AI-based chronic disease biochemical data analysis system and method

Abstract

The invention provides an AI-based chronic disease biochemical data analysis system and method. The system comprises a data acquisition module, a model construction module, a prediction module and a visual interaction module, the data acquisition module is used for acquiring biochemical examination data and life behavior data of a patient; the model construction module is used for constructing a behavior influence model and an index influence model and fusing the behavior influence model and the index influence model to realize multi-modal health risk modeling; the prediction module is used for inputting data acquired by the data acquisition module into the behavior influence model and the index influence model to output a chronic disease risk assessment result, and the visual interaction module is used for interactively displaying the chronic disease risk assessment result for a user; according to the method, the life behavior data of the patient and the biochemical examination data are fused, so that multi-mode, explainable and personalized chronic disease risk assessment is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of chronic disease data analysis, and in particular to an AI-based chronic disease biochemical data analysis system and method. Background Art

[0002] With the development of medical informatization, chronic diseases have gradually become the main type of disease threatening public health; early detection, dynamic monitoring and personalized intervention of chronic diseases have become important research directions of modern smart medical systems.

[0003] Currently, a large amount of chronic disease-related data is stored in the medical system, covering structured biochemical examination indicators and unstructured life behavior records; traditional methods mostly rely on a single data type for modeling, which makes it difficult to fully characterize the multidimensional factors that form chronic disease risks, limiting the accuracy and individualization capabilities of predictions.

[0004] Among existing technical solutions, some studies have attempted to extract chronic disease-related symptoms and treatment information from electronic health records based on natural language processing technology, and use neural network models for classification and display. However, such methods are mostly limited to text mining and static knowledge extraction, lack the ability to jointly model patient behavioral habits and physiological indicators, and fail to achieve the time series expression of behavioral risk scores and the adaptive update mechanism of model fusion.

[0005] Therefore, there is an urgent need for a chronic disease risk analysis system that integrates life behavior data and biochemical indicator data and has dynamic learning and personalized modeling capabilities to improve the scientificity and practicality of intelligent predictions and intervention recommendations.

[0006] A review of related published technical proposals reveals that publication number CN112287665B proposes a method and system for analyzing chronic disease data based on natural language processing and ensemble training. The system comprises a data preprocessing module, a data recognition module, a data training module, and a data visualization module. The data preprocessing module extracts chronic disease data from an external chronic disease database, generates corresponding word vectors, and then quantizes these word vectors to serve as training samples. The data recognition module inputs the word vectors of the training samples into a bidirectional long-short-term memory network for training, generating hidden vectors that are then transferred to a conditional probability field to calculate character labels. The data training module performs classification training to extract a ternary association model between chronic disease symptoms, biochemical pathological indicators, and treatments. This model is then transferred to the data visualization module for statistical analysis and presentation to an external user interface module. This approach fully utilizes large amounts of unstructured electronic medical data, combines natural language processing with ensemble training neural networks, and performs causal analysis and disease prediction for chronic diseases, thereby providing targeted treatment. However, this approach fails to achieve deep integration of biochemical examination data with natural language information and lacks a multimodal data evaluation mechanism, limiting the comprehensiveness and accuracy of its risk assessment model. Summary of the Invention

[0007] The purpose of this invention is to address the current deficiencies and propose an AI-based chronic disease biochemical data analysis system and method.

[0008] The present invention adopts the following technical solutions:

[0009] A chronic disease biochemical data analysis system based on AI, comprising a data acquisition module, a model building module, a prediction module and a visualization interaction module.

[0010] The data acquisition module is used to collect patient biochemical examination data and life behavior data; the model construction module is used to construct a behavior impact model and an indicator impact model and integrate the two to achieve multimodal health risk modeling; the prediction module is used to input the data collected by the data acquisition module into the behavior impact model and the indicator impact model to output the chronic disease risk assessment results; the visualization interaction module is used to interactively display the chronic disease risk assessment results for users.

[0011] The data acquisition module includes a biochemical indicator acquisition unit and a behavior data acquisition unit; the biochemical indicator acquisition unit is used to acquire the patient's biochemical examination data; the behavior data acquisition unit is used to acquire the patient's life behavior data, and the life behavior data is a natural language filled text.

[0012] The model construction module includes a behavior influence model construction unit, an indicator influence model construction unit and a model fusion unit; the behavior influence model construction unit is used to construct a behavior influence model for predicting the development of chronic diseases based on the patient's life behavior data; the indicator influence model construction unit is used to construct an indicator influence model for predicting the development of chronic diseases based on the patient's biochemical examination data; the model fusion unit is used to fuse the behavior influence model and the indicator influence model and output the final risk assessment for the patient's chronic disease.

[0013] Furthermore, the behavior impact model is constructed as follows:

[0014] S211: Establishing a behavior vocabulary and a degree vocabulary, wherein the behavior vocabulary contains word labels for typical behaviors related to chronic disease management, and the degree vocabulary contains word labels for identifying the severity of behaviors;

[0015] S212: Obtaining the life behavior data and the corresponding chronic disease risk descriptions from real user health records and extracting sample data from them. Each sample data item includes a typical behavior word label recorded by a patient within a specified time window, a severity word label describing the typical behavior word label, and a chronic disease risk label. The chronic disease risk label is pre-labeled by experts based on the chronic disease risk description.

[0016] S213: Use the sample data obtained in the previous step to form a training set to train and build a behavior impact model. The input of the behavior impact model is typical behavior word labels and severity word labels that describe the typical behavior word labels. The output is a chronic disease risk label, and the model output result is mapped to the time dimension according to the time information in the life behavior data.

[0017] Furthermore, the indicator impact model is constructed as follows:

[0018] S221: Obtain biochemical examination data of real users and their corresponding chronic disease risk descriptions from medical institutions and electronic health record systems;

[0019] S222: Extracting sample data from the data acquired in the previous step, wherein each piece of sample data includes a biochemical examination data sequence of a patient within a specified time window and a corresponding chronic disease risk label, wherein the chronic disease risk label is pre-labeled by an expert based on a description of the chronic disease risk;

[0020] S223: Use the sample data obtained in the previous step to form a training set to train and build an indicator influence model, where the input of the indicator influence model is a structured biochemical examination data sequence and the output is a chronic disease risk label.

[0021] Furthermore, the model fusion unit completes the fusion of the behavior impact model and the indicator impact model in the following manner:

[0022] R t =α B R1(t)+α s R2(t);

[0023] Among them, R t is the final output of the model fusion unit at time point t, α B is the behavior impact model weight, α s is the weight of the indicator influence model, R1(t) is the final output of the behavior influence model at time point t, and R2(t) is the final output of the indicator influence model at time point t; satisfying:

[0024]

[0025] Among them, ΔR1(t) is the output change of the behavior influence model at time point t, which is obtained by the difference between the output of the behavior influence model at time point t and the previous time point t-1; ΔR2(t) is the output change of the indicator influence model at time point t, which is obtained by the difference between the output of the indicator influence model at time point t and the previous time point t-1; ω1 and ω2 are adjustable weight parameters, which can be pre-set initially.

[0026] Furthermore, the prediction module includes a feature extraction unit and a prediction output unit; the feature extraction unit is used to convert the patient's biochemical examination data and life behavior data into input types for the behavior influence model and the indicator influence model; the prediction output unit is used to input the converted input types into the behavior influence model and the indicator influence model, and finally obtain the final chronic disease risk label as the final output through the model fusion unit; and the final output is transmitted to the visualization interaction module for display;

[0027] An AI-based method for analyzing biochemical data of chronic diseases, comprising the following steps:

[0028] S1: Obtain the patient's biochemical examination data and life behavior data;

[0029] S2: Constructing a behavior influence model and an indicator influence model. The behavior influence model evaluates and analyzes the risk of chronic diseases of patients based on their life behavior data, and the indicator influence model evaluates and analyzes the risk of chronic diseases of patients based on their biochemical examination data.

[0030] S3: Integrate the behavior impact model and the indicator impact model;

[0031] S4: After feature processing, the patient's biochemical examination data and life behavior data are input into the fused behavior influence model and indicator influence model to output the risk assessment of the patient's chronic disease.

[0032] The beneficial effects achieved by the present invention are:

[0033] The present invention combines the patient's biochemical examination data and life behavior data, integrates the patient's subjective behavior description and objective biochemical indicators, and establishes a multimodal fusion model to achieve intelligent assessment of chronic disease risks; through the dynamic fusion of behavior impact model and indicator impact model, the system can give full play to the advantages of each modality data and improve the accuracy and stability of prediction; by introducing an adaptive weighted fusion mechanism based on the degree of fluctuation of model output, the fusion weight is automatically adjusted with the degree of fluctuation of model output, thereby enhancing the responsiveness of the model under different health states. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but rather the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0035] Figure 1 It is a schematic diagram of the overall module of the present invention.

[0036] Figure 2This is a flow chart of the AI-based chronic disease biochemical data analysis method of the present invention.

[0037] Figure 3 A schematic diagram of the process for constructing the behavior impact model of the present invention.

[0038] Figure 4 Schematic diagram of the process of constructing the indicator impact model of the present invention.

[0039] Figure 5 When |ΔR2(t)| is fixed to 0.5, α B Schematic diagram of the function changing with |ΔR1(t)|. DETAILED DESCRIPTION

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with its embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention; for those skilled in the art, after reviewing the following detailed description, other systems, methods and / or features of the present embodiment will become apparent; it is intended that all such additional systems, methods, features and advantages are included in this specification; included within the scope of the present invention and protected by the appended claims; additional features of the disclosed embodiments are described in the following detailed description, and these features will be apparent from the following detailed description.

[0041] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. This is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or component referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0042] Example 1:

[0043] like Figure 1 As shown, this embodiment provides an AI-based chronic disease biochemical data analysis system, which includes a data acquisition module, a model building module, a prediction module, and a visualization interaction module;

[0044] The data acquisition module is used to collect biochemical examination data and life behavior data of patients; the model construction module is used to construct a behavior influence model and an indicator influence model and integrate the two to achieve multimodal health risk modeling; the prediction module is used to input the data collected by the data acquisition module into the behavior influence model and the indicator influence model to output the chronic disease risk assessment results; the visualization interaction module is used to interactively display the chronic disease risk assessment results for users;

[0045] The data acquisition module includes a biochemical indicator acquisition unit and a behavior data acquisition unit; the biochemical indicator acquisition unit is used to collect the patient's biochemical examination data; the behavior data acquisition unit is used to collect the patient's life behavior data, and the life behavior data is a natural language filled text;

[0046] The model construction module includes a behavior influence model construction unit, an indicator influence model construction unit and a model fusion unit; the behavior influence model construction unit is used to construct a behavior influence model for predicting the development of chronic diseases based on the patient's life behavior data; the indicator influence model construction unit is used to construct an indicator influence model for predicting the development of chronic diseases based on the patient's biochemical examination data; the model fusion unit is used to fuse the behavior influence model and the indicator influence model and output a final risk assessment for the patient's chronic disease;

[0047] Further, such as Figure 3 As shown, the behavior impact model is constructed as follows:

[0048] S211: Establish a behavior vocabulary and a degree vocabulary. The behavior vocabulary contains word tags for typical behaviors related to chronic disease management, and the degree vocabulary contains word tags for identifying the severity of behaviors. Specifically, the typical behavior vocabulary tags in the behavior vocabulary include, but are not limited to, staying up late, sitting for a long time, drinking, smoking, and lack of exercise. The word tags in the degree vocabulary include frequency, time, and intensity word tags, including, but not limited to:

[0049] Frequency categories: occasionally, sometimes, often, daily, and continuously;

[0050] Time category: recent days, within three months and within one year;

[0051] Intensity: slightly, somewhat, very, extremely, and severely;

[0052] S212: Obtaining the life behavior data and the corresponding chronic disease risk descriptions from real user health records and extracting sample data from them. Each sample data item includes a typical behavior word label recorded by a patient within a specified time window, a severity word label describing the typical behavior word label, and a chronic disease risk label. The chronic disease risk label is pre-labeled by experts based on the chronic disease risk description.

[0053] S213: Using the sample data obtained in the previous step to form a training set, training and constructing a behavior impact model, wherein the input of the behavior impact model is a typical behavior word label and a severity word label that describes the typical behavior word label, and the output is a chronic disease risk label. The model output result is mapped to the time dimension according to the time information in the life behavior data;

[0054] Further, such as Figure 4 As shown, the indicator impact model is constructed as follows:

[0055] S221: Obtain biochemical examination data of real users and their corresponding chronic disease risk descriptions from medical institutions and electronic health record systems;

[0056] S222: Extracting sample data from the data acquired in the previous step, wherein each piece of sample data includes a biochemical examination data sequence of a patient within a specified time window and a corresponding chronic disease risk label, wherein the chronic disease risk label is pre-labeled by an expert based on a description of the chronic disease risk;

[0057] S223: Using the sample data obtained in the previous step to form a training set to train and construct an indicator influence model, the input of the indicator influence model is a structured biochemical examination data sequence, and the output is a chronic disease risk label;

[0058] Further, such as Figure 5 As shown, the model fusion unit completes the fusion of the behavior impact model and the indicator impact model in the following ways:

[0059] R t =α B R1(t)+α s R2(t);

[0060] Among them, R t is the final output of the model fusion unit at time point t, α B is the behavior impact model weight, α s is the weight of the indicator influence model, R1(t) is the final output of the behavior influence model at time point t, and R2(t) is the final output of the indicator influence model at time point t; satisfying:

[0061]

[0062] Wherein, ΔR1(t) is the output change of the behavior influence model at time point t, which is obtained by the difference between the output of the behavior influence model at time point t and the previous time point t-1; ΔR2(t) is the output change of the indicator influence model at time point t, which is obtained by the difference between the output of the indicator influence model at time point t and the previous time point t-1; ω1 and ω2 are adjustable weight parameters that can be pre-set initially;

[0063] By adopting the above-mentioned adaptive weighted fusion mechanism based on the degree of model output fluctuation, the fusion weight is automatically adjusted with the degree of model output fluctuation, thereby enhancing the model's responsiveness under different health conditions; if a model output fluctuates violently, the system will automatically increase its weight, so that the fusion result is more focused on the current risk source, effectively improving the adaptability, stability and individualized prediction ability of multimodal model fusion in chronic disease risk assessment, and has high practicality and broad application prospects.

[0064] Furthermore, the prediction module includes a feature extraction unit and a prediction output unit; the feature extraction unit is used to convert the patient's biochemical examination data and life behavior data into input types for the behavior influence model and the indicator influence model; the prediction output unit is used to input the converted input types into the behavior influence model and the indicator influence model, and finally obtain the final chronic disease risk label as the final output through the model fusion unit; and the final output is transmitted to the visualization interaction module for display;

[0065] Specifically, the chronic disease risk label is used to indicate the risk level of a patient suffering from or developing a certain chronic disease at a specific time point or time period. The label can be in the form of a continuous numerical score, such as a risk numerical score between 0 and 10.

[0066] like Figure 2 As shown, a method for analyzing biochemical data of chronic diseases based on AI comprises the following steps:

[0067] S1: Obtain the patient's biochemical examination data and life behavior data;

[0068] S2: Constructing a behavior influence model and an indicator influence model. The behavior influence model evaluates and analyzes the risk of chronic diseases of patients based on their life behavior data, and the indicator influence model evaluates and analyzes the risk of chronic diseases of patients based on their biochemical examination data.

[0069] S3: Integrate the behavior impact model and the indicator impact model;

[0070] S4: After feature processing, the patient's biochemical examination data and life behavior data are input into the fused behavior influence model and indicator influence model to output the risk assessment of the patient's chronic disease.

[0071] Example 2:

[0072] This embodiment should be understood to include at least all the features of any of the aforementioned embodiments and be further improved thereon;

[0073] This embodiment provides an AI-based chronic disease biochemical data analysis system, which includes a data acquisition module, a model building module, a prediction module, and a visualization interaction module;

[0074] The data acquisition module is used to collect biochemical examination data and life behavior data of patients; the model construction module is used to construct a behavior influence model and an indicator influence model and integrate the two to achieve multimodal health risk modeling; the prediction module is used to input the data collected by the data acquisition module into the behavior influence model and the indicator influence model to output the chronic disease risk assessment results; the visualization interaction module is used to interactively display the chronic disease risk assessment results for users;

[0075] The model construction module includes a behavior influence model construction unit, an indicator influence model construction unit and a model fusion unit; the behavior influence model construction unit is used to construct a behavior influence model for predicting the development of chronic diseases based on the patient's life behavior data; the indicator influence model construction unit is used to construct an indicator influence model for predicting the development of chronic diseases based on the patient's biochemical examination data; the model fusion unit is used to fuse the behavior influence model and the indicator influence model and output a final risk assessment for the patient's chronic disease;

[0076] Furthermore, the adjustable weight parameters in the model fusion unit can be adjusted in combination with the model output and the real feedback. The specific adjustment process is as follows:

[0077] S31: Obtain the difference between the output of the model fusion unit and the actual corresponding chronic disease risk label, and construct the loss function:

[0078]

[0079] Among them, L T is a loss function used to measure the degree of deviation between the output of the model fusion unit at the current time T and the true label; R T is the output of the model fusion unit at the current time T, is the real chronic disease risk label at the current time T, obtained through the real medical test results at the current time T;

[0080] S32: When the loss function value is greater than a preset threshold, the adjustable weight parameters are gradually adjusted by the gradient descent method;

[0081] By combining the output of the model fusion unit with the actual corresponding chronic disease risk label in the above way, the adjustable weight parameters are adaptively updated, thereby achieving refined, personalized and intelligent optimization of the model fusion strategy without relying on human intervention, effectively improving the accuracy of chronic disease risk assessment and the sustainable learning ability of the system.

[0082] The contents disclosed above are only preferred feasible embodiments of the present invention and do not limit the scope of protection of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention description and drawings are included in the scope of protection of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. An AI-based chronic disease biochemical data analysis system, characterized by: The system includes a data acquisition module, a model building module, a prediction module and a visualization interaction module; The data acquisition module is used to collect biochemical examination data and life behavior data of patients; the model construction module is used to construct a behavior impact model and an indicator impact model and integrate the two to achieve multimodal health risk modeling; The prediction module is used to input the data collected by the data collection module into the behavior influence model and the indicator influence model to output the chronic disease risk assessment results, and the visualization interaction module is used to interactively display the chronic disease risk assessment results for users; The data acquisition module includes a biochemical indicator acquisition unit and a behavioral data acquisition unit; The biochemical indicator acquisition unit is used to collect biochemical examination data of patients; The behavior data collection unit is used to collect the patient's life behavior data, and the life behavior data is a text filled in natural language; The model construction module includes a behavior influence model construction unit, an indicator influence model construction unit and a model fusion unit; the behavior influence model construction unit is used to construct a behavior influence model for predicting the development of chronic diseases based on the patient's life behavior data; the indicator influence model construction unit is used to construct an indicator influence model for predicting the development of chronic diseases based on the patient's biochemical examination data; the model fusion unit is used to fuse the behavior influence model and the indicator influence model and output the final risk assessment for the patient's chronic disease.

2. The AI-based chronic disease biochemical data analysis system according to claim 1, characterized in that: The behavior impact model is constructed as follows: S211: Establishing a behavior vocabulary and a degree vocabulary, wherein the behavior vocabulary contains word labels for typical behaviors related to chronic disease management, and the degree vocabulary contains word labels for identifying the severity of behaviors; S212: Obtaining the life behavior data and the corresponding chronic disease risk descriptions from real user health records and extracting sample data from them. Each sample data item includes a typical behavior word label recorded by a patient within a specified time window, a severity word label describing the typical behavior word label, and a chronic disease risk label. The chronic disease risk label is pre-labeled by experts based on the chronic disease risk description. S213: Use the sample data obtained in the previous step to form a training set to train and build a behavior impact model. The input of the behavior impact model is typical behavior word labels and severity word labels that describe the typical behavior word labels. The output is a chronic disease risk label, and the model output result is mapped to the time dimension according to the time information in the life behavior data.

3. The AI-based chronic disease biochemical data analysis system according to claim 1, characterized in that: The indicator impact model is constructed as follows: S221: Obtain biochemical examination data of real users and their corresponding chronic disease risk descriptions from medical institutions and electronic health record systems; S222: Extracting sample data from the data acquired in the previous step, each piece of sample data includes a biochemical examination data sequence of a patient within a specified time window and a corresponding chronic disease risk label, wherein the chronic disease risk label is pre-labeled by an expert based on a description of the chronic disease risk; S223: Use the sample data obtained in the previous step to form a training set to train and build an indicator influence model, where the input of the indicator influence model is a structured biochemical examination data sequence and the output is a chronic disease risk label.

4. The AI-based chronic disease biochemical data analysis system according to claim 1, characterized in that: The model fusion unit specifically completes the fusion of the behavior impact model and the indicator impact model in the following ways: R t =α B ·R1(t)+α s ·R2(t); Among them, R t is the final output of the model fusion unit at time point t, α B is the behavior impact model weight, α s is the weight of the indicator influence model, R1(t) is the final output of the behavior influence model at time point t, and R2(t) is the final output of the indicator influence model at time point t; satisfying: Among them, ΔR1(t) is the output change of the behavior influence model at time point t, which is obtained by the difference between the output of the behavior influence model at time point t and the previous time point t-1; ΔR2(t) is the output change of the indicator influence model at time point t, which is obtained by the difference between the output of the indicator influence model at time point t and the previous time point t-1; ω1 and ω2 are adjustable weight parameters, which can be pre-set initially.

5. The AI-based chronic disease biochemical data analysis system according to claim 1, characterized in that: The prediction module includes a feature extraction unit and a prediction output unit; the feature extraction unit is used to convert the patient's biochemical examination data and life behavior data into input types of the behavior influence model and the indicator influence model; the prediction output unit is used to input the converted input types into the behavior influence model and the indicator influence model, and finally obtain the final chronic disease risk label as the final output through the model fusion unit; and pass the final output to the visualization interaction module for display.

6. An AI-based chronic disease biochemical data analysis method, applied to the AI-based chronic disease biochemical data analysis system according to claim 1, characterized in that: The method comprises the following steps: S1: Obtain the patient's biochemical examination data and life behavior data; S2: Constructing a behavior influence model and an indicator influence model. The behavior influence model evaluates and analyzes the risk of chronic diseases of patients based on their life behavior data, and the indicator influence model evaluates and analyzes the risk of chronic diseases of patients based on their biochemical examination data. S3: Integrate the behavior impact model and the indicator impact model; S4: After feature processing, the patient's biochemical examination data and life behavior data are input into the fused behavior influence model and indicator influence model to output the risk assessment of the patient's chronic disease.

Images