program of the present invention;

[0047] Figure 8 This is a system architecture diagram for predicting and intervening in myopia progression according to the present invention. DETAILED DESCRIPTION

[0048] Please refer to the attached Figure 1-7 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Example 1

[0050] like Figure 1 As shown, the method for predicting and intervening in the progression of myopia in adolescents provided by the present invention comprises the following steps:

[0051] Acquisition steps: acquiring genetic data, environmental data, lifestyle data, eye behavior data and vision data.

[0052] Specifically, genetic data includes family myopia history, parents' myopia degree, etc.; environmental data includes daily light intensity, indoor humidity, etc.; lifestyle habit data includes sleep duration, diet structure, etc.; eye behavior data includes close-range eye use time, outdoor activity time, etc.; vision data includes naked eye vision and best corrected vision.

[0053] Preferably, the present invention uses a combination of specialized data collection equipment and questionnaires to obtain the aforementioned data. For example, environmental data can be collected using a light sensor to record daily changes in light intensity; lifestyle data can be collected using a smart bracelet and a structured questionnaire; and vision data can be obtained using a standard eye chart.

[0054] The prediction step is to build a deep learning model based on the data to predict the development trend of myopia.

[0055] The deep learning model employed in this study combines the strengths of long short-term memory (LSTM) and convolutional neural networks (CNN). LSTM is well-suited for processing time series data, capturing temporal patterns in myopia progression; CNN excels at extracting spatial features and analyzing eye behavior patterns. The combination of these two forms a hybrid model with powerful predictive capabilities.

[0056] The mathematical expression of the model is as follows:

[0057] ,

[0058] ,

[0059] ,

[0060] in, Represents the input feature vector at time t, which contains data such as genetics, environment, and lifestyle habits; Represents the hidden state of LSTM; Representing eye behavior image data; Represents the features extracted by CNN; and are the weight matrix and bias term respectively; is the prediction output, which represents the probability distribution of different myopia progression situations.

[0061] The model training uses the back propagation algorithm, and the loss function uses the cross entropy function:

[0062] ,

[0063] in, is the true label, is the predicted value.

[0064] Preferably, in order to improve the generalization ability of the model, the present invention adopts batch normalization and dropout technology, and the dropout rate is set to 0.5. This value has been verified by experiments to effectively prevent overfitting and improve the performance of the model on new samples.

[0065] The monitoring step uses multimodal fusion technology to monitor eye behavior in real time and analyze the eye behavior data.

[0066] like Figure 2 As shown, the present invention specifically includes collecting eye behavior data by independently observing the viewing angles of both eyes, collecting gaze angle information and blinking frequency within 60 seconds at a time interval of 0.5 seconds, and using a multimodal fusion eye behavior analysis method to generate eye behavior data.

[0067] A sampling interval of 0.5 seconds is the optimal value determined through extensive experimentation. Shorter intervals generate redundant data and increase computational burden, while longer intervals may miss important eye movement information. A 60-second acquisition cycle covers most common eye movement scenarios, ensuring representative and complete data.

[0068] like Figure 3 As shown in FIG, the multimodal fusion eye behavior analysis method includes the following steps:

[0069] First, the subject's eyes are used as two independent observation perspectives, and two cameras are set up to capture the subject's gaze angle;

[0070] Secondly, the data collected by the two cameras is uploaded in real time and fused into an animation in chronological order to determine whether the outdoor activity situation meets the standards;

[0071] Finally, deep learning-based eye movement data analysis technology is used to process the collected eye movement data into a gaze angle curve to determine whether the close-range eye use time meets the standard.

[0072] The deep learning eye movement data analysis algorithm uses a temporal convolutional network, which is mathematically expressed as follows:

[0073] ,

[0074] in, represents the feature map of layer I, and are the convolution kernel and bias of this layer respectively, As the activation function, the ReLU function is used: ,* represents the convolution operation.

[0075] like Figure 4As shown, the standard for determining the time spent using the eyes at close range is as follows: if it is detected that the subject is less than 40 cm away from the display screen, the timer is started; if it is detected that the subject is greater than or equal to 40 cm away from the display screen, the timer is stopped; and whether the time spent using the eyes at close range meets the standard is determined based on the timing data of the timer.

[0076] The critical value of 40 cm is based on ophthalmological research. Prolonged use of the eyes below this distance can lead to eye fatigue and worsening myopia. This invention uses a deep learning-based object detection algorithm to achieve distance measurement with an accuracy of ±2 cm, meeting practical application requirements.

[0077] like Figure 5 As shown, the animations include outdoor activity animations and close-up eye use animations. The output curve for the outdoor activity animation is: using the average outdoor eye use behavior data of the subject's two eyes as a reference standard, the collected gaze angle curve data of the subject is normalized, and the standardized gaze angle curve animation is output. The output curve for the close-up eye use animation is: using the gaze angle curve data of the subject before myopia as a reference standard, the collected gaze angle curve data of the subject is normalized, and the standardized gaze angle curve data is output as a reference standard.

[0078] The Z-score method is used for standardization:

[0079] ,

[0080] in, is the original data, is the mean, is the standard deviation, The data are standardized.

[0081] Intervention steps

[0082] Based on the prediction results of the deep learning model, a myopia intervention plan is generated.

[0083] like Figure 6 As shown, the present invention uses a reinforcement learning algorithm to construct a close-up eye animation in the intervention step, specifically:

[0084] Penalties are imposed when eye behavior deviates from healthy standards, while rewards are given when it meets them. This is the only way to properly guide users to develop healthy eye habits and reduce unhealthy eye behaviors. Through this precise reward and punishment mechanism, the system enables users to gradually learn and proactively adopt healthy eye behaviors, effectively preventing the progression of myopia.

[0085] The reinforcement learning algorithm adopts the Q-learning method, and its update formula is:

[0086] ,

[0087] in, Indicates that the status Take action The value function of For immediate rewards (determined by whether the behavior meets health standards); is the discount factor, set to 0.9; is the learning rate, set to 0.1. These parameters are determined through a large number of experimental optimizations and can achieve a good balance between learning efficiency and stability.

[0088] like Figure 7 As shown, the myopia progression intervention program includes:

[0089] Remind subjects to engage in outdoor activities;

[0090] Analyze the subjects' close-up eye use time data;

[0091] Based on the deep learning model, the progression of myopia in the subjects is analyzed and future intervention plans are generated.

[0092] Optimally, the present invention provides personalized outdoor activity recommendations based on user characteristics, such as morning exercises and after-school outdoor games, while also recommending a reasonable duration (at least two hours per day). Recommendations for managing close-up eye use include specific instructions such as taking a five-minute break from distant viewing every 30 minutes.

[0093] In the prediction step, the present invention uses the constructed deep learning model to predict the future development trend of the subject based on eye behavior data, close-range eye use time data, myopia development time and myopia degree data, and generates a future intervention plan; if it is predicted that the subject's future myopia degree will be greater than 600 degrees, the subject is reminded to consider surgical treatment.

[0094] The critical value of 600 degrees is determined based on clinical research - patients with high myopia (≥600 degrees) have a significantly increased risk of complications such as retinal detachment and macular degeneration, and early surgical intervention may achieve better treatment results.

[0095] Evaluation step to assess the effectiveness of the intervention and determine whether a stable state has been achieved.

[0096] Specifically, the evaluation step of the present invention predicts the intervention effect and the appropriate intervention time point based on the subject's eye behavior data, eye usage time data, and myopia progression degree data during the intervention process;

[0097] When the degree of myopia progression decreases by more than 0.90D (i.e. 0.90 diopters), the intervention is judged to be successful and the intensity of intervention can be gradually reduced; when the degree of myopia progression decreases by a smaller amount but more than 0D, the intervention is judged to be effective and the intervention should continue; when the degree of myopia progression does not change, the intervention is judged to be ineffective and a new intervention plan needs to be formulated.

[0098] This 0.90D improvement threshold was determined through analysis of a large number of clinical cases and has empirical support. This degree of improvement indicates that the intervention measures have produced significant results, and it can be considered to gradually reduce the intensity of intervention. For situations where the effect is not obvious, the system will automatically adjust the intervention strategy, such as increasing the recommendation for outdoor activity time or more strictly controlling close eye use. This quantitative standard makes the evaluation more objective and scientific, and supports the dynamic adjustment and optimization of the intervention plan. For situations where the effect is not obvious, the present invention will automatically adjust the intervention strategy, such as increasing the recommendation for outdoor activity time, more strictly controlling close eye use, etc.

[0099] Example 2

[0100] like Figure 8 As shown, the present invention also provides a system for predicting and intervening in the progression of myopia in adolescents for implementing the above method, comprising:

[0101] Genetic module 1, used to obtain genetic data that affects the development of myopia;

[0102] Environmental module 2, used to obtain environmental data that influences the development of myopia;

[0103] Lifestyle Module 3, used to obtain data on lifestyle habits that affect the development of myopia;

[0104] Eye behavior module 4, used to obtain eye behavior data that affects the development of myopia;

[0105] Prediction module 5, used to build a deep learning model based on the acquired data and predict the development trend of myopia;

[0106] Control module 6, for controlling the eye behavior module 4 to monitor eye behavior in real time using multimodal fusion technology and analyze eye behavior data when the subject wears glasses or orthokeratology lenses;

[0107] Intervention module 7, used to generate a myopia intervention plan based on the prediction results of the deep learning model;

[0108] Evaluation module 8 is used to evaluate the intervention effect and determine whether a stable state has been reached.

[0109] The system architecture of the present invention adopts a modular design, and data is exchanged between modules through standard interfaces, ensuring the scalability and maintainability of the system. Preferably, the system can be deployed in the cloud and interact with users through mobile terminal applications to achieve real-time data collection, analysis and feedback.

[0110] Genetic Module 1 collects data on family myopia history and parental refractive error through questionnaires and medical records. For example, detailed information such as the degree of myopia in parents, age of onset, and whether they have high myopia is recorded. This data is valuable for predicting the risk of myopia in children.

[0111] Environmental Module 2 uses various sensors to collect environmental parameters such as light intensity and indoor humidity. Light intensity is measured in lux. Outdoor sunlight typically ranges from 10,000 to 25,000 lux, while indoor lighting is only 300 to 500 lux. This difference is a significant factor influencing the development of myopia.

[0112] Lifestyle Module 3 records users' sleep schedules, dietary preferences, and other information. Studies have shown that lack of sleep and a high-sugar diet are positively correlated with the development of myopia, and this module can capture these potential risk factors.

[0113] Eye Behavior Module 4 is the core component of the system and consists of a binocular image acquisition unit and a close-up eye time detection unit. The binocular image acquisition unit uses two high-precision cameras to track the movements of the left and right eyes respectively, while the close-up eye time detection unit monitors the distance between the user and the screen in real time and records the accumulated eye time.

[0114] Prediction Module 5 integrates the data collected by each module and runs a deep learning algorithm for analysis and prediction. This module uses transfer learning technology to quickly adapt to the characteristics of different users and improve prediction accuracy.

[0115] Control Module 6 coordinates the work of various parts of the system to ensure smooth data flow. In particular, when the user wears glasses or orthokeratology lenses, the module automatically adjusts parameters to ensure that monitoring accuracy is not affected.

[0116] Intervention Module 7 generates personalized intervention recommendations based on the prediction results, including behavioral adjustments, environmental optimization, and medical intervention. This module uses an adaptive algorithm to continuously optimize intervention strategies based on user feedback.

[0117] Evaluation Module 8 continuously monitors the effectiveness of the intervention, generates progress reports, and determines whether a stable state has been reached. If the intervention is ineffective, the system automatically adjusts the strategy to ensure the effectiveness of the intervention.

[0118] The system of the present invention supports multi-user management and is suitable for use in various scenarios such as schools, homes, and medical institutions. The system has a user-friendly interface and is easy to operate, which greatly reduces the user threshold and improves compliance.

[0119] In summary, the method and system for predicting and intervening in adolescent myopia progression provided by this invention builds a complete myopia prevention and control ecosystem through multi-source data fusion, deep learning prediction, multimodal analysis, and reinforcement learning intervention. This system can accurately predict myopia progression trends, provide personalized intervention plans, and scientifically evaluate intervention effects, providing a new technical approach for adolescent myopia prevention and control.

[0120] Example 3: School scenario application

[0121] In a school environment, the adolescent myopia progression prediction and intervention system of the present invention can be deployed in batches to achieve myopia monitoring and intervention for an entire class or grade. The system in this embodiment includes:

[0122] Central monitoring server 10, responsible for data storage, model operation and result analysis;

[0123] a classroom environment monitoring unit 11, comprising a light sensor and an air quality sensor;

[0124] Student terminal device 12, including an eye movement data acquisition device and a distance monitoring module;

[0125] Teacher management terminal 13, used to view the overall situation of the class and early warning information of individual students;

[0126] The parent feedback module 14 is used to receive students' eye use reports and provide family intervention feedback.

[0127] In this embodiment, the system workflow is as follows:

[0128] First, at the beginning of the semester, basic information is collected from all students, including vision tests, questionnaires (collecting genetic and lifestyle data), and initial eye movement data. This data is uploaded to the central monitoring server 10 via student terminals 12 to create a basic data profile for each student.

[0129] Secondly, during daily teaching, the classroom environment monitoring unit 11 records the classroom light intensity and air quality in real time. Preferably, when the classroom light level is detected to be below 500 lux, the system automatically sends a reminder to the teacher management terminal 13, suggesting that the lighting conditions be adjusted. Research shows that insufficient classroom lighting is one of the main environmental factors that contribute to student visual fatigue.

[0130] During classroom learning and self-study, the student terminal device 12 continuously monitors the student's distance from the book or electronic device, as well as eye movement data. Unlike Example 1, in the school scenario, the system automatically adjusts monitoring parameters based on the type of course. For example, in a reading class, the system focuses more on close-up eye use; in a physical education class, it focuses more on outdoor light exposure.

[0131] In particular, this embodiment introduces a class-level group analysis function. The prediction module 5 builds an improved deep learning model based on the whole class data:

[0132] ,

[0133] in, Indicates the class myopia risk assessment, For personal data characteristics, For environmental characteristics, This multi-level modeling approach enables the system to identify class-level risk factors, such as whether a particular teacher's teaching style leads to excessive close eye use.

[0134] In terms of intervention strategies, this embodiment incorporates the characteristics of school schedules and provides for distance-gazing activities during breaks. Specifically, after 45 consecutive minutes of class, the system prompts teachers via the teacher management terminal 13 to organize students for a "5-minute distance-gazing" activity. This activity requires students to gaze out the window at a distant object (at least 6 meters away), effectively relieving ciliary muscle fatigue. Research has shown that taking a 5-minute break every 45 minutes can reduce the risk of developing myopia by approximately 25%.

[0135] This example also incorporates a class competition mechanism, ranking classes by their outdoor activity time and myopia progression rate, encouraging students to actively participate in vision protection activities. This competition mechanism, combined with a reinforcement learning algorithm, significantly improves student participation and compliance.

[0136] In terms of assessment, in addition to individual assessments, the system also provides class-wide assessment reports to help school administrators identify possible systemic issues. For example, if myopia progression in one class is significantly faster than in other classes, the system will analyze possible causes (such as classroom lighting and course schedules) and provide targeted recommendations.

[0137] Example 4: Clinical follow-up application

[0138] This embodiment provides a myopia prediction and intervention system for ophthalmology clinics to support doctors' decision-making. The system includes:

[0139] Medical-grade data acquisition module 20, used to accurately measure clinical indicators such as refraction and axial length;

[0140] Medical history management module 21, integrating patients’ previous examination records and treatment plans;

[0141] Professional predictive analysis module 22, running an enhanced deep learning model;

[0142] Treatment plan generation module 23, which provides evidence-based intervention recommendations;

[0143] Follow-up management module 24 arranges regular review and records treatment feedback.

[0144] In this embodiment, the system first uses the medical-grade data acquisition module 20 to acquire more comprehensive clinical data, including parameters such as refractive power (±0.25D accuracy), corneal curvature, and axial length (±0.01mm accuracy). This high-precision data is the basis for accurate prediction of myopia.

[0145] In terms of prediction models, the professional prediction analysis module 22 uses a more complex deep learning architecture that integrates clinical measurement data with daily monitoring data:

[0146] ,

[0147] in, represents the probability of myopia progression, is the clinical measurement feature vector, is the daily monitoring feature vector, and is the corresponding weight matrix, is the bias term, is the sigmoid activation function. Clinical measurement data is typically updated every 3-6 months, while routine monitoring data is continuously updated. The combination of the two can provide a more comprehensive basis for prediction.

[0148] In particular, this embodiment establishes differentiated prediction models for different types of myopia (such as simple myopia and pathological myopia). For patients at high risk of pathological myopia (such as those with both parents experiencing severe myopia), the system will provide more frequent monitoring recommendations and more proactive intervention plans.

[0149] In terms of intervention strategies, the treatment plan generation module 23 can recommend a variety of intervention options based on the prediction results, including:

[0150] Optical intervention: frame glasses, orthokeratology lenses or multifocal soft contact lenses, etc., with specific fitting parameters;

[0151] Drug intervention: Low-concentration atropine (e.g., 0.01%, 0.025%, or 0.05%), with frequency of use and precautions;

[0152] Behavioral interventions: customized outdoor activity programs and near vision management strategies.

[0153] The system calculates the expected effects of each intervention plan based on factors such as the patient's age, degree of myopia, and rate of progression, and presents this information to the physician in a visual format to assist in clinical decision-making. For example, for children aged 8-10 years with myopia between -1.00D and -3.00D and a rate of progression greater than -0.50D / year, the system may recommend orthokeratology lenses combined with behavioral intervention as the preferred option, which is expected to slow myopia progression by approximately 60%.

[0154] The Follow-up Management Module 24 tracks patient treatment outcomes and dynamically adjusts prediction models and intervention recommendations based on actual progress. This module utilizes intelligent reminders to automatically notify doctors and patients' families when patients miss follow-up appointments or when monitoring data indicates abnormal progression.

[0155] This embodiment also innovatively introduces the concept of a time-sensitive intervention window. Research shows that there is a "golden period" for myopia intervention, typically 12-24 months after the onset of myopia. Therefore, the system will specifically flag patients in this period and recommend that doctors prioritize their attention.

[0156] The assessment module further refines the assessment criteria in clinical applications, including changes in myopia degree and also includes:

[0157] Changes in axial length (increase <0.1mm / year is the ideal control effect);

[0158] changes in regulatory function (assessed by regulatory hysteresis);

[0159] Improved quality of life (assessed by standardized questionnaires).

[0160] This multidimensional evaluation method makes the evaluation of intervention effects more comprehensive and objective, and provides a scientific basis for the adjustment of long-term treatment strategies.

[0161] In summary, this embodiment provides a more professional and accurate myopia prediction and intervention system for clinical application scenarios, realizes digital management of the entire process from diagnosis to treatment to follow-up, and significantly improves the clinical effect and medical efficiency of adolescent myopia management.

[0162] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for predicting and intervening in the progression of myopia in adolescents, characterized by: The following steps are involved: Acquisition steps: Acquisition of genetic data, environmental data, lifestyle data, eye behavior data, and vision data; Prediction step: building a deep learning model based on the data to predict the development trend of myopia; Monitoring step: using multimodal fusion technology to monitor eye behavior in real time and analyze the eye behavior data; Intervention step: generating a myopia intervention plan based on the prediction results of the deep learning model; Evaluation step: Evaluate the effectiveness of the intervention and determine whether a stable state has been achieved.

2. The method for predicting and intervening in myopia progression in adolescents according to claim 1, characterized in that: The acquisition of the eye behavior data specifically includes: collecting the visual angle information and blinking frequency within 60 seconds through independent observation of both eyes at a time interval of 0.5 seconds, and generating the eye behavior data using a multimodal fusion eye behavior analysis method.

3. The method for predicting and intervening in myopia progression in adolescents according to claim 2, characterized in that: The multimodal fusion eye behavior analysis method comprises the following steps: The subject's eyes are used as two independent observation perspectives, and two cameras are set up to capture the subject's gaze angle; The data collected by the two cameras are uploaded in real time and fused into an animation in chronological order to determine whether the outdoor activity situation meets the standards; Eye movement data analysis technology based on deep learning is used to process the collected eye movement data into a gaze angle curve to determine whether the time spent using eyes at close range meets the standard.

4. The method for predicting and intervening in myopia progression in adolescents according to claim 3, wherein: The standard for determining the time spent using the eyes at close range is as follows: if it is detected that the subject is less than 40 cm away from the display screen, the timer is started; if it is detected that the subject is greater than or equal to 40 cm away from the display screen, the timer is stopped; and whether the time spent using the eyes at close range meets the standard is determined based on the timing data of the timer.

5. The method for predicting and intervening in the progression of myopia in adolescents according to claim 4, characterized in that: The animation includes outdoor activity animation and close-up eye animation; The output curve of the outdoor activity animation is: using the average outdoor eye behavior data of the subject's two eyes as a reference standard, the collected gaze angle curve data of the subject is normalized, and the normalized gaze angle curve animation is output; The output curve of the close-up eye animation is: using the gaze angle curve data of the subject before myopia as a reference standard, standardizing the collected gaze angle curve data of the subject, and outputting the standardized gaze angle curve data as a reference standard.

6. The method for predicting and intervening in the progression of myopia in adolescents according to claim 5, characterized in that: The intervention step utilizes a reinforcement learning algorithm to construct the close-up eye animation, specifically: When eye use behavior deviates from healthy standards, it will be punished; when it meets the standards, it will be rewarded.

7. The method for predicting and intervening in the progression of myopia in adolescents according to claim 6, characterized in that: The myopia progression intervention program includes: Remind subjects to engage in outdoor activities; Analyze the subjects' close-up eye use time data; Based on the deep learning model, the progression of myopia in the subjects is analyzed and future intervention plans are generated.

8. The method for predicting and intervening in the progression of myopia in adolescents according to claim 7, characterized in that: In the prediction step, the deep learning model is used to predict the future development trend of the subject based on the eye behavior data, close-range eye use time data, myopia development time and myopia degree data, and generate a future intervention plan; if it is predicted that the subject's future myopia degree will be greater than 600 degrees, the subject is reminded to consider surgical treatment.

9. The method for predicting and intervening in the progression of myopia in adolescents according to claim 8, characterized in that: The evaluation steps are specifically as follows: Based on the subjects' eye behavior data, eye usage time data, and myopia progression data during the intervention process, the intervention effect and appropriate intervention time point are predicted; When the degree of myopia progression decreases by more than 0.90D (i.e. 0.90 diopters), the intervention is judged to be successful and the intensity of intervention can be gradually reduced; when the degree of myopia progression decreases by a smaller amount but more than 0D, the intervention is judged to be effective and the intervention should continue; when the degree of myopia progression does not change, the intervention is judged to be ineffective and a new intervention plan needs to be formulated.

10. A system for predicting and intervening in the progression of myopia in adolescents implementing the method according to any one of claims 1 to 9, comprising: Genetic module, used to obtain genetic data that influences the development of myopia; Environmental module, used to obtain environmental data that affects the development of myopia; Lifestyle habit module, used to obtain data on lifestyle habits that affect the development of myopia; Eye behavior module, used to obtain eye behavior data that affects the development of myopia; A prediction module, which is used to build a deep learning model based on the acquired data and predict the development trend of myopia; A control module is used to control the eye behavior module to monitor eye behavior in real time using multimodal fusion technology and analyze eye behavior data when the subject wears glasses or orthokeratology lenses; An intervention module, configured to generate a myopia intervention plan based on the prediction results of the deep learning model; The evaluation module is used to evaluate the intervention effect and determine whether a stable state has been reached.

Images