A method for monitoring data aggregation and intelligent analysis of refractive errors in children and adolescents
By constructing a multi-source heterogeneous dataset and using ST-RNN and LSTM models for feature extraction and prediction, the problem of multi-source heterogeneous data alignment was solved, achieving efficient and accurate refractive error monitoring data analysis and providing support for future risk prediction and prevention and control decisions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG GENERAL HOSPITAL
- Filing Date
- 2026-03-20
- Publication Date
- 2026-07-14
AI Technical Summary
When processing refractive error monitoring data across the country or large regions, existing technologies struggle to efficiently align heterogeneous data from multiple sources, leading to statistical biases and wasted computational resources. They also fail to accurately quantify the impact of environmental factors on myopia and lack effective analysis and prediction of future refractive error distribution.
We employ an intelligent analysis method that aggregates refractive error monitoring data from children and adolescents. By constructing a multi-source heterogeneous dataset, we use spatial weight matrices and Bayesian networks for data matching, and combine ST-RNN and LSTM models for feature extraction and prediction, thereby conducting multi-dimensional risk extrapolation and simulation analysis.
It achieves efficient alignment of multi-source heterogeneous data, improves the accuracy of data fusion, can accurately identify high-risk areas and key time points, and outputs a probability distribution map of future refractive error prevalence, providing data support for prevention and control.
Smart Images
Figure QLYQS_15
Abstract
Description
Technical Field
[0001] This invention relates to a disease monitoring and analysis method, and more particularly to a method for analyzing refractive error monitoring data. Background Technology
[0002] Existing refractive error monitoring data is usually compiled at the administrative division level, such as schools or regions; while environmental data, such as light intensity and air pollution, are usually continuous data based on geographic grids. The two exhibit spatial heterogeneity, either "area-area" or "point-area," and direct superposition analysis can lead to statistical bias, making it difficult to accurately quantify the impact of environmental factors on myopia.
[0003] When processing high-resolution environmental data and massive clinical data across a nationwide or large region, direct pixel-by-pixel or item-by-item matching is computationally intensive. Traditional aggregation methods often ignore the proximity effect and the weight of population movement in geographic space, resulting in wasted computational resources and an inability to reflect the true effects of environmental exposure. Furthermore, the acquired monitoring data should not only reflect the current and past distribution of refractive errors, but also be used to analyze and predict the future development and distribution of refractive errors based on historical data. This allows for proactive early intervention based on the prediction results, achieving better prevention and control effects.
[0004] Therefore, there is an urgent need for a refractive error data aggregation and analysis method that can achieve efficient alignment of multi-source heterogeneous spatiotemporal data. Summary of the Invention
[0005] To address the technical challenge of efficiently aligning multi-source heterogeneous data, including clinical medical data, environmental data, and genetic data, when processing refractive error (RE) prevention and control data in children and adolescents, this paper presents an intelligent analysis method for aggregating and analyzing refractive error monitoring data in children and adolescents.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for intelligent analysis of refractive error monitoring data in children and adolescents, the method comprising the following steps: Step 1. Construct a multi-source heterogeneous refractive error monitoring dataset Obtain clinical statistics on refractive errors in children and adolescents in the target region, and record them as clinical data of administrative division units. Environmental data, referred to as environmental raster data. And based on this, a multi-source heterogeneous refractive error monitoring dataset was constructed; Step 2. Dynamic aggregation of heterogeneous data based on multidimensional spatial weights Clinical data for administrative divisions Calculate its relationship with environmental raster data Spatial topological relationships to solve the problem of "face-to-face" data mismatch; Step 3. Spatiotemporal feature embedding based on ST-RNN and Bayesian network The aggregated structured data is input into the spatial-temporal recurrent neural network ST-RNN, and combined with a Bayesian network for feature extraction and correlation analysis. Step 4. In-depth data mining and risk projection analysis based on aggregated data The refractive error monitoring data after aggregation in step 2 and feature extraction in step 3 are used to predict the time-series features of the monitoring data based on the ST-RNN architecture and to extrapolate environmental intervention scenarios based on Monte Carlo simulation. The results of risk prediction analysis are obtained, which not only realizes the prediction of future trends, but also focuses on quantifying the effect of environmental intervention through simulation analysis, so as to provide data support for prevention and control decisions.
[0007] Further, step 1 includes:
[0008] 1.1 Clinical Data Acquisition: Collect demographic information and key clinical refractive indicators for the target area; demographic information includes age, gender, etc.; key clinical refractive indicators include axial length, corneal curvature, lens thickness, refractive power, etc.
[0009] 1.2 Environmental Data Acquisition: Acquire natural environmental data and socio-economic environmental data of the target area; natural environmental data include light intensity, PM2.5 concentration, temperature, humidity, altitude, etc.
[0010] 1.3 Basic Database Construction: A unified spatial reference system is established using Geographic Information System (GIS), and a multi-source heterogeneous refractive error monitoring dataset containing spatiotemporal dimensions is constructed.
[0011] Further, step 2 includes: 2.1 Grid Attribution Determination: For each administrative division Determine the set of all environmental grid cells covering its range. ; 2.2 Construction of Composite Weights: Introducing a Spatial Weight Matrix Based on the coverage ratio of the grid cells within the administrative region and population density distribution Construct composite weights ;in This is the adjustment coefficient; 2.3 Weighted Aggregate Calculation: This involves combining environmental variables... Aggregated to administrative districts value The calculation method is as follows: Thus, high-dimensional heterogeneous environmental raster data is reduced in dimensionality and mapped to a structured tensor isomorphic to clinical data.
[0012] Further, step 3 includes: 3.1 Spatial Embedding: A regional adjacency matrix is constructed using a graph convolutional network (GCN) to capture the spatial spillover effect between adjacent administrative regions, such as the impact of inter-city population flow on the spread of myopia. 3.2 Temporal Recursion: The Long Short-Term Memory (LSTM) network is used to capture the trend of refractive error prevalence over time, such as the annual growth trend; 3.3 Bayesian Association Analysis: Based on the features extracted by ST-RNN, a Bayesian network is constructed to quantify the causal relationship between environmental factors and the evolution of refractive errors. The environmental factors include light and pollution.
[0013] Furthermore, the time-series feature prediction of monitoring data based on the ST-RNN architecture in step 4 specifically involves: Using the Long Short-Term Memory (LSTM) module of the spatial-temporal recurrent neural network (ST-RNN) constructed in step 3, the aggregated multidimensional monitoring data is trained, including: Input features: The coupling features of key environmental factors such as light and pollution with key clinical refractive indicators selected by the Bayesian network in step 3 are used. Predictive modeling: Train an LSTM model to learn the nonlinear evolution of refractive error prevalence and key clinical refractive indicators over time, and establish a high-precision time-series prediction model.
[0014] Furthermore, step 4, which involves the environmental intervention scenario simulation based on Monte Carlo simulation, includes: Scenario setting: Set up a baseline scenario without intervention, an environmental improvement scenario, and an environmental degradation scenario; the environmental improvement scenario includes increasing outdoor light and reducing pollution; Random sampling simulation: The Monte Carlo method is used to randomly sample environmental variable parameters to simulate the evolution path of refractive error data under different external conditions; Risk projection analysis: Outputs a probability distribution map of future refractive error prevalence, identifying high-risk areas and key time points.
[0015] The beneficial effects of this invention are as follows: By introducing a dynamic aggregation algorithm with dual weights of area and population, this invention effectively solves the statistical bias problem of "surface-to-surface" mismatch between clinical data and environmental data, thus improving the accuracy of data fusion. Through multi-dimensional risk extrapolation and prediction of the aggregated monitoring data, a probability distribution map of refractive error prevalence over the next 5-10 years can be output, accurately identifying high-risk areas and key time points, achieving a closed-loop analysis from data monitoring to intelligent prevention and control. Detailed Implementation
[0016] The present invention will be further described in detail below with reference to embodiments. A method for intelligent analysis of refractive error monitoring data in children and adolescents includes the following four core steps:
[0017] Step 1: Construct a multi-source heterogeneous refractive error monitoring dataset Obtain clinical statistics on refractive errors in children and adolescents in the target region, and record them as clinical data of administrative division units. Environmental data is recorded as environmental raster data. .
[0018] Specific operations: Clinical data acquisition: Collection includes key clinical refractive indicators such as axial length, corneal curvature, lens thickness, and refractive power, as well as demographic information such as age, gender, and BMI.
[0019] Environmental data acquisition: Acquire natural environmental data of the target area, such as light intensity, PM2.5 concentration, temperature, humidity, altitude, and socio-economic environmental data.
[0020] Basic database construction: A unified spatial reference system is established using Geographic Information System (GIS) to construct a basic database for refractive error monitoring that includes spatiotemporal dimensions.
[0021] Step 2: Dynamic aggregation of heterogeneous data based on multidimensional spatial weights Clinical data for administrative divisions Calculate its relationship with the environment grid. The spatial topology relationship solves the problem of "face-to-face" data mismatch.
[0022] Specific operations: Grid affiliation determination: For each administrative division Determine the set of all environmental grid cells covering its range. .
[0023] Composite weight construction: Introducing a spatial weight matrix Based on the coverage area ratio of the grid cells within the administrative region. and population density distribution Construct composite weights ,in This is the adjustment coefficient.
[0024] Weighted aggregation calculation: This involves combining environmental variables... Aggregated to administrative districts value The calculation is as follows: This step reduces the dimensionality of high-dimensional heterogeneous environmental raster data and maps it to a structured tensor isomorphic to clinical data.
[0025] Step 3: Spatiotemporal Feature Embedding Based on ST-RNN and Bayesian Network The aggregated structured data is input into a spatial-temporal recurrent neural network (ST-RNN) and combined with a Bayesian network for feature extraction and correlation analysis.
[0026] Specific operations: Spatial embedding: Using graph convolutional networks (GCNs) to construct regional adjacency matrices, we can capture the spatial spillover effect between adjacent administrative regions, such as the impact of inter-city population flow on the spread of myopia.
[0027] Temporal recursion: Utilizing the Long Short-Term Memory (LSTM) network to capture trends in the prevalence of refractive errors over time, such as annual growth trends.
[0028] Bayesian association analysis: Based on the features extracted by ST-RNN, a Bayesian network is constructed to quantify the causal association strength between environmental factors, such as light, pollution and the evolution of refractive errors.
[0029] Step 4: In-depth data mining and risk simulation analysis based on aggregated data This step aims to perform final intelligent analysis and application on the refractive error monitoring data after aggregation in step two and feature extraction in step three. It not only enables the prediction of future trends, but also focuses on quantifying the effects of environmental interventions through simulation analysis, providing data support for prevention and control decisions.
[0030] Specific operations: 1. Prediction of temporal features of monitoring data based on ST-RNN architecture: The Long Short-Term Memory (LSTM) module in the ST-RNN spatial-temporal recurrent neural network constructed in step 3 is used to train the aggregated multidimensional monitoring data.
[0031] Input features: The key environmental factors screened by the Bayesian network in step three, such as the coupling features of light, pollution and key clinical refractive indicators.
[0032] Predictive modeling: Train the LSTM model to learn the prevalence of refractive errors and key clinical refractive indicators, such as the nonlinear evolution of axial length growth over time, and establish a high-precision time-series prediction model.
[0033] 2. Monte Carlo simulation-based environmental intervention scenario extrapolation: To evaluate the effectiveness of different prevention and control strategies, the Monte Carlo simulation method is introduced to perform multi-dimensional risk extrapolation on the aggregated monitoring data.
[0034] Scenario setting: Set up a baseline scenario without intervention and an environmental improvement scenario, such as increasing outdoor light, reducing pollution, and environmental degradation scenarios.
[0035] Random sampling simulation: The Monte Carlo method is used to randomly sample environmental variable parameters to simulate the evolution path of refractive error data under different external conditions.
[0036] Risk output: Outputs a probability distribution map of the prevalence of refractive errors in the next 5-10 years, accurately identifies high-risk areas and key time nodes, and realizes a closed-loop analysis from "data monitoring" to "intelligent prevention and control".
[0037] The above content is only used to illustrate the technical solution of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.
Claims
1. A method for intelligent analysis of refractive error monitoring data in children and adolescents, characterized in that: The method includes the following steps: Step 1. Construct a multi-source heterogeneous refractive error monitoring dataset Obtain clinical statistics on refractive errors in children and adolescents in the target region, and record them as clinical data of administrative division units. Environmental data, referred to as environmental raster data. And based on this, a multi-source heterogeneous refractive error monitoring dataset was constructed; Step 2. Dynamic aggregation of heterogeneous data based on multidimensional spatial weights Clinical data for administrative divisions Calculate its relationship with environmental raster data Spatial topological relationships; Step 3. Spatiotemporal feature embedding based on ST-RNN and Bayesian network The aggregated structured data is input into the spatial-temporal recurrent neural network ST-RNN, and combined with a Bayesian network for feature extraction and correlation analysis. Step 4. In-depth data mining and risk projection analysis based on aggregated data The refractive error monitoring data after aggregation in step 2 and feature extraction in step 3 are used to predict the time-series features of the monitoring data based on the ST-RNN architecture and to extrapolate environmental intervention scenarios based on Monte Carlo simulation, so as to obtain risk prediction and analysis results.
2. The intelligent analysis method for aggregating refractive error monitoring data in children and adolescents according to claim 1, characterized in that: Step 1 includes: 1.1 Clinical Data Acquisition: Collect demographic information and key clinical refractive indicators for the target area; 1.2 Environmental Data Acquisition: Acquire natural environmental data and socio-economic environmental data for the target area; 1.3 Basic Database Construction: A unified spatial reference system is established using Geographic Information System (GIS), and a multi-source heterogeneous refractive error monitoring dataset containing spatiotemporal dimensions is constructed.
3. The intelligent analysis method for aggregating refractive error monitoring data in children and adolescents according to claim 1, characterized in that: Step 2 includes: 2.1 Grid Attribution Determination: For each administrative division Determine the set of all environmental grid cells covering its range. ; 2.2 Construction of Composite Weights: Introducing a Spatial Weight Matrix Based on the coverage ratio of the grid cells within the administrative region and population density distribution Construct composite weights ;in This is the adjustment coefficient; 2.3 Weighted Aggregate Calculation: This involves combining environmental variables... Aggregated to administrative districts value The calculation method is as follows: .
4. The intelligent analysis method for aggregating refractive error monitoring data of children and adolescents according to claim 1, characterized in that: Step 3 includes: 3.1 Spatial Embedding: A regional adjacency matrix is constructed using a graph convolutional network (GCN) to capture the spatial spillover effect between adjacent administrative regions; 3.2 Temporal Recursion: Using a Long Short-Term Memory (LSTM) network to capture the trend of refractive error prevalence over time; 3.3 Bayesian association analysis: Based on the features extracted by ST-RNN, a Bayesian network is constructed to quantify the causal association strength between environmental factors and the evolution of refractive errors.
5. The intelligent analysis method for aggregating refractive error monitoring data of children and adolescents according to claim 1, characterized in that: The time-series feature prediction of monitoring data based on the ST-RNN architecture in step 4 specifically involves: Using the Long Short-Term Memory (LSTM) module of the spatial-temporal recurrent neural network (ST-RNN) constructed in step 3, the aggregated multidimensional monitoring data is trained, including: Input features: The coupling features of key environmental factors and key clinical refractive indicators selected by the Bayesian network in step 3 are used. Predictive modeling: Train an LSTM model to learn the nonlinear evolution of refractive error prevalence and key clinical refractive indicators over time, and establish a high-precision time-series prediction model.
6. The intelligent analysis method for aggregating refractive error monitoring data of children and adolescents according to claim 1, characterized in that: Step 4, which involves environmental intervention scenario simulation based on Monte Carlo simulation, includes: Scenario setting: Set up a baseline scenario with no intervention, an environmental improvement scenario, and an environmental degradation scenario; Random sampling simulation: The Monte Carlo method is used to randomly sample environmental variable parameters to simulate the evolution path of refractive error data under different external conditions; Risk projection analysis: Outputs a probability distribution map of future refractive error prevalence, identifying high-risk areas and key time points.