Blood vessel plaque prediction system based on physical examination big data

Abstract

The invention relates to the technical field of health management, in particular to a blood vessel plaque prediction system based on health examination big data. Comprising a multi-modal data acquisition module, a data preprocessing and fusion module, a feature selection and risk factor mining module, a plaque prediction model construction module, a personalized health intervention suggestion module and a visualization and decision support module. The XGBoost algorithm is used for screening key risk factors and performing importance ranking, a plaque risk prediction model is constructed based on a long-short-term memory network, the risk probability of future plaque formation is generated, and the personalized health intervention suggestion module provides diet, exercise and medicine intervention suggestions according to a prediction result. The visualization module displays a risk assessment result and an intervention scheme through a graphical interface; according to the method, accurate prediction and personalized management of the vascular plaque risk are realized, and the method has relatively high clinical application value.

Description

Technical Field

[0001] The present invention relates to the field of health management technology, and in particular to a vascular plaque prediction system based on health examination big data. Background Art

[0002] Vascular plaque is an important cause of cardiovascular and cerebrovascular diseases. Its formation is related to multiple factors, including dyslipidemia, unhealthy lifestyle, and changes in vascular wall structure. Early detection and intervention of vascular plaque is the key to preventing cardiovascular and cerebrovascular diseases.

[0003] Existing technologies mainly focus on a single data source or local analysis, such as plaque risk assessment based on blood lipid indicators or imaging features, which cannot fully cover the multidimensional factors that affect plaque formation. In addition, traditional prediction methods mostly rely on linear models or simple decision tree algorithms, which make it difficult to capture the complex nonlinear relationships and temporal dynamic characteristics between multimodal data, resulting in insufficient accuracy and generalization of prediction results. At the same time, existing technologies lack systematicity in personalized intervention, and most of them only provide general health advice, failing to fully combine users' dynamic health data and key risk factors for precise intervention. Summary of the invention

[0004] The present invention provides a vascular plaque prediction system based on health examination big data.

[0005] The vascular plaque prediction system based on health examination big data includes multimodal data acquisition module, data preprocessing and fusion module, feature selection and risk factor mining module, plaque prediction model construction module, personalized health intervention suggestion module and visualization and decision support module, among which;

[0006] The multimodal data acquisition module is used to collect health-related multimodal data from multiple data sources;

[0007] The data preprocessing and fusion module is used to preprocess the collected multimodal data, including data cleaning and standardization, and use a deep belief network algorithm to extract features from the preprocessed multimodal data, and fuse the extracted features into a unified feature vector;

[0008] The feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, extracts key risk factors related to vascular plaque formation, and sorts the extracted key risk factors by calculating the feature importance score, and analyzes the correlation between the risk factors. The key risk factors include biochemical risk factors, such as marker lipoprotein (a) [Lp(a)], low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), fasting blood sugar (FBS), lifestyle risk factors (such as smoking frequency, lack of exercise) and vascular imaging risk factors, such as vascular intima-media thickness (IMT);

[0009] The plaque prediction model building module builds a vascular plaque risk prediction model based on a long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future;

[0010] The personalized health intervention suggestion module generates personalized intervention suggestions according to the risk probability output by the vascular plaque risk prediction model, and the suggestions include a diet adjustment plan, an exercise intervention plan, and a drug intervention suggestion;

[0011] The visualization and decision support module displays prediction results, importance distribution of key risk factors and personalized intervention recommendations through a graphical user interface, and provides risk assessment reports for medical practitioners.

[0012] Optionally, the multimodal data acquisition module includes:

[0013] Physical examination report data collection: obtain the user's historical and current physical examination report data through the hospital information system interface;

[0014] Test data collection: collect users' test data;

[0015] Lifestyle data collection: Users input their daily diet structure, exercise frequency, smoking and drinking records through mobile applications or wearable devices;

[0016] Vascular imaging data acquisition: Obtain the user's vascular ultrasound image, CT image and MRI image data through medical imaging equipment;

[0017] Data timestamp and labeling step: Add a timestamp to each piece of collected multimodal data to ensure the time series integrity of the data, and add labels to the data according to the data type and source.

[0018] Optionally, the data preprocessing and fusion module includes:

[0019] Data cleaning: Clean the collected multimodal data, remove outliers, fill in missing values, and standardize the data format to ensure data integrity and consistency;

[0020] Data standardization: Standardize the cleaned multimodal data.

[0021] Optionally, the data preprocessing and fusion module uses a deep belief network algorithm to extract features from the preprocessed multimodal data, including:

[0022] Deep belief network model construction: A deep belief network model is constructed to extract deep features of multimodal data layer by layer. The deep belief network model consists of multiple layers of restricted Boltzmann machines, and the model is optimized through unsupervised pre-training and global fine-tuning.

[0023] Data input: input the preprocessed multimodal data into the constructed deep belief network model for feature extraction;

[0024] Feature fusion: The multimodal feature vectors output by the deep belief network model are fused to generate a unified feature representation.

[0025] Optionally, the feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, including:

[0026] Model construction: A feature selection model is constructed based on the XGBoost algorithm to measure the contribution of each feature to the prediction result by fitting the target variable and to evaluate the importance of the feature;

[0027] Data input: Input the feature vector into the constructed feature selection model;

[0028] Feature importance calculation: Utilize the feature importance evaluation mechanism of the feature selection model to calculate the gain score of each input feature based on the feature selection model;

[0029] Screening features: Sort the features according to the gain score and select the features that contribute most to the prediction of the target variable, such as markers Lp(a), LDL-C, HDL-C, FBS, and IMT values, as key risk factors.

[0030] Optionally, the feature selection and risk factor mining module sorts and performs correlation analysis on the extracted key risk factors, including:

[0031] Ranking of key risk factors: The extracted key risk factors are ranked according to their contribution to the risk of vascular plaque formation;

[0032] Risk factor correlation analysis: Analyze the correlation between key risk factors and reveal the potential interactions between multiple factors.

[0033] Optionally, the plaque prediction model building module includes:

[0034] Model construction: Construct a plaque risk prediction model based on the long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future;

[0035] Model training: The feature vectors extracted by the feature selection and risk factor mining module are input into the constructed plaque risk prediction model for training to optimize the model parameters;

[0036] Model validation: After the plaque risk prediction model training is completed, the plaque risk prediction model is evaluated;

[0037] Model deployment: The trained plaque risk prediction model is deployed into the system to predict the risk probability of future vascular plaque formation in real time.

[0038] Optionally, the personalized health intervention suggestion module includes:

[0039] Risk grading: Based on the output probability of the plaque risk prediction model, users are divided into different risk levels, including low risk, medium risk and high risk, providing a basis for formulating personalized intervention plans;

[0040] Diet adjustment plan generation: Develop a diet adjustment plan based on the user’s risk level;

[0041] Exercise intervention plan generation: Generate an exercise intervention plan based on the user's risk level and lifestyle data;

[0042] Drug intervention recommendation generation: For high-risk users, drug intervention recommendations are generated based on vascular imaging data and physical examination report data;

[0043] Personalized report generation: Integrate the generated intervention recommendations into a user-specific personalized health management report.

[0044] Optionally, the visualization and decision support module includes:

[0045] Visualization of prediction results: Display the output of the plaque risk prediction model through a graphical user interface;

[0046] Importance distribution display of key risk factors: Visual display of key risk factors extracted by feature selection and risk factor mining modules;

[0047] Visual display of personalized intervention recommendations: Diet, exercise and drug intervention recommendations are presented to users through a graphical interface;

[0048] Generation and presentation of risk assessment reports: Provide risk assessment reports to medical practitioners to assist in clinical decision-making.

[0049] Beneficial effects of the present invention:

[0050] The present invention, through the multimodal data acquisition module and the data preprocessing and fusion module, comprehensively integrates the physical examination report data, test data, living habits data and vascular imaging data, uses the deep belief network to extract high-level features, and uses the XGBoost algorithm to screen out the key risk factors that are highly correlated with the formation of vascular plaques, which significantly improves the efficiency of data processing and the accuracy of prediction. The plaque prediction model based on the long short-term memory network can capture the dynamic characteristics of time series data, realize the efficient prediction of future risks, and provide reliable support for dynamic risk monitoring and precise management.

[0051] The present invention generates dietary adjustment, exercise plan and drug intervention suggestions based on the output results of the plaque prediction model and the user's dynamic health data through a personalized health intervention suggestion module. By analyzing dynamic data such as markers Lp(a), LDL-C, HDL-C, FBS, IMT value and exercise frequency, a highly operational health intervention plan is provided, and the drug use plan is optimized in combination with the detection information to ensure the scientificity, safety and personalization of the intervention measures, effectively helping users reduce the risk of vascular plaque formation and improve overall health.

[0052] In this invention, the visualization and decision support module displays plaque risk prediction results, the importance distribution of key risk factors, and personalized intervention suggestions through a graphical user interface, and generates a comprehensive assessment report to provide users with dynamic health risk trends and intervention execution guidance. The system provides medical practitioners with detailed risk assessment reports, supports real-time adjustment of diagnosis and treatment plans, improves clinical decision-making efficiency, and comprehensively enhances the scientificity and convenience of health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 A schematic diagram of a system flow of an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of the feature selection and risk factor mining module flow in an embodiment of the present invention;

[0056] Figure 3 The characteristic curve for predicting the risk of vascular plaque occurrence in the present invention;

[0057] Figure 4 A calibration curve for predicting the risk of vascular plaque occurrence in the present invention;

[0058] Figure 5 Decision analysis curve for predicting the risk of vascular plaque development. DETAILED DESCRIPTION

[0059] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0060] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0061] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0062] like Figure 1-Figure 5 As shown, the vascular plaque prediction system based on health examination big data includes a multimodal data acquisition module, a data preprocessing and fusion module, a feature selection and risk factor mining module, a plaque prediction model construction module, a personalized health intervention suggestion module and a visualization and decision support module, among which;

[0063] The multimodal data acquisition module is used to collect health-related multimodal data from multiple data sources. The multimodal data includes physical examination report data, test data, life habit data and vascular imaging data, among which;

[0064] The data preprocessing and fusion module is used to preprocess the collected multimodal data, including data cleaning and standardization, and uses the deep belief network algorithm to extract features from the preprocessed multimodal data, and fuse the extracted features into a unified feature vector;

[0065] The feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, extracts the key risk factors related to vascular plaque formation, and sorts the extracted key risk factors by calculating the feature importance score, and analyzes the correlation between the risk factors. The key risk factors include biochemical risk factors (such as markers Lp(a), LDL-C, HDL-C, FBS), lifestyle risk factors (such as smoking frequency, lack of exercise) and vascular imaging risk factors (such as IMT);

[0066] The plaque prediction model building module builds a vascular plaque risk prediction model based on the long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future;

[0067] The personalized health intervention suggestion module generates personalized intervention suggestions based on the risk probability output by the vascular plaque risk prediction model. The suggestions include diet adjustment plan, exercise intervention plan and drug intervention suggestions.

[0068] The visualization and decision support module displays prediction results, importance distribution of key risk factors and personalized intervention recommendations through a graphical user interface, and provides risk assessment reports for medical practitioners.

[0069] The multimodal data acquisition module includes:

[0070] Physical examination report data collection: obtain the user's historical and current physical examination report data through the hospital information system interface. The physical examination report data includes blood lipid indicators;

[0071] Data collection includes:

[0072] Diet structure: daily intake ratio of total calories, fat, protein and carbohydrates;

[0073] Exercise frequency: the number of times you exercise per week and the duration of each exercise;

[0074] Smoking records: daily smoking amount and cumulative smoking time;

[0075] Drinking records: number of drinks per week and type of drinks (e.g., beer, wine, spirits);

[0076] Vascular imaging data acquisition: Obtain the user's vascular ultrasound image, CT image and MRI image data through medical imaging equipment;

[0077] Data timestamp and labeling step: Add a timestamp to each piece of collected multimodal data to ensure the time series integrity of the data, and add labels to the data according to the data type and source (such as "physical examination report data", "test data");

[0078] The multimodal data acquisition module provides the system with a high-quality, structured health-related data foundation through the comprehensive collection of physical examination report data, test data, lifestyle data and vascular imaging data, ensuring the diversity and integrity of the data and providing solid input support for subsequent modules.

[0079] The data preprocessing and fusion modules include:

[0080] Data cleaning: Clean the collected multimodal data, remove outliers, fill in missing values, and standardize the data format to ensure the integrity and consistency of the data, including:

[0081] (1) Clean the physical examination report data:

[0082] (1.1) Detect outliers: Use the triple standard deviation method to eliminate outlier data points that are beyond the reasonable range;

[0083] (1.2) Filling missing values: For missing data of blood glucose, blood lipids and BMI, mean filling method was used to reduce data bias;

[0084] (2) Clean the test data:

[0085] (3) Identify incomplete records: delete lifestyle data entries that do not contain key fields, such as entries without drinking records;

[0086] (4) Cleaning of vascular image data:

[0087] (4.1) Detect image defects: Use pixel histogram method to mark low-quality images;

[0088] (4.2) Repair image data: Use bilinear interpolation algorithm to fill in damaged pixel values;

[0089] Data standardization: Standardize the cleaned multimodal data to ensure that different data types have consistent scales and distributions, including:

[0090] Standardize lifestyle data: discretize exercise frequency, drinking frequency, etc. into level data, for example, "high, medium, low" correspond to the values ​​"2, 1, 0";

[0091] Standardize the vascular imaging data: normalize the pixel values ​​and adjust the pixel value range of the imaging data to 0,1;

[0092] The data cleaning and standardization steps ensure the consistency of multimodal data quality by removing outliers, filling missing values, and standardizing data formats and scales, providing a high-quality data foundation for subsequent feature extraction.

[0093] The data preprocessing and fusion module uses a deep belief network algorithm to extract features from the preprocessed multimodal data, including:

[0094] Deep belief network model construction: Build a deep belief network model to extract deep features of multimodal data layer by layer. The deep belief network model consists of multiple layers of restricted Boltzmann machines. The model is optimized through unsupervised pre-training and global fine-tuning, including:

[0095] (1) Define the model structure: set the input layer dimension. The input data includes the feature vectors of physical examination report data, test data, lifestyle data, and vascular imaging data.

[0096] Define the number of hidden layers and neurons, and the number of hidden units in each layer decreases layer by layer to capture high-level abstract features;

[0097] (2) Initialization parameters: Randomly initialize the model’s weight matrix W and bias values ​​b, c;

[0098] (3) Global fine-tuning: Use the back-propagation algorithm to perform supervised fine-tuning on all RBM layers to ensure the global optimality of the extracted features;

[0099] Data input: The preprocessed multimodal data is input into the constructed deep belief network model for feature extraction, including:

[0100] (1) Input of physical examination report data: standardized continuous variables such as Lp(a), LDL-C, HDL-C, and FBS were directly input into the model;

[0101] (2) Input of test data: One-hot encoded information is input as discrete variables;

[0102] (3) Input of lifestyle data: discretized data such as diet structure and exercise frequency are used as time series feature input;

[0103] (4) Input of vascular imaging data: using the principal component feature vector extracted by principal component analysis (PCA) as input;

[0104] Feature fusion: The multimodal feature vectors output by the deep belief network model are fused to generate a unified feature representation, including:

[0105] Save feature vector: store the fused feature vector in the database to provide input for subsequent modules;

[0106] By constructing a deep belief network model and inputting multimodal data to complete feature extraction and weighted fusion, this module can effectively extract high-level features of multimodal data and generate a unified feature vector, providing comprehensive and accurate data support for subsequent risk factor mining and model building.

[0107] The feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, including:

[0108] Model construction: A feature selection model is constructed based on the XGBoost algorithm to measure the contribution of each feature to the prediction result by fitting the target variable and to evaluate the importance of the feature, including:

[0109] (1) Define the target variable: Set the target variable y to represent the actual value of the risk of vascular plaque formation (e.g., whether there is a high risk, with a value range of 0 or 1);

[0110] (2) Model construction: A feature selection model is constructed based on the XGBoost algorithm, and a loss function is defined, which is expressed as:

[0111]

[0112] in, is the loss between the predicted value and the true value of the i-th sample, Ω(f k ) is a regularization term that controls the model complexity, and K is the total number of decision trees;

[0113] Use the gradient boosting method to optimize the model parameters and gradually generate multiple weak learners (i.e. decision trees);

[0114] Data input: Input the feature vector into the constructed feature selection model, including:

[0115] (1) Input of physical examination report data: standardized continuous variables such as Lp(a), LDL-C, HDL-C, and FBS are provided as input features;

[0116] (2) Input of test data: provide one-hot encoding information as input features;

[0117] (3) Input of lifestyle data: provide discretized behavioral features such as exercise frequency and smoking frequency as input features;

[0118] (4) Input of vascular imaging data: providing dimensionality reduction features of imaging data (such as IMT values) as input features;

[0119] Feature importance calculation: Using the feature importance evaluation mechanism of the feature selection model, based on the feature selection model, the gain score of each input feature is calculated, including:

[0120] Screening features: Sort features according to gain scores and select features that contribute most to the prediction of the target variable, such as markers Lp(a), LDL-C, and IMT values, as key risk factors;

[0121] The feature selection and risk factor mining module constructs a feature selection model through the XGBoost algorithm to automatically identify key features that have a significant impact on vascular plaque formation, laying the foundation for subsequent risk factor analysis and prediction model construction.

[0122] The feature selection and risk factor mining module sorts and analyzes the correlation of the extracted key risk factors, including:

[0123] Ranking of key risk factors: The extracted key risk factors are ranked according to their contribution to the risk of vascular plaque formation, including:

[0124] Calculate the feature contribution ratio: According to the feature gain score, calculate the contribution ratio of each risk factor, expressed as:

[0125]

[0126] Among them, Gain i is the gain score of the ith risk factor, is the sum of all risk factor gain scores, and m is the total number of risk factors;

[0127] Generate ranking results: sort from high to low according to contribution ratio;

[0128] Risk factor correlation analysis: Analyze the correlation between key risk factors and reveal the potential interactions between multiple factors, including:

[0129] Calculate the correlation matrix: The Pearson correlation coefficient is used to calculate the correlation between risk factors, expressed as:

[0130]

[0131] Among them, r ij represents the correlation between the i-th and j-th risk factors, X ik and X jk is the value of the risk factor in the kth sample, and is the mean;

[0132] Analyze the correlation results: mark the highly correlated combinations between risk factors, for example, the markers Lp(a), LDL-C and IMT values ​​may be highly positively correlated, and the smoking frequency and lack of exercise may be negatively correlated. Visualize the correlation matrix and generate a heat map to intuitively display the relationship between the risk factors;

[0133] Through the ranking and correlation analysis of key risk factors, the system can not only reveal the relative importance of each factor, but also analyze the combined impact of multiple factors on vascular plaque formation, providing a scientific basis for personalized intervention and precise optimization of prediction models.

[0134] The patch prediction model building modules include:

[0135] Model construction: Construct a plaque risk prediction model based on the long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future, including:

[0136] (1) Define the model structure:

[0137] (1.1) Input layer: Set the input layer dimension to be consistent with the extracted features, including key risk factors such as markers Lp(a), LDL-C, HDL-C, FBS, IMT values ​​and their time series data;

[0138] (1.2) Hidden layer: construct several long short-term memory network units, each unit contains a forget gate, an input gate and an output gate to capture time series features;

[0139] (1.3) Output layer: Set the output as the risk probability of future vascular plaque formation, ranging from 0 to 1;

[0140] (2) Initialize model parameters: Randomly initialize the weight matrix and bias value to ensure the stability of the initial state of model training;

[0141] Model training: The feature vectors extracted by the feature selection and risk factor mining module are input into the constructed plaque risk prediction model for training and optimization of model parameters, including:

[0142] (1) Input data preparation: Input features include the following optimized key risk factors:

[0143] Time series of physical examination report data: such as historical changes in markers Lp(a), LDL-C, HDL-C, FBS, and blood pressure;

[0144] Detect key features of data;

[0145] Time series of lifestyle data: such as exercise frequency, drinking frequency, and smoking records;

[0146] Time series of vascular imaging data: such as the dynamic changes of IMT values;

[0147] (2) Define the calculation formula of long short-term memory network unit:

[0148] (2.1) Forget gate: The calculation formula is expressed as:

[0149] f t =σ(W f ·[h t―1 ,x t ]+b f );

[0150] Among them, x t is the current input, h t―1 is the hidden state of the previous moment, W f is the weight matrix, b f is the bias value, σ is the activation function;

[0151] (2.2) Input gate: The calculation formula is expressed as:

[0152] i t =σ(W i ·[h t―1 ,x t ]+b i ),

[0153] Among them, i t is the input gate state, is the candidate memory content, W i , W C and b i , b C are weight matrix and bias value respectively;

[0154] (2.3) Update the memory unit state: The calculation formula is expressed as:

[0155]

[0156] (2.4) Output gate: The calculation formula is expressed as:

[0157] o t =σ(W o ·[h t―1 ,x t ]+b o ),h t =o t tanh(C t );

[0158] Among them, t is the output gate state, h t is the hidden state at the current moment, W o and b o is the weight matrix and bias value;

[0159] (3) Define the loss function: Use the cross entropy loss function to measure the error between the predicted result and the true value. The formula is:

[0160]

[0161] Where N is, y i is the true label, is the predicted probability output by the model;

[0162] (4) Optimize model parameters: Use the Adam optimization algorithm to update model parameters, reduce the loss function value, and improve prediction accuracy;

[0163] Model validation: After the plaque risk prediction model is trained, the plaque risk prediction model is evaluated to ensure that the model has good prediction performance, including:

[0164] (1) Evaluation metrics: Use metrics such as accuracy, precision, recall, and F1 score to measure model performance;

[0165] (2) Overfitting processing: Regularization and early stopping mechanisms are used to avoid model overfitting and ensure the generalization ability of the model;

[0166] Model deployment: Deploy the trained plaque risk prediction model to the system to predict the risk probability of future vascular plaque formation in real time, including:

[0167] Input data processing: converting multimodal data collected in real time into feature vectors and inputting them into the model;

[0168] Prediction result output: The model output is the future risk probability, ranging from 0 to 1, and corresponds to the risk level standard (such as low risk, medium risk, high risk);

[0169] Through the plaque risk prediction model constructed based on the long short-term memory network algorithm, the system can effectively capture the dynamic change characteristics in time series data, provide accurate predictions of future risks, and provide a scientific basis for personalized health management and intervention.

[0170] The personalized health intervention recommendation module includes:

[0171] Risk grading: Based on the output probability of the plaque risk prediction model, users are divided into different risk levels, including low risk, medium risk and high risk, providing a basis for formulating personalized intervention plans, including:

[0172] Low risk (0≤P<0.3): No drug intervention is recommended, only maintaining a healthy lifestyle;

[0173] Medium risk (0.3≤P<0.7): moderate adjustment of diet and exercise plan is recommended;

[0174] High risk (P ≥ 0.7): A strict intervention plan is recommended and drug treatment may be required;

[0175] Diet adjustment plan generation: Develop a diet adjustment plan based on the user's risk level, including:

[0176] (1) Nutritional needs assessment: Evaluate the user's current dietary structure and analyze whether the ratio of fat, protein, and carbohydrates is reasonable through intake data;

[0177] If the marker LDL-C is high, it is recommended to reduce the intake of saturated fatty acids and increase the intake of foods rich in unsaturated fatty acids (such as fish and nuts);

[0178] (2) Generate a personalized diet plan: Recommend daily food types and intakes based on the user’s metabolic needs and risk factors, such as:

[0179] High-risk users: It is recommended to reduce red meat intake and increase high-fiber foods (such as whole grains, green leafy vegetables);

[0180] Medium-risk users: It is recommended to control total calorie intake appropriately and maintain a balanced nutrition;

[0181] Provide sample menus, including recommended options for breakfast, lunch, dinner, and snacks;

[0182] Exercise intervention plan generation: Generate an exercise intervention plan based on the user's risk level and lifestyle data, including:

[0183] (1) Sports ability assessment: assess the user's basic sports ability based on the exercise frequency, heart rate and BMI in the time series data;

[0184] If you are not exercising enough and your BMI is too high, it is recommended to gradually increase the amount of exercise;

[0185] (2) Generate exercise plan:

[0186] High-risk users: Low-intensity aerobic exercise (such as walking, yoga) is recommended and the intensity is gradually increased. It is recommended that the daily exercise time is not less than 30 minutes;

[0187] Moderate-risk users: Moderate-intensity aerobic exercise (such as jogging and cycling) is recommended, at least 5 times a week, 40 minutes each time;

[0188] Low-risk users: Maintain the current exercise frequency and undergo regular physical examinations;

[0189] Drug intervention recommendation generation: For high-risk users, drug intervention recommendations are generated based on vascular imaging data and physical examination report data, including:

[0190] (1) Drug recommendations corresponding to risk factors: If the marker LDL-C is significantly high, lipid-lowering drugs (such as statins) are recommended; if blood pressure is at a high level for a long time, antihypertensive drugs (such as ACEIs and CCBs) are recommended;

[0191] (2) Drug safety assessment: combining test data to avoid side effects caused by drug interactions;

[0192] (3) Dynamic adjustment of medication plan: Dynamically adjust medication dosage or medication discontinuation plan based on the monitoring results of time series data (such as the downward trend of the marker LDL-C);

[0193] Personalized report generation: The generated intervention recommendations are integrated into a personalized health management report for the user, including:

[0194] (1) Risk assessment results: Displays the user's current risk level and dynamic trends of time series data;

[0195] (2) Summary of intervention recommendations: including specific implementation plans for dietary adjustments, exercise plans, and drug intervention recommendations;

[0196] The personalized health intervention recommendation module dynamically generates diet, exercise and drug intervention recommendations by combining the output of the plaque risk prediction model and the user's multimodal time series data, providing users with scientific, real-time and personalized health management plans to effectively reduce the risk of vascular plaque formation.

[0197] Visualization and decision support modules include:

[0198] Visualization of prediction results: The output of the plaque risk prediction model is displayed through a graphical user interface, including:

[0199] (1) Risk probability chart display: Display the user's current risk level and corresponding risk probability in the form of a risk level bar chart or pie chart;

[0200] (2) Dynamic display of time series data: Use line graphs to display the time series changes of physical examination report data (such as markers Lp(a), LDL-C, HDL-C, FBS, and IMT values) to help users understand the impact of dynamic trends on risk;

[0201] Importance distribution display of key risk factors: Visualize the key risk factors extracted by the feature selection and risk factor mining modules to help users and medical practitioners understand the main influencing factors, including:

[0202] Risk factor weight chart: Use a bar chart or radar chart to display the importance score of each risk factor (such as the characteristic importance of markers Lp(a), LDL-C, and IMT values) and arrange them in descending order;

[0203] Correlation heat map: draw a correlation heat map between risk factors to show the positive and negative correlations between factors (e.g. markers Lp(a), LDL-C and IMT values ​​may be positively correlated);

[0204] Visual display of personalized intervention recommendations: Diet, exercise and drug intervention recommendations are presented to users through a graphical interface, including:

[0205] (1) Diet plan schedule presentation: Generate a sample menu for the diet adjustment plan and present it in the form of a schedule, including the recommended daily food types, portions, and nutrients (such as the ratio of fat, protein, and carbohydrates);

[0206] (2) Graphic display of exercise plan: Display the recommended daily exercise type, duration, and intensity in the form of a calendar, such as "jogging for 30 minutes" or "yoga for 45 minutes";

[0207] (3) Graphical support for drug recommendations: Generate a drug use plan for high-risk users and dynamically remind them of the time and dosage of medication on the interface;

[0208] Generation and presentation of risk assessment reports: Providing risk assessment reports to medical practitioners to assist clinical decision-making, including:

[0209] Comprehensive assessment report generation: Summarize the output results of the plaque risk prediction model, key risk factor analysis and trend changes of time series data to generate a comprehensive assessment report;

[0210] The report content includes risk level, time series change chart, key risk factor ranking, correlation analysis and intervention suggestions. It provides export function and supports storage or printing of reports in PDF format for user health record record;

[0211] The visualization and decision support module displays prediction results, key risk factor analysis and personalized health intervention recommendations through a multi-dimensional and intuitive graphical interface, and generates professional risk assessment reports, providing scientific basis and efficient support for user health management and clinical decision-making of medical practitioners.

[0212] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0213] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A vascular plaque prediction system based on health examination big data, characterized in that: It includes multimodal data acquisition module, data preprocessing and fusion module, feature selection and risk factor mining module, plaque prediction model construction module, personalized health intervention recommendation module and visualization and decision support module, among which; The multimodal data acquisition module is used to acquire health-related multimodal data from multiple data sources, wherein the multimodal data includes vascular imaging data, wherein; The vascular imaging data are vascular ultrasound, CT and MRI image data acquired by medical imaging equipment; The data preprocessing and fusion module is used to preprocess the collected multimodal data, including data cleaning and standardization, and use a deep belief network algorithm to extract features from the preprocessed multimodal data, and fuse the extracted features into a unified feature vector; The feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, extracts key risk factors related to vascular plaque formation, and sorts the extracted key risk factors by calculating feature importance scores, and analyzes the correlation between the risk factors. The key risk factors include biochemical risk factors, lifestyle risk factors, and vascular imaging risk factors; The plaque prediction model building module builds a vascular plaque risk prediction model based on a long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future; The personalized health intervention suggestion module generates personalized intervention suggestions according to the risk probability output by the vascular plaque risk prediction model, and the suggestions include a diet adjustment plan, an exercise intervention plan, and a drug intervention suggestion; The visualization and decision support module displays prediction results, importance distribution of key risk factors and personalized intervention recommendations through a graphical user interface, and provides risk assessment reports for medical practitioners.

2. The vascular plaque prediction system based on health examination big data according to claim 1 is characterized in that: The multimodal data acquisition module comprises: Physical examination report data collection: obtain the user's historical and current physical examination report data through the hospital information system interface; Test data collection: collect users' test data; Lifestyle data collection: Users input their daily diet structure, exercise frequency, smoking and drinking records through mobile applications or wearable devices; Vascular imaging data acquisition: Obtain the user's vascular ultrasound image, CT image and MRI image data through medical imaging equipment; Data timestamp and labeling step: Add a timestamp to each piece of collected multimodal data and add labels to the data according to the data type and source.

3. The vascular plaque prediction system based on health examination big data according to claim 2 is characterized in that: The data preprocessing and fusion module includes: Data cleaning: Clean the collected multimodal data, remove outliers, fill in missing values, and standardize the data format to ensure data integrity and consistency; Data standardization: Standardize the cleaned multimodal data.

4. The vascular plaque prediction system based on health examination big data according to claim 3 is characterized in that: The data preprocessing and fusion module uses a deep belief network algorithm to extract features from the preprocessed multimodal data, including: Deep belief network model construction: A deep belief network model is constructed to extract deep features of multimodal data layer by layer. The deep belief network model consists of multiple layers of restricted Boltzmann machines, and the model is optimized through unsupervised pre-training and global fine-tuning. Data input: input the preprocessed multimodal data into the constructed deep belief network model for feature extraction; Feature fusion: The multimodal feature vectors output by the deep belief network model are fused to generate a unified feature representation.

5. The vascular plaque prediction system based on health examination big data according to claim 4 is characterized in that: The feature selection and risk factor mining module performs feature selection on the fused feature vector based on the XGBoost algorithm, including: Model construction: A feature selection model is constructed based on the XGBoost algorithm to measure the contribution of each feature to the prediction result by fitting the target variable and to evaluate the importance of the feature; Data input: Input the feature vector into the constructed feature selection model; Feature importance calculation: Utilize the feature importance evaluation mechanism of the feature selection model to calculate the gain score of each input feature based on the feature selection model; Screening features: Sort the features according to the gain score and select the features that contribute most to the prediction of the target variable as key risk factors.

6. The vascular plaque prediction system based on health examination big data according to claim 5 is characterized in that: The feature selection and risk factor mining module sorts and analyzes the correlation of the extracted key risk factors, including: Ranking of key risk factors: The extracted key risk factors are ranked according to their contribution to the risk of vascular plaque formation; Risk factor correlation analysis: Analyze the correlation between key risk factors and reveal the potential interactions between multiple factors.

7. The vascular plaque prediction system based on health examination big data according to claim 6 is characterized in that: The plaque prediction model building module includes: Model construction: Construct a plaque risk prediction model based on the long short-term memory network algorithm to predict the risk probability of vascular plaque formation in the future; Model training: The feature vectors extracted by the feature selection and risk factor mining module are input into the constructed plaque risk prediction model for training; Model validation: After the plaque risk prediction model training is completed, the plaque risk prediction model is evaluated; Model deployment: The trained plaque risk prediction model is deployed into the system to predict the risk probability of future vascular plaque formation in real time.

8. The vascular plaque prediction system based on health examination big data according to claim 7 is characterized in that: The personalized health intervention suggestion module includes: Risk grading: Based on the output probability of the plaque risk prediction model, users are divided into different risk levels, including low risk, medium risk and high risk; Diet adjustment plan generation: Develop a diet adjustment plan based on the user’s risk level; Exercise intervention plan generation: Generate an exercise intervention plan based on the user's risk level and lifestyle data; Drug intervention recommendation generation: For high-risk users, drug intervention recommendations are generated based on vascular imaging data and physical examination report data; Personalized report generation: Integrate the generated intervention recommendations into a user-specific personalized health management report.

9. The vascular plaque prediction system based on health examination big data according to claim 8, characterized in that: The visualization and decision support module includes: Visualization of prediction results: Display the output of the plaque risk prediction model through a graphical user interface; Importance distribution display of key risk factors: Visual display of key risk factors extracted by feature selection and risk factor mining modules; Visual display of personalized intervention recommendations: Diet, exercise and drug intervention recommendations are presented to users through a graphical interface; Generation and presentation of risk assessment reports: Provide risk assessment reports to medical practitioners to assist in clinical decision-making.

Images