Dementia risk prediction system based on associative information processing model

Abstract

The present invention relates to the field of health information technology, and in particular to a dementia risk prediction system based on an associative information processing model. The system first extracts multimodal latent variable features through a data processing module, and optimizes and generates an individualized risk feature matrix using causal reasoning and association rules. Subsequently, the system uses a reinforcement learning optimization algorithm to dynamically generate personalized intervention paths and achieves long-term risk prediction through a recurrent neural network. The system outputs a personalized prediction report, including dementia risk assessment, trend analysis, and intervention recommendations. Experiments have shown that the present invention can significantly improve the accuracy of dementia risk prediction, the dynamic adaptability of intervention strategies, and the efficiency of health management.

Description

Technical Field

[0001] The present invention relates to the field of health information technology, and in particular to a dementia risk prediction system based on an association information processing model. Background Art

[0002] Dementia is a common neurodegenerative disease, particularly among the elderly. Its morbidity and severity place a significant burden on individuals, their families, and society. Early prediction of dementia risk and development of personalized intervention strategies can significantly slow disease progression, improve patients' quality of life, and reduce socioeconomic burdens.

[0003] The "risk factor scoring method" and "Cox regression model probability method" used in existing technologies (Chinese invention patent, publication number: CN117373668A, title: A method for establishing a risk prediction model for the onset of dementia in the elderly) have certain application value in traditional dementia risk assessment, but they still have the following shortcomings: the risk factor scoring method and the Cox regression model probability method rely on static data analysis, ignore the dynamic correlations between multimodal data, and fail to reflect the risk trends of individual characteristics over time; the risk factor scoring method mainly calculates the risk index through a simple weighted accumulation method, while the Cox regression model mainly relies on the influence of single-variable covariates and lacks adaptability to higher-order correlations and nonlinear dependencies. Both methods are based on fixed models and cannot dynamically update or optimize prediction strategies. Their sensitivity and specificity in risk assessment are also limited, making them difficult to support real-time intervention and optimization. Summary of the Invention

[0004] To address the numerous issues with the aforementioned existing technologies, the present invention provides a dementia risk prediction system based on an associative information processing model. Based on multimodal data fusion, the system utilizes a reinforcement learning optimization algorithm to dynamically generate personalized intervention pathways and implements long-term risk prediction through a recurrent neural network. Through multi-level data analysis and optimization, the system generates precise, personalized prediction reports, significantly improving the accuracy of dementia risk prediction and the practicality of intervention measures.

[0005] A dementia risk prediction system based on an associative information processing model, comprising:

[0006] The data processing module is used to collect multimodal data, including genetic data, health data, behavioral data, and environmental data, and process all the data to generate latent variable data;

[0007] A causal inference module is used to construct a causal model based on latent variable data, calculate causal weights, and generate an individualized risk signature matrix, where the signature matrix is ​​generated by comprehensively analyzing the weighted contributions of different causal relationship variables; based on the signature matrix, a nonlinear regression algorithm is applied to calculate individualized risk index data to reflect the likelihood of dementia for each individual;

[0008] An intervention optimization module, which builds a reinforcement learning environment based on individualized risk index data and generates dynamically adjusted intervention strategy data through a deep Q-network-based optimization algorithm. The optimization algorithm uses a reinforcement learning framework, combined with real-time updated risk indicators and intervention feedback, to optimize intervention pathways to reduce dementia risk;

[0009] The prediction and generation module is used to perform long-term risk prediction, short-term trend analysis and individual stratification screening based on the dynamically adjusted intervention strategy data, and generate personalized prediction report data, which includes risk assessment and intervention recommendations.

[0010] Preferably, the data processing module includes a missing data filling unit, which processes missing values ​​in the multimodal data by the following steps:

[0011] For continuous numerical data, mean interpolation is used to fill missing values ​​based on the mean of other data records;

[0012] For categorical variables, the nearest neighbor algorithm-based imputation method is used to select the most common category value in similar data records for imputation;

[0013] The padded multimodal data generates initial latent variable data.

[0014] Preferably, the data processing module performs time series analysis on the behavioral data and the environmental data, extracts trend features and updates latent variable data through the following steps:

[0015] Behavioral data and environmental data are divided into time windows of fixed length, and features are calculated independently for the data in each time window;

[0016] Calculate the mean of the data in each time window to generate the mean parameter in the trend feature;

[0017] Calculate the rate of change of the mean values ​​in adjacent time windows to generate the rate of change parameter in the trend feature;

[0018] Applying linear regression analysis to the data within the time window, calculating the trend slope parameter, and generating the slope parameter in the trend feature;

[0019] The initial latent variable data is updated based on the extracted trend features to obtain updated latent variable data.

[0020] Preferably, the individualized risk signature matrix generated by the causal reasoning module is calculated by the following steps:

[0021] Determine the causal weight and calculate the causal weight using the following formula:

[0022]

[0023] in, Representation characteristics Features The causal weight of Characterized by and features The covariance of Characterized by variance;

[0024] The eigenvalues ​​in the feature matrix are calculated based on the causal weights, which is achieved through the following formula:

[0025]

[0026] in, is the eigenvalue of the matrix, Representation characteristics In the characteristic matrix, the eigenvalues The contribution weight of is the latent variable feature, and n is the total number of all feature variables involved in the data calculation process.

[0027] Preferably, the causal reasoning module optimizes the feature matrix in combination with the association rules, and is implemented by the following steps:

[0028] Use the Apriori algorithm to mine association rules between features from latent variable data;

[0029] Adjust the feature matrix according to the mined rules and weight the association rules to causal weights;

[0030] The adjusted feature matrix is ​​regenerated by combining the causal weights and the association weights.

[0031] Preferably, the individualized risk index is calculated by the following formula:

[0032]

[0033] in, represents the individualized risk index, Characterized by The regression coefficient of is the latent variable eigenvalue, is the constant term of regression, and n is the total number of all characteristic variables involved in the data calculation process.

[0034] Preferably, the reinforcement learning environment of the intervention optimization module includes the following parts:

[0035] State space, consisting of individualized risk index data and high-risk factor annotations;

[0036] Action sets, including dietary adjustments, exercise recommendations, and medications;

[0037] The reward function is calculated by the following formula:

[0038]

[0039] in, Represents the reward value, is the change in risk index before and after intervention, The economic and time costs of the intervention.

[0040] Preferably, the optimization algorithm of the reinforcement learning adopts a deep Q network, and the training process includes the following steps:

[0041] Initialize the parameters of the reinforcement learning model and determine the initial state through individualized risk index data;

[0042] use - Greedy strategies, which balance between exploring new interventions and exploiting existing ones;

[0043] The Q value is updated in multiple rounds of iterations, and the intervention path is continuously optimized through the reward function.

[0044] Preferably, the long-term risk prediction task of the prediction and generation module is implemented through a recurrent neural network, the input features include the individualized risk index and high-risk factor labels in the intervention strategy data after dynamic adjustment, and the output is a risk curve in the future time.

[0045] Preferably, the personalized prediction report includes the following contents:

[0046] Summary of dementia risk assessment; long-term trend analysis results, including changes in the risk curve; short-term trend analysis results; individual stratification results and key high-risk factors; and specific intervention recommendations for individuals.

[0047] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0048] This invention uses multimodal data fusion technology to achieve accurate extraction of latent variable features and dynamic association modeling, overcoming the limitations of traditional methods that rely solely on univariate analysis and enhancing the ability to comprehensively characterize factors affecting dementia risk.

[0049] This invention uses a reinforcement learning-driven intervention optimization algorithm to achieve real-time adjustment of personalized intervention paths, effectively solving the problem of fixed intervention strategies in existing technologies and improving dynamic risk management capabilities.

[0050] This invention uses the long-term risk prediction technology of recurrent neural networks to accurately predict future risk trends, providing scientific support for planning intervention measures in advance.

[0051] The present invention achieves comprehensive optimization of stratified screening and intervention recommendations through a personalized prediction report generation module, achieving both high efficiency and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a structural block diagram of the system of the present invention;

[0053] Figure 2 This is a data flow diagram of the causal reasoning module in the present invention;

[0054] Figure 3 This is a schematic diagram of the reinforcement learning process of the intervention optimization module in the present invention;

[0055] Figure 4 A schematic diagram of the process of generating a personalized forecast report in the present invention. DETAILED DESCRIPTION

[0056] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure.

[0057] like Figure 1 As shown, a dementia risk prediction system based on an associative information processing model includes:

[0058] The data processing module is used to collect multimodal data, including genetic data, health data, behavioral data, and environmental data, and process all the data to generate latent variable data;

[0059] Genetic data refers to genetic information extracted from an individual's genome that is associated with dementia risk. Genetic variations (such as single nucleotide polymorphisms (SNPs)) can reflect an individual's genetic susceptibility to neurodegenerative diseases. This invention analyzes genetic data to identify gene loci and their expression relationships that influence the onset of dementia, thereby providing a genetic basis for risk assessment.

[0060] Genetic data is typically obtained through high-throughput sequencing technologies (such as whole-genome sequencing or gene chip technology). Genetic loci such as APOE4, CLU, and PICALM have been shown to be closely associated with dementias such as Alzheimer's disease. Genetic data is represented in a coded format; for example, each locus can be coded as 0 (wild type), 1 (heterozygous), or 2 (mutant). Analysis of genetic data can identify the contribution of genetic factors to dementia risk, with particularly high predictive value for individuals with a family history or genetic susceptibility. The inclusion of genetic data provides a foundational dimension for assessing dementia risk.

[0061] In this example, an individual's genetic data indicates that their APOE4 locus is 2 (mutated), a site that has been shown to be a high-risk factor for dementia. In the present invention, the weight of this locus is assigned in the latent variable matrix, directly influencing the calculation of the individualized risk index.

[0062] Health data refers to continuous or categorical variables related to an individual's physical health status, reflecting their physiological state and disease risk. Common health data include blood sugar, blood pressure, BMI (body mass index), heart rate, and cholesterol levels. These indicators are often closely related to cardiovascular and cerebrovascular health and have a significant impact on the onset of dementia.

[0063] Health data can be collected through physical examination reports, medical devices (such as blood pressure monitors, blood glucose meters), or wearable devices. Health data are usually stored as continuous variables and processed using standardized formulas before being input into the model. For example, hypertension and diabetes are important risk factors for dementia. Long-term elevated blood pressure may cause damage to brain blood vessels, thereby increasing the risk of dementia. Health data can supplement an individual's current health status that cannot be reflected in genetic data, and provide a controllable dimension for dementia risk assessment (such as improving blood pressure or blood glucose levels through drug intervention).

[0064] For example, a person's fasting blood glucose level of 7.5 mmol / L (high) combined with their BMI (27, indicating overweight) suggests a predisposition to metabolic syndrome. These indicators are then combined with behavioral and genetic data in a latent variable matrix to comprehensively assess an individual's risk of dementia.

[0065] Behavioral data reflects an individual's lifestyle and daily habits, which indirectly influence dementia risk by impacting physical and brain health. Common behavioral data include exercise levels, dietary habits, smoking history, and alcohol consumption. Behavioral data can be obtained through self-reports, wearable devices (such as pedometers), or smart monitoring devices. Quantitative data (such as walking distance) within behavioral data is stored as a time series, while qualitative data (such as dietary preferences) is processed using vectorization.

[0066] Regular exercise can reduce the risk of dementia, and walking frequency and distance are important indicators; long-term smoking and excessive drinking are high-risk behavioral factors for dementia; high-fat, high-sugar diets may increase inflammatory responses and accelerate cognitive decline.

[0067] Behavioral data provide the most modifiable risk factors for dementia, and behavioral interventions (such as improving diet and increasing exercise) can significantly reduce risk.

[0068] For example, a person's daily step count, recorded using a smart bracelet, shows only 2,000 steps (below the recommended 8,000). Furthermore, their dietary history indicates a high proportion of high-fat intake. In the model, these behavioral traits are flagged as high-risk factors, increasing their individualized risk index.

[0069] Environmental data refers to the impact of external environmental factors on dementia risk, including air pollution, residential environment (e.g., urban vs. rural), and social support networks. These factors indirectly influence dementia risk through long-term effects on individual health or psychological well-being. Environmental data can be obtained through public databases, meteorological monitoring stations, or questionnaires. Processing methods include: directly inputting quantitative data (e.g., PM2.5 concentration) into the model; and processing qualitative data (e.g., residential type) through classification and vectorization.

[0070] Long-term exposure to high levels of PM2.5 may increase brain inflammation, accelerating cognitive decline. Loneliness and lack of social support are significant sources of psychological stress, indirectly increasing the risk of dementia. Environmental data can capture the long-term impact of an individual's external environment, providing an uncontrollable but critical reference factor for risk assessment.

[0071] For example, consider an individual living in an urban area with PM2.5 concentrations consistently exceeding 80 micrograms per cubic meter and having a low social support network index (e.g., living alone and infrequent socializing). The model assigns high weight to these factors, making them crucial for assessing individual risk.

[0072] Preferably, the data processing module includes a missing data filling unit, which processes missing values ​​in the multimodal data by the following steps:

[0073] For continuous numerical data, mean interpolation is used to fill missing values ​​based on the mean of other data records;

[0074] For categorical variables, the nearest neighbor algorithm-based imputation method is used to select the most common category value in similar data records for imputation;

[0075] The padded multimodal data generates initial latent variable data.

[0076] In this invention, the missing data filling unit of the data processing module is used to address the problem of missing values ​​in multimodal data, ensuring the completeness and accuracy of subsequent analysis. Multimodal data comes from complex sources, and different types of data (such as genetic data, health data, behavioral data, and environmental data) may contain missing values ​​during the collection process. Without filling, these missing values ​​can lead to reduced model performance or even inability to process data.

[0077] For continuous numerical data (such as blood sugar, blood pressure and other health indicators), mean interpolation is used to fill in the missing values. Mean interpolation calculates the mean of other records in the dataset to fill in the missing values, thereby minimizing the bias introduced by the interpolation. The formula is as follows:

[0078]

[0079] in, To fill the value, is the value of the non-missing record, The total number of all characteristic variables involved in the data calculation process.

[0080] For categorical variables (such as smoking status, residential area category, etc.), the nearest neighbor algorithm is used to fill in the missing values. The nearest neighbor algorithm calculates the similarity between each record and selects the most common category value in the most similar records to fill in the missing values. The nearest neighbor distance calculation formula is:

[0081]

[0082] in, Represents a record and The Euclidean distance between and Record and The eigenvalues ​​of is the feature dimension. By comparing the distances of all records, the nearest The category values ​​of the records are voted and filled with the category with the highest frequency.

[0083] After imputation, all data records are reassembled to generate a multimodal dataset without missing values. The imputed data are further converted into initial latent variable data to provide input for the subsequent causal inference module.

[0084] This approach addresses missing values ​​in multimodal data, providing consistent data input for subsequent analysis. Mean imputation for continuous variables avoids excessive computational complexity and reduces the impact of data imputation on the overall distribution. The nearest neighbor imputation method for categorical variables ensures the rationality of imputed values ​​through a voting mechanism. The design of the missing data imputation unit improves the system's ability to handle complex data and prevents model failures caused by incomplete data.

[0085] In an embodiment, in a dementia risk prediction scenario, an individual's genetic data, health data, behavioral data, and environmental data are collected. Among them, genetic data: all key gene sites have been measured, and there are no missing values. Health data: blood pressure records are missing, and other records are normal. The system uses the mean interpolation method to use the mean blood pressure of the same group of people (such as 120 / 80 mmHg) as the filling value. Behavioral data: exercise frequency data is missing. The system uses the nearest neighbor algorithm to compare the similarity of other records, and selects the most common exercise frequency to fill in (such as "walking 5 times a week") based on the walking distance and the category of daily activities. Environmental data: some air pollution index records are missing, and the system fills them with the mean of the same area and time period.

[0086] The imputed data are then integrated to generate a complete, multimodal dataset, which is then used to generate the initial latent variable data. By calculating the latent variable matrix, key high-risk factors for individuals are identified, providing support for subsequent risk assessment and intervention strategies.

[0087] Preferably, the data processing module performs time series analysis on the behavioral data and the environmental data, extracts trend features and updates latent variable data through the following steps:

[0088] Behavioral data and environmental data are divided into time windows of fixed length, and features are calculated independently for the data in each time window;

[0089] Calculate the mean of the data in each time window to generate the mean parameter in the trend feature;

[0090] Calculate the rate of change of the mean values ​​in adjacent time windows to generate the rate of change parameter in the trend feature;

[0091] Applying linear regression analysis to the data within the time window, calculating the trend slope parameter, and generating the slope parameter in the trend feature;

[0092] The initial latent variable data is updated based on the extracted trend features to obtain updated latent variable data.

[0093] In this paper, behavioral and environmental data are key dimensions of dynamic change. Extracting trend features through time series analysis effectively captures the temporal patterns of these data and further updates latent variable data, providing more timely and dynamic input for subsequent causal inference modules. Time series analysis slices the data using fixed-length time windows and extracts trend features, including mean, rate of change, and slope, through statistical and regression methods. These features reflect the dynamic characteristics of behavioral and environmental data from different perspectives.

[0094] Behavioral data (such as exercise frequency) and environmental data (such as air pollution index) are stored in time series format. To ensure timely and consistent analysis, the system divides time series data into windows of fixed length, with data in each window processed independently. The window length is set based on the actual application scenario, such as daily, weekly, or monthly, and the specific choice requires a balance between timeliness and stability.

[0095] In each time window, the mean parameter is calculated for the behavioral data and environmental data to reflect the average state of the time period. The mean calculation formula is:

[0096]

[0097] in, is the mean, is the observation value in the window, The mean parameter is used to capture long-term stable features, such as the average number of steps per day or air pollution levels over a period of time.

[0098] The rate of change is calculated for the mean values ​​within adjacent time windows to reflect the direction and magnitude of trend changes in the short term. The formula for calculating the rate of change is:

[0099]

[0100] in, is the rate of change, and are the mean values ​​of the current window and the next window, respectively. The rate of change parameter can reveal the dynamic characteristics of data fluctuations. For example, a rapid increase in the air pollution index may be associated with short-term risks.

[0101] In each time window, the trend slope is calculated by linear regression method to describe the long-term trend of the data. The slope calculation formula is:

[0102]

[0103] in, is the slope, and time and variable value respectively, and is the mean of the time and variable, and n is the total number of all characteristic variables involved in the data calculation process. The slope parameter is used to identify the long-term change direction of the data, such as the gradual decrease in walking distance within a year.

[0104] The trend features (mean, rate of change, and slope) extracted through the above steps are input into the latent variable model as new feature vectors, updating the initial latent variable data and generating updated latent variable data. This latent variable data more comprehensively reflects the dynamic characteristics and supports subsequent causal reasoning and personalized risk assessment.

[0105] Time series analysis of behavioral and environmental data can capture dynamic changes over time, supplementing information that static data cannot capture. Trend characteristics (such as slope and rate of change) can reveal potential relationships between behavioral and environmental data and dementia risk, enhancing the strength of subsequent model analysis. Dynamic updates of latent variable data allow the model to reflect current risk status, thereby enhancing the system's adaptability to time-sensitive characteristics.

[0106] In an embodiment, in a dementia risk prediction application, behavioral data and environmental data of an individual are collected and processed as follows:

[0107] The system records the individual's daily step count and the air pollution index of the corresponding area, and divides the data into a time window per week. For example:

[0108] Week 1: Steps [5000, 5200, 4800, 5500, 5300, 5100, 4900], Air Pollution Index [60, 65, 62, 70, 72, 68, 66];

[0109] Week 2: Steps [4800, 4700, 4500, 4600, 4300, 4400, 4200], Air Pollution Index [75, 78, 80, 85, 88, 86, 82].

[0110] Mean: The average number of steps in the first week was 5100, and the average air pollution index was 65.57; the average number of steps in the second week was 4500, and the average air pollution index was 82.

[0111] Change rate: The step count change rate is -11.76%, and the air pollution index change rate is +25.07%.

[0112] Trend slope: Calculated by linear regression, the slope of step count is -100 (a decrease of 100 steps per day) and the slope of the API is +3.71 (an increase of 3.71 points per day).

[0113] The extracted trend features (mean, rate of change, and slope) are input into the latent variable model, and the updated latent variable data contains the dynamic characteristics of behavioral and environmental data. For example:

[0114] Week 1: latent variable features [5100, +0, -100, 65.57, +0, +3.71];

[0115] Week 2: Latent variable features [4500, -11.76%, -100, 82, +25.07%, +3.71].

[0116] The updated latent variable data indicated that the individual's physical activity decreased significantly and air pollution increased significantly, suggesting a potential increase in dementia risk in the short term. This data will be further input into the causal inference module for risk assessment and intervention optimization.

[0117] like Figure 2 As shown, the causal reasoning module is used to build a causal model based on latent variable data, calculate causal weights and generate an individualized risk feature matrix, wherein the feature matrix is ​​generated by comprehensively analyzing the weighted contributions of different causal relationship variables; based on the feature matrix, a nonlinear regression algorithm is applied to calculate individualized risk index data to reflect the likelihood of dementia for each individual;

[0118] The causal inference module is the core module of the dementia risk prediction system based on the associative information processing model. Its main function is to construct a causal model using latent variable data, calculate causal weights, generate an individualized risk feature matrix, and ultimately calculate individualized risk index data using a nonlinear regression algorithm. Causal inference analyzes the potential causal relationships between multimodal data to quantify the impact of each variable on dementia risk, providing a foundation for accurate risk assessment and intervention decisions.

[0119] A causal model is a graph structure that describes the causal relationships between variables, where nodes represent latent variable characteristics and edges represent causal relationships. Causal models are constructed based on latent variable data, using maximum entropy optimization combined with conditional independence tests to determine the direction and strength of causal relationships between variables. For example, genetic variation may further increase dementia risk by affecting cognitive function, while chronic physical inactivity may indirectly increase risk by affecting cardiovascular health.

[0120] Causal weights reflect the strength of one variable's causal contribution to another. Based on the calculated causal weights, the weighted contributions of multiple variables are comprehensively analyzed to generate an individualized risk signature matrix. Each row of the matrix corresponds to a latent variable characteristic, and each column corresponds to an individualized eigenvalue. Based on the generated signature matrix, a nonlinear regression algorithm is applied to calculate the individualized risk index.

[0121] Preferably, the individualized risk signature matrix generated by the causal reasoning module is calculated by the following steps:

[0122] Determine the causal weight and calculate the causal weight using the following formula:

[0123]

[0124] in, Representation characteristics Features The causal weight of Characterized by and features The covariance of Features variance;

[0125] The eigenvalues ​​in the feature matrix are calculated based on the causal weights, which is achieved through the following formula:

[0126]

[0127] in, is the eigenvalue of the matrix, Representation characteristics In the characteristic matrix, the eigenvalues The contribution weight of is the latent variable feature, and n is the total number of all feature variables involved in the data calculation process.

[0128] In this paper, the causal inference module calculates causal weights from latent variable data and generates a personalized risk signature matrix based on the causal weights. This signature matrix quantifies the causal relationships between latent variable features and aims to identify the weighted contributions of different latent variables to the risk of dementia, thereby providing high-quality input for subsequent risk index calculation.

[0129] Causal weight is an important indicator for quantifying the strength of the causal relationship between features. By analyzing the statistical relationship between latent variable features, the causal weight between each pair of features is calculated. The specific formula is: The causal weight is quantified by the ratio of covariance to variance. right influence rather than simple correlation, thus avoiding the interference of non-causal relationships.

[0130] After calculating the relationship between latent variable features using causal weights, an individualized risk feature matrix is ​​generated. Each row of the feature matrix corresponds to a latent variable feature, and each column corresponds to an individualized eigenvalue. The eigenvalue is calculated using the following formula: .

[0131] During the feature matrix generation process, causal weights can be updated in real time based on new data, and the eigenvalues ​​of latent variables can also be adjusted based on dynamic changes, so that the system can adapt to the differences between different individuals and dynamic changes over time.

[0132] By calculating causal weights, we can quantify the causal relationships between latent variable features, avoiding the bias that may be introduced by simple correlation analysis. By weighting the feature matrix, the generated eigenvalues ​​can more accurately reflect the contribution of individual characteristics to dementia risk, thereby improving the accuracy of model predictions. The ability to update causal weights and feature matrices in real time enables the system to respond quickly to individual and temporal changes, improving overall performance.

[0133] In an example, in a practical application of a dementia risk prediction system, an individual's latent variable data includes genetic data (such as APOE4 gene locus status), health data (such as blood pressure and BMI), behavioral data (such as daily step count), and environmental data (such as air pollution index). The causal inference module performs the following processing on this individual's data:

[0134] The causal weight is calculated by the covariance and variance between latent variable features:

[0135] The causal weight of the APOE4 gene on the risk of dementia is:

[0136] The causal weight of daily steps on cardiovascular health is:

[0137] The causal weight of air pollution index on dementia risk is:

[0138] Generate an individualized risk signature matrix based on the causal weights and latent variable data. For example, the signature matrix of an individual contains the following values:

[0139] like and but:

[0140]

[0141] Assume that the air pollution index increases significantly to , the system will recalculate the dementia risk characteristic value in real time:

[0142]

[0143] This indicates an increased risk. Through the above processing, the system generates an accurate feature matrix based on individual characteristics, providing high-quality input for subsequent risk index calculations.

[0144] Preferably, the causal reasoning module optimizes the feature matrix in combination with the association rules, and is implemented by the following steps:

[0145] Use the Apriori algorithm to mine association rules between features from latent variable data;

[0146] Adjust the feature matrix according to the mined rules and weight the association rules to causal weights;

[0147] The adjusted feature matrix is ​​regenerated by combining the causal weights and the association weights.

[0148] In this invention, the causal inference module not only calculates causal weights but also optimizes the feature matrix by combining association rules, thereby further improving the feature matrix's ability to represent latent variable relationships. Association rules exploit frequent patterns in latent variable data and the correlations between features, supplementing nonlinear or high-order dependencies that causal weights cannot fully capture. The optimized feature matrix combines causal weights with association rule weights to generate more accurate and comprehensive input data, providing more reliable support for subsequent risk assessment.

[0149] Association rule mining uses the Apriori algorithm to identify frequent item sets and association rules from latent variable data. The core of the Apriori algorithm is to discover strong association rules between features by gradually expanding frequent item sets. For example, certain genetic variants and specific lifestyle habits (such as smoking or a high-fat diet) may jointly have a significant impact on dementia risk, but this rule may not be directly expressed using a single variable.

[0150] By setting the minimum support threshold, we can filter out the frequent combinations that appear in the latent variable features. For example:

[0151] Gene and behavioral characteristics If it appears in 70% of the data records at the same time, its support is 0.7, which meets the minimum support threshold.

[0152] Generate association rules based on frequent item sets and evaluate the strength of the rules by calculating confidence and lift. For example:

[0153] → The confidence level is 0.8 and the lift is 1.5, indicating that the two are strongly correlated.

[0154] The mined association rules are introduced into the feature matrix by assigning weights. The association rule weights reflect the strength of the rules and are combined with the causal weights to generate the optimized feature matrix:

[0155]

[0156] in, Represents the optimized weight, which integrates causal weight and association rule weight; represents causal weight, represents latent variable right causal contribution of Represents the association rule weight, which is calculated by confidence and lift; It represents the adjustment coefficient of the association weight, which is used to balance the contribution of causal weight and association weight.

[0157] Combining causal weights and association rule weights generates an optimized feature matrix, where each column represents the contribution of a latent variable feature to the overall risk. This optimized matrix, based on causal reasoning, further enhances the model's ability to express higher-order dependencies.

[0158] By combining causal weights and association rule weights, the feature matrix more comprehensively reflects the complex relationships between latent variable features, improving the model's ability to identify nonlinear dependencies. The optimized feature matrix enables the model to capture the impact of more latent features on dementia risk, thereby generating a more accurate risk index. The introduction of association rules effectively compensates for the limitations of causal reasoning for higher-order interaction patterns, making it particularly suitable for analyzing multimodal data.

[0159] In the embodiment, in a certain dementia risk prediction system, the specific application of the causal reasoning module combined with the association rule to optimize the feature matrix is ​​as follows:

[0160] Mining frequent item sets and association rules between features from latent variable data. The following example shows:

[0161] Gene , behavioral characteristics , environmental characteristics It appears in 60% of the data records, with a support of 0.6;

[0162] → , the confidence level is 0.8 and the lift is 1.3;

[0163] → , with a confidence level of 0.7 and a lift of 1.2.

[0164] The association rule weights are introduced into the feature matrix and the optimized weights are calculated using the formula. For example:

[0165] Raw causal weights: ; .

[0166] Association rule weight: ; .

[0167] Optimized weights:

[0168]

[0169] After integrating the causal weights and the association weights, the column eigenvalues ​​of the feature matrix are updated:

[0170]

[0171] Assumptions and ,but:

[0172]

[0173] The optimized feature matrix reflects more high-order correlations, making the model's prediction of the individual's dementia risk more accurate.

[0174] Preferably, the individualized risk index is calculated by the following formula:

[0175]

[0176] in, represents the individualized risk index, Characterized by The regression coefficient of is the latent variable eigenvalue, is the constant term of regression, and n is the total number of all characteristic variables involved in the data calculation process.

[0177] In this study, the personalized risk index (PRI) is a comprehensive assessment of each individual's likelihood of developing dementia, calculated based on latent variable characteristics and corresponding regression coefficients. Using a nonlinear regression model, the PRI integrates the influence of each latent variable characteristic into a quantified risk value, reflecting the individual's likelihood of developing the disease.

[0178] Regression coefficient and constant term It is obtained through training with historical data and can minimize the error between the predicted risk index and the actual observed value.

[0179] Each latent variable eigenvalue The contribution to the individualized risk index is given by its regression coefficient For example, high-weighted traits (such as the APOE4 gene) have a more significant impact on the risk index, while low-weighted traits (such as dietary preferences) contribute less. This weighting method allows the model to highlight the impact of key traits while retaining the combined effects of other minor factors.

[0180] The training process of the model includes the following steps: using historical multimodal data as input, latent variable features The feature matrix generated by the previous module; setting the actual risk label (such as clinical diagnosis results or medical scores) as the output; using nonlinear regression optimization algorithms (such as gradient descent) to adjust the regression coefficients and constant term , to minimize the error between the model prediction results and the actual risk labels.

[0181] The individualized risk index can be recalculated based on the updated latent variable features in real time, adapting to the dynamic changes in individual characteristics. For example, an increase in an individual's exercise frequency may reduce the risk index, and the system will automatically update the index value based on the new data.

[0182] In an embodiment, in a dementia risk prediction scenario, the latent variable characteristics of an individual include the following:

[0183] Feature data input:

[0184] APOE4 gene locus status ( ):1.2;

[0185] Daily steps ( ):0.8;

[0186] BMI ( ):0.6.

[0187] Regression coefficients:

[0188] The regression coefficient of APOE4 gene ( ):0.9;

[0189] The regression coefficient of daily steps ( ): -0.5 (negative values ​​indicate protective effects);

[0190] The regression coefficient of BMI ( ):0.7.

[0191] Constant term:

[0192] The regression constant term ( ):0.3.

[0193] Risk index calculation:

[0194]

[0195] Substitute specific values:

[0196]

[0197] The calculation results show that the individual's risk index is 1.4, which is at a medium risk level.

[0198] Assume that an individual increases their step count from 0.8 to 1.2 by increasing their exercise. The system will recalculate the risk index based on the new eigenvalue:

[0199]

[0200] After the update, the risk index dropped from 1.4 to 1.2, showing a significant effect of improved exercise on risk reduction.

[0201] like Figure 3 As shown, the intervention optimization module is used to build a reinforcement learning environment based on individualized risk index data. It generates dynamically adjusted intervention strategy data through an optimization algorithm based on a deep Q network. The optimization algorithm uses a reinforcement learning framework, combined with real-time updated risk indicators and intervention feedback, to optimize the intervention path to reduce the risk of dementia.

[0202] In this invention, the intervention optimization module generates dynamically adjusted intervention strategy data using a reinforcement learning model. This module uses individualized risk index data as input and, combined with real-time updated risk indicators and intervention feedback, optimizes intervention pathways to reduce an individual's dementia risk. The core implementation uses a reinforcement learning algorithm based on a deep Q-network (DQN) to select the optimal intervention strategy through simulation and feedback mechanisms.

[0203] The intervention optimization module transforms the dementia risk prediction problem into a reinforcement learning problem, using a reinforcement learning environment to describe individual states, possible intervention measures, and their effects.

[0204] DQN is a reinforcement learning algorithm based on Q-learning, combined with deep neural networks, for solving optimization problems in continuous state spaces. In the intervention optimization module, the core task of DQN is to learn the optimal intervention strategy, which is achieved through the following steps:

[0205] Initialize the weights of the deep neural network and randomly select actions for initial exploration;

[0206] use Greedy strategy, which balances between exploring new actions and exploiting existing strategies;

[0207] Update the strategy using the Q-value function;

[0208] Through multiple rounds of iterative network training, the Q value gradually converges and the optimal intervention strategy is generated.

[0209] The module updates the individual's status based on real-time data, recalculates the risk index, and optimizes intervention paths. For example, if an individual's risk index decreases due to increased exercise frequency, the module will deprioritize further exercise and instead recommend other interventions, such as improving diet.

[0210] Preferably, the reinforcement learning environment of the intervention optimization module includes the following parts:

[0211] State space, consisting of individualized risk index data and high-risk factor annotations;

[0212] Action sets, including dietary adjustments, exercise recommendations, and medications;

[0213] The reward function is calculated by the following formula:

[0214]

[0215] in, Represents the reward value, is the change in risk index before and after intervention, The economic and time costs of the intervention.

[0216] In this paper, the intervention optimization module's reinforcement learning environment is used to construct an adaptable decision-making framework. By organically combining the state space, action set, and reward function, it generates dynamically adjusted intervention strategy data. The module's design goal is to maximize the effect of intervention measures on reducing dementia risk while controlling economic and time costs.

[0217] State Space It is the core component of the reinforcement learning environment and is used to describe the individual's current health status and risk level. In this invention, the state space consists of individualized risk index and high-risk factor labels:

[0218]

[0219] in, represents the individualized risk index at the current moment; Represents the characteristic values ​​of high-risk factors (such as BMI, daily steps, gene mutation status, etc.). The design of the state space can fully reflect the individual's current health status and provide information support for the selection of intervention strategies.

[0220] The action set A defines the intervention measures that the system can take, including the following:

[0221] Dietary adjustments: such as reducing fat and sugar intake and increasing antioxidant intake;

[0222] Exercise suggestions: such as setting daily walking goals and increasing low- to moderate-intensity aerobic exercise;

[0223] Drug treatment: such as the use of anti-inflammatory drugs or smoking cessation aids.

[0224] Each intervention measure in the action set is highly correlated with health risk factors and can indirectly reduce the dementia risk index by adjusting the individual's high-risk factor characteristic values.

[0225] Reward Function The calculation formula used to quantify the effect of each intervention is:

[0226]

[0227] The design of the reward function can achieve multi-objective optimization, obtain positive rewards by reducing the risk index, and avoid negative benefits caused by high-cost intervention.

[0228] By accurately describing an individual's health status through a state space, the reinforcement learning environment can generate personalized intervention strategies, thus avoiding a "one-size-fits-all" approach. The real-time updating of the state space and reward function enables the system to rapidly respond to changes in an individual's health, generating real-time adjusted intervention strategies. Through the design of the reward function, the system can find the optimal balance between risk reduction and cost control, thereby improving the efficiency and feasibility of actual interventions. Each intervention measure in the action set directly targets high-risk factors. The reinforcement learning module optimizes the intervention path through multiple rounds of iterations, significantly reducing an individual's risk of dementia.

[0229] In the embodiment, the initial state of an individual is:

[0230] Risk Index : 2.0 (high risk);

[0231] High-risk factor characteristic values: BMI is 30 (obesity), daily step count is 1500 (low activity), and smoking status is current smoking.

[0232] Reinforcement learning environment initialization, state space initialization is ;

[0233] The action set includes the following interventions:

[0234] : Dietary modification (reducing calorie intake);

[0235] : Increase exercise (daily walking goal 8,000 steps);

[0236] : Quit smoking (using auxiliary medicine).

[0237] First intervention:

[0238] System selection action (Add exercise), the adjusted state becomes:

[0239] The risk index dropped to , the reward function calculates the reward value: Among them, the economic cost Indicates time investment.

[0240] Second intervention:

[0241] In the new state, the system selects an action (Dietary adjustment) to further optimize the status:

[0242] The risk index dropped to , the reward value is:

[0243] Based on the real-time updated state and reward function, the system, after multiple rounds of iteration, ultimately generated the optimal strategy path: Step 1: Increase exercise (daily walking goal: 8,000 steps); Step 2: Dietary adjustment (increasing dietary fiber intake); Step 3: Quit smoking (using supplemental medication). After these interventions, the individual's risk index dropped from 2.0 to 1.0, significantly reducing the risk of dementia.

[0244] Preferably, the optimization algorithm of the reinforcement learning adopts a deep Q network, and the training process includes the following steps:

[0245] Initialize the parameters of the reinforcement learning model and determine the initial state through individualized risk index data;

[0246] use Greedy strategies, which balance between exploring new interventions and exploiting existing ones;

[0247] The Q value is updated in multiple rounds of iterations, and the intervention path is continuously optimized through the reward function.

[0248] In this study, a reinforcement learning optimization algorithm uses a deep Q-network (DQN) to generate dynamically adjusted personalized intervention strategies. This algorithm aims to significantly reduce an individual's dementia risk by iteratively optimizing the intervention pathway through multiple rounds of training. By leveraging the high adaptability of reinforcement learning models, the algorithm strikes a balance between exploring the effectiveness of different interventions and optimizing existing strategies, thereby reducing individual risk indices while controlling intervention costs.

[0249] The reinforcement learning model uses individualized risk index data and high-risk factor labels as initial states :

[0250]

[0251] in, is the initial value of the individualized risk index, are the characteristic values ​​of high-risk factors related to risk (such as BMI, daily steps, etc.). These states describe the health status of the individual and provide the initial conditions for the optimization process.

[0252] To strike a balance between exploring new interventions (expanding the policy space) and exploiting existing policies (improving performance), DQN adopts Greedy strategy.

[0253] The exploration phase is based on probability Randomly select intervention actions , try new strategies;

[0254] Utilization stage with probability Select the current best action .

[0255] The decision formula is:

[0256]

[0257] This strategy ensures that the model explores more potential effective paths in the early stages and focuses on optimizing the existing best path in the later stages.

[0258] After each round of intervention, the Q value is updated based on the immediate feedback. The Q value update is achieved by the following formula:

[0259]

[0260] in, Indicates that the status Select Action the expected returns to be obtained; Represents the immediate reward, which is calculated by the following formula , is the change in risk index, The economic and time costs of intervention, represents the discount factor, which indicates the weight of future rewards; Represents the learning rate, which is used to control the step size of Q value update.

[0261] During reinforcement learning training, the system repeatedly simulates different intervention strategy combinations and adjusts the Q value through reward feedback. As training rounds increase, the Q value gradually converges, ultimately generating the optimal intervention strategy path, significantly reducing the individual risk index at the lowest cost.

[0262] Through real-time state updates, DQN can automatically adjust intervention strategies based on individual health changes, making decision-making more flexible and accurate. By comprehensively considering risk reduction and intervention costs through the reward function, it ensures that the optimized path significantly reduces risk while minimizing resource consumption. The introduction of a greedy strategy enables the model to explore more possible strategic paths early on while ensuring it focuses on the optimal path later on, improving overall optimization efficiency. Through multiple rounds of training and iterative optimization, the resulting intervention pathways can significantly reduce individual risk indices, thereby enhancing the scientific and practical nature of health management.

[0263] In an embodiment, in a dementia risk prediction application, the initial state of an individual is:

[0264] Individualized risk index : 2.0 (high risk);

[0265] High-risk factor characteristics: BMI is 30 (obesity), daily step count is 1500 (low activity), and smoking status is current smoking.

[0266] Reinforcement learning optimization process:

[0267] Initial state .

[0268] Action Collection: : Adjust your diet; : Increase exercise; : Quit smoking.

[0269] Round 1 intervention:

[0270] use Greedy strategy selection action (Increase exercise).

[0271] After adjustment, the number of steps per day increased to 8,000, and the risk index decreased to .

[0272] Instant Rewards: , where intervention cost Indicates the time investment cost.

[0273] Round 2 intervention:

[0274] New Status . Select Action (Dietary adjustments).

[0275] After adjusting for BMI down to 28, the risk index was further reduced to .

[0276] Instant Rewards:

[0277] Round 3 intervention:

[0278] New Status . Select Action (Quit smoking).

[0279] After adjustment, the smoking status becomes 0 (successful smoking cessation), and the risk index is reduced to .

[0280] Instant Rewards:

[0281] After multiple rounds of training, the model generated the optimal strategy path: Step 1: Increase exercise; Step 2: Adjust diet; Step 3: Quit smoking. The final result: The risk index dropped from 2.0 to 1.0, a 50% reduction in risk.

[0282] Preferably, the long-term risk prediction task of the prediction and generation module is implemented through a recurrent neural network, the input features include the individualized risk index and high-risk factor labels in the intervention strategy data after dynamic adjustment, and the output is a risk curve in the future time.

[0283] In this paper, the long-term risk prediction task of the prediction and generation module is implemented using a recurrent neural network (RNN). Its core goal is to predict individualized risk trends in the future based on dynamically adjusted intervention strategy data. With its powerful time series modeling capabilities, RNN captures the dynamic correlation between individualized risk indices and high-risk factor characteristics, providing scientific predictions for long-term health management.

[0284] The input features of the long-term risk prediction task include the individualized risk index and high-risk factor labels in the dynamically adjusted intervention strategy data, specifically in the form of a time series:

[0285]

[0286] in, Indicates the current time Individualized risk index; Represents high-risk factor annotations (such as BMI, daily steps, smoking status, etc.). The input features capture the dynamic changes in individual health status and intervention measures over time, providing the necessary information for future risk prediction.

[0287] RNN processes time series data with its recursive network structure, through hidden states Record historical information and predict future risk curves. The update formula for the hidden state is:

[0288]

[0289] in, Indicates the current time The hidden state of Represents the weight matrix from hidden state to hidden state; Represents the weight matrix input to the hidden state; Indicates the current input features (including and ); represents the bias term; Represents the activation function (such as tanh or ReLU). The output prediction value is the future risk curve:

[0290]

[0291] The model aims to minimize the error between the predicted risk curve and the true value, and uses the mean square error (MSE) as the loss function:

[0292]

[0293] RNN optimizes parameters through backpropagation and gradient descent methods 、 and , in order to improve the accuracy of prediction.

[0294] RNNs can capture the dynamic patterns of individualized risk indices and high-risk factor labels over time, generating future risk curves and providing forward-looking support for health management. The module combines an individual's current health status and intervention history to generate tailored long-term risk predictions for each individual, avoiding a one-size-fits-all approach. By predicting future risk changes, the module provides a quantitative basis for developing proactive intervention strategies to mitigate long-term risks. RNN's time-dependent modeling capabilities enable predictions to reflect changing risk trends, providing comprehensive information support for decision-making.

[0295] In the example, the initial state of an individual is as follows:

[0296] Current Risk Index =2.0; high-risk factor characteristic values: BMI is 30, daily step count is 2000, and smoking status is current smoking.

[0297] Long-term risk prediction process: The input sequence includes past individualized risk indices and high-risk factor annotations:

[0298]

[0299] in, =2.0 indicates the initial risk index; =1.8 indicates the updated risk index after the intervention.

[0300] Through the recursive calculation of RNN, the temporal dependency between input features is captured and the future risk curve is predicted: , assuming the output is: , indicating that with continuous intervention, the risk index will gradually decrease.

[0301] Use historical data to verify model performance by optimizing the loss function:

[0302]

[0303] Adjusting network weights 、 and bias , making the prediction results more accurate.

[0304] like Figure 4 As shown, the prediction and generation module is used to perform long-term risk prediction, short-term trend analysis and individual stratification screening based on the dynamically adjusted intervention strategy data, and generate personalized prediction report data, which includes risk assessment and intervention recommendations.

[0305] Preferably, the personalized prediction report includes the following contents:

[0306] Summary of dementia risk assessment; long-term trend analysis results, including changes in the risk curve; short-term trend analysis results; individual stratification results and key high-risk factors; and specific intervention recommendations for individuals.

[0307] In this invention, the prediction and generation module comprehensively analyzes dynamically adjusted intervention strategy data to perform long-term risk prediction, short-term trend analysis, and individual stratification screening, ultimately generating personalized prediction reports. This module not only comprehensively assesses an individual's current health status but also predicts future risk trends, providing a scientific basis for the development of intervention measures. The core goal of this module design is to achieve personalized, dynamic, and forward-looking health management.

[0308] Based on the individualized risk index and high-risk factor characteristics from the dynamically adjusted intervention strategy data, we predict future risk trends. Input features include the individual's current risk index and historical dynamic intervention data, such as daily step count and BMI changes. Output: A future risk curve, showing the individual's risk trend over a specific timeframe (e.g., six months or a year). This provides data support for preemptive intervention planning. For example, if the risk curve indicates a potential increase in risk over the next three months, a more rigorous health management plan can be developed in advance.

[0309] Short-term trend analysis focuses on changes in an individual's health following a recent intervention. It aims to assess the effectiveness of the intervention by analyzing the magnitude and direction of recent changes in risk indices. Input features include: recent changes in risk indices and high-risk factor characteristics (e.g., changes in daily step count and dietary intake over the past month). Output: Short-term trend analysis results, including the rate of risk change and an assessment of intervention effectiveness. For example, if an individual recently increased their exercise level and the rate of risk change decreased, the system could provide feedback indicating the positive effects of the exercise intervention.

[0310] Based on individualized risk indices and high-risk factor characteristics, individuals are stratified for screening (e.g., low-risk, medium-risk, and high-risk groups). Input characteristics: Risk index and specific high-risk factor labels. Output: Stratified screening results and a list of key high-risk factors. This provides guidance for allocating medical resources and prioritizing interventions. For example, high-risk individuals may require additional medical resources, while low-risk individuals can maintain existing interventions.

[0311] Based on the results of long-term risk prediction, short-term trend analysis and individual stratification screening, a personalized prediction report is generated. The report content includes: a summary of dementia risk assessment; long-term trend analysis results, including changes in the risk curve; short-term trend analysis results; individual stratification results and a list of key high-risk factors; and specific intervention recommendations for individuals (such as adjusting diet, increasing exercise, drug treatment, etc.).

[0312] Through long-term forecasting and short-term trend analysis, the module provides scientific risk assessment and intervention effectiveness evaluation tools, making health management more precise and efficient. Personalized forecast reports provide tailored health management plans for each individual, enhancing the relevance of intervention measures. Through individual stratified screening, the module prioritizes medical resources based on risk level, providing more support for high-risk individuals. Short-term trend analysis provides real-time insights into intervention effectiveness, providing a basis for adjusting intervention strategies.

[0313] In this embodiment, after an individual receives the dementia risk prediction system service, the system's prediction and generation module generates the following results:

[0314] Input data, dynamically adjusted intervention strategy data: current individualized risk index =BMI was 28 (high); daily step count was 4000 (low activity); current smoking status. Intervention history data: step count increased from 2000 to 4000 and BMI decreased from 30 to 28 over the past month.

[0315] The long-term risk prediction module predicts the risk curve for the next six months based on the input characteristics: after one month, the risk index decreases to 1.6; after three months, the risk index stabilizes at 1.4; and after six months, the risk index further decreases to 1.2. The prediction results show that with existing intervention measures, the risk is expected to gradually decrease.

[0316] Short-term trend analysis showed that the risk index decreased from 2.0 to 1.8 over the past month, a 10% decrease. An increase in step count contributed the most to the risk reduction, reaching 50%, followed by dietary adjustments at 30%, and a decrease in BMI at 20%. These results indicate that exercise intervention is currently the most effective measure.

[0317] Individuals are stratified into the "medium-risk" group based on their current risk index and high-risk factor profile. Key high-risk factors include BMI, step count, and smoking status. Screening results recommend further BMI reduction and smoking cessation.

[0318] Generate personalized forecast reports. The reports generated by the system include the following:

[0319] Risk Assessment Summary: The individual is currently at medium risk, with a risk index of 1.8.

[0320] Long-term trend analysis: The risk index will gradually decrease to 1.2 in the next 6 months, indicating that existing intervention measures are relatively effective.

[0321] Short-term trend analysis: The risk index decreased by 10% in the past month, mainly due to exercise intervention.

[0322] Stratification results and key high-risk factors: medium-risk group, high-risk factors are BMI, number of steps and smoking status.

[0323] Intervention recommendations: Increase the number of steps per day to 6,000, further improve diet (reduce high-fat intake), and use medication to assist in quitting smoking.

[0324] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A dementia risk prediction system based on an associative information processing model, characterized in that: include: a data processing module for collecting multimodal data, including genetic data, health data, behavioral data, and environmental data, and generating initial latent variable data based on the multimodal data; dividing the behavioral data and environmental data into time windows of fixed length, independently calculating features for the data in each time window, calculating the mean for the data in each time window, generating a mean parameter in the trend feature, calculating the rate of change of the means in adjacent time windows, generating a rate of change parameter in the trend feature, applying linear regression analysis to the data in the time window, calculating the trend slope parameter, generating a slope parameter in the trend feature, and updating the initial latent variable data based on the extracted trend features to obtain updated latent variable data; A causal inference module is used to construct a causal model based on the updated latent variable data, calculate causal weights, and generate an individualized risk signature matrix, where the signature matrix is ​​generated by comprehensively analyzing the weighted contributions of different causal relationship variables; and to calculate individualized risk index data based on the signature matrix using a nonlinear regression algorithm to reflect the likelihood of dementia for each individual. The individualized risk signature matrix is ​​calculated by the following steps: Determine the causal weight and calculate the causal weight using the following formula: ,in, Representation characteristics Features The causal weight of Characterized by and features The covariance of Characterized by variance; The eigenvalues ​​in the feature matrix are calculated based on the causal weights, which is achieved through the following formula: ,in, is the eigenvalue of the matrix, Representation characteristics In the characteristic matrix, the eigenvalues The contribution weight of is the latent variable feature, n is the total number of latent variable features, that is, the number of different latent variable features considered in the feature matrix; The causal reasoning module optimizes the feature matrix in combination with association rules, which is achieved by the following steps: using the Apriori algorithm to mine association rules between features from latent variable data; adjusting the feature matrix based on the mined rules, and weighting the weights of the association rules to causal weights; and regenerating the adjusted feature matrix by combining the causal weights and association weights; The individualized risk index is calculated by the following formula: ,in, represents the individualized risk index, Characterized by The regression coefficient of is the latent variable eigenvalue, is the constant term of the regression, and n is the total number of latent variable characteristics considered in the individualized risk assessment; An intervention optimization module, which builds a reinforcement learning environment based on individualized risk index data and generates dynamically adjusted intervention strategy data through a deep Q-network-based optimization algorithm. The optimization algorithm uses a reinforcement learning framework, combined with real-time updated risk indicators and intervention feedback, to optimize intervention pathways to reduce dementia risk; The prediction and generation module is used to perform long-term risk prediction, short-term trend analysis and individual stratification screening based on the dynamically adjusted intervention strategy data, and generate personalized prediction report data. The report includes risk assessment and intervention recommendations. The long-term risk prediction task is implemented through a recurrent neural network. The input features include the individualized risk index and high-risk factor labels in the dynamically adjusted intervention strategy data, and the output is the risk curve in the future time.

2. The dementia risk prediction system based on the association information processing model according to claim 1, characterized in that: The data processing module includes a missing data filling unit, which processes missing values ​​in the multimodal data by the following steps: For continuous numerical data, mean interpolation is used to fill missing values ​​based on the mean of other data records; For categorical variables, the nearest neighbor algorithm-based imputation method is used to select the most common category value in similar data records for imputation; The padded multimodal data generates initial latent variable data.

3. The dementia risk prediction system based on the association information processing model according to claim 1, characterized in that: The reinforcement learning environment of the intervention optimization module consists of the following components: State space, consisting of individualized risk index data and high-risk factor annotations; Action sets, including dietary adjustments, exercise recommendations, and medications; The reward function is calculated by the following formula: , where R represents the reward value, is the change in risk index before and after intervention, and C is the economic cost and time cost of the intervention measures.

4. The dementia risk prediction system based on the association information processing model according to claim 1, characterized in that: The optimization algorithm for reinforcement learning uses a deep Q network, and the training process includes the following steps: Initialize the parameters of the reinforcement learning model and determine the initial state through individualized risk index data; use Greedy strategies, which balance between exploring new interventions and exploiting existing ones; The Q value is updated in multiple rounds of iterations, and the intervention path is continuously optimized through the reward function.

5. The dementia risk prediction system based on the association information processing model according to claim 1, characterized in that: The personalized prediction report includes the following contents: Dementia Risk Assessment Summary; Long-term trend analysis results, including changes in the risk curve; short-term trend analysis results; individual stratification results and key high-risk factors; and specific intervention recommendations for individuals.

Images