Large model-based medical health risk assessment method, apparatus and device, and medium
The large model-based health risk assessment method integrates and processes multi-channel health data to provide accurate, automated, and efficient health risk predictions and personalized recommendations, overcoming limitations of traditional methods.
Patent Information
- Application Number
- CN202510450961.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-11
- Publication Date
- 2025-07-15
AI Technical Summary
Traditional health risk assessment methods cannot achieve large-scale automated processing, low analysis efficiency, and difficult to discover complex health patterns.
A large-model-based medical and health risk assessment method is adopted to integrate, preprocess, feature extraction and fusion of multi-channel health data, and quantitative analysis is performed using preset health risk assessment models, and a personalized health assessment report is generated.
Achieve more accurate and rapid health status and potential risk predictions, improve data processing efficiency and analytical accuracy, provide customized health risk assessments and intervention recommendations, and reduce the work burden of medical professionals.
Smart Images
Figure CN120319475A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and particularly to a method, device, equipment and medium for medical and health risk assessment based on a large model. Background Art
[0002] With the enhancement of people's health awareness and the development of medical technology, the demand for health risk assessment is increasing day by day. Traditional health risk assessment methods mainly rely on questionnaires, physical examination data and doctors' clinical experience. To a certain extent, these methods can help identify an individual's health status and potential risks, but there are still great limitations, such as the inability to achieve large-scale automated processing, low analysis efficiency, and difficulty in discovering complex health patterns. Summary of the Invention
[0003] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for medical and health risk assessment based on a large model, which can more accurately and quickly predict an individual's health status and potential risks by combining large model technology. The specific solutions are as follows:
[0004] In the first aspect, the present application discloses a method for medical and health risk assessment based on a large model, which is applied to a medical and health risk assessment system and includes:
[0005] Integrate and preprocess the health data from multiple channels of the target object, and extract and fuse features from the preprocessed multi-channel health data to obtain fused feature data;
[0006] Perform quantitative analysis on the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data;
[0007] Match according to the health risk score and the user data corresponding to the target object in a preset risk index database to obtain the health problems and health development trends corresponding to the target object;
[0008] Generate a health assessment report based on the health problems and the health development trends, and convert the professional medical terms in the health assessment report into corresponding target phrases through natural language processing to obtain a target health assessment report.
[0009] Optionally, the integrating and preprocessing the health data from multiple channels of the target object, and extracting and fusing features from the preprocessed multi-channel health data to obtain fused feature data includes:
[0010] Aggregating the health data of the target object from several channels through a preset data receiving framework to obtain multi-channel health data corresponding to the target object;
[0011] Performing multi-layer data filtering on the multi-channel health data to eliminate abnormal data in the multi-channel health data, and performing data interpolation on the obtained multi-channel health data after elimination to obtain interpolated multi-channel data;
[0012] Converting the data format of the interpolated multi-channel data into a preset data format, and performing data enhancement on the converted data to obtain preprocessed multi-channel health data;
[0013] Identify medical entities in the preprocessed multi-channel health data through a preset medical dictionary and a regular expression, and construct a plurality of medical entity triples based on the medical entities through a preset entity rule template;
[0014] The plurality of medical entity triplets are respectively coded and mapped by using a preset medical knowledge base to convert the plurality of medical entity triplets into standardized medical features, and feature fusion is performed on the obtained plurality of standardized medical features to obtain fused feature data.
[0015] Optionally, the performing multi-layer data filtering on the multi-channel health data to remove abnormal data in the multi-channel health data includes:
[0016] Performing a first data filtering on the multi-channel health data by using a plurality of preset health indicator physiological limit values to obtain first filtered data;
[0017] Identify first abnormal data having a contradictory combination of indicators in the first filtered data by using a preset machine learning algorithm, and remove the first abnormal data to obtain second filtered data;
[0018] The preset knowledge graph is used to screen the second abnormal data with logical contradictions in the second filtered data, and the second abnormal data is eliminated to obtain the eliminated multi-channel health data.
[0019] Optionally, the quantitative analysis of the fused feature data by a preset health risk assessment model to determine a health risk score corresponding to each indicator in the fused feature data includes:
[0020] Inputting the fused feature data into a preset health risk assessment model, so as to quantitatively analyze the fused feature data through the preset health risk assessment model to obtain the disease probability corresponding to each indicator in the fused feature data;
[0021] Map the probability of getting the disease through a preset piecewise linear function to convert the probability of getting the disease into a corresponding health risk score.
[0022] Optionally, matching in a preset risk index database according to the health risk score and the user data corresponding to the target object to obtain the health problems and health development trends corresponding to the target object, including:
[0023] Obtain the user data of the target object, and match in the preset risk index database according to the health risk score and the user data to match the target diseases corresponding to each index in the preset risk index database corresponding to the health risk score;
[0024] Analyze the user data and each of the target diseases to generate the health development trend of the target object;
[0025] Wherein, the user data includes physiological data, psychological data and life data.
[0026] Optionally, converting the professional medical terms in the health assessment report into corresponding target phrases through natural language processing to obtain a target health assessment report, including:
[0027] Perform word segmentation processing on the health assessment report through a preset word segmentation tool to obtain a number of texts to be converted;
[0028] Convert the professional medical terms in the number of texts to be converted into unified target medical terms in the preset medical dictionary to obtain a first target health assessment report;
[0029] Match the medical explanation information corresponding to the target medical terms from the preset medical information database, and add the medical explanation information to the first target health assessment report to obtain a second target health assessment report.
[0030] Optionally, the medical and health risk assessment method based on a large model further includes:
[0031] Collect the current health status and risk assessment evaluation corresponding to the target health assessment report feedback by the target object, and save the current health status and risk assessment evaluation as evaluation accuracy data to the local risk assessment database;
[0032] If the model evaluation time node is reached, determine the historical evaluation accuracy data of the preset time period corresponding to the model evaluation time node in the risk assessment database, and determine whether model performance optimization is required based on the historical evaluation accuracy data.
[0033] Second aspect, the present application discloses a medical and health risk assessment device based on a large model, which is applied to a medical and health risk assessment system, including:
[0034] A feature processing module, configured to perform data integration and preprocessing on the health data from multiple channels of a target object, and perform feature extraction and feature fusion on the preprocessed multi-channel health data obtained to obtain fused feature data;
[0035] A risk scoring module, configured to perform quantitative analysis on the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data;
[0036] A health evaluation module, configured to match according to the health risk score and user data corresponding to the target object in a preset risk index database to obtain the health problems and health development trends corresponding to the target object;
[0037] An evaluation report generation module, configured to generate a health evaluation report based on the health problems and the health development trends, and convert the professional medical terms in the health evaluation report into corresponding target phrases through natural language processing to obtain a target health evaluation report.
[0038] Third aspect, the present application discloses an electronic device, including:
[0039] A memory, configured to store a computer program;
[0040] A processor, configured to execute the computer program to implement the medical and health risk assessment method based on a large model as described above.
[0041] Fourth aspect, the present application discloses a computer-readable storage medium, configured to store a computer program, wherein the computer program, when executed by a processor, implements the medical and health risk assessment method based on a large model as described above.
[0042] In this application, the health data from multiple channels of the target object can be integrated and preprocessed, and feature extraction and feature fusion can be performed on the preprocessed multi-channel health data obtained to obtain the fused feature data; the fused feature data can be quantitatively analyzed through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data; according to the health risk score and the user data corresponding to the target object, a match is made in a preset risk index database to obtain the health problems and health development trends corresponding to the target object; a health assessment report is generated based on the health problems and the health development trends, and the professional medical terms in the health assessment report are converted into corresponding target phrases through natural language processing to obtain a target health assessment report.
[0043] It can be seen that through the method of this application, after receiving the multi-channel health data of the target object, it can be integrated and preprocessed, and feature extraction and feature fusion can be performed on the preprocessed multi-channel health data obtained to obtain the fused feature data; subsequently, the fused feature data is quantitatively analyzed through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data, so as to match the corresponding health problems and health development trends in a preset risk index database based on the obtained health risk score and the user's data. Finally, a health assessment report can be generated according to the obtained health problems and health development trends, and in order to facilitate the understanding of users who do not understand medical expertise, the professional medical terms in the health assessment report can be converted into corresponding target phrases through natural language processing. In this way, by applying the large model technology, the subtle patterns in the data can be captured and learned, so as to more accurately predict the individual's health status and potential risks, and by analyzing the multi-dimensional health data of the individual, customized health risk assessment and intervention suggestions can be provided for each patient, realizing precision medicine and personalized health management. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.
[0045] Figure 1 It is a flowchart of a medical health risk assessment method based on a large model disclosed in this application;
[0046] Figure 2 It is a flowchart of data preprocessing disclosed in this application;
[0047] Figure 3 A flowchart of feature fusion disclosed in this application;
[0048] Figure 4 A flowchart of health risk assessment disclosed in this application;
[0049] Figure 5 A timing diagram of medical health risk assessment based on a large model disclosed in this application;
[0050] Figure 6 A schematic structural diagram of a medical health risk assessment device based on a large model disclosed in this application;
[0051] Figure 7 A structural diagram of an electronic device disclosed in this application. Detailed implementation manners
[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0053] In the prior art, the medical health risk assessment of medical users is usually achieved by artificial methods such as questionnaires, physical examination data, and doctors' clinical experience. To a certain extent, these methods can help identify the health status and potential risks of individuals, but there are still great limitations, such as the inability to achieve large-scale automated processing, low analysis efficiency, and difficulty in discovering complex health patterns.
[0054] To overcome the above problems, this application discloses a medical health risk assessment method, device, equipment, and medium based on a large model, which can more accurately and quickly predict the health status and potential risks of individuals by combining large model technology.
[0055] See Figure 1 As shown, an embodiment of the present invention discloses a medical health risk assessment method applied to a medical health risk assessment system, including:
[0056] Step S11: Integrate and preprocess the health data from multiple channels of the target object, and perform feature extraction and feature fusion on the preprocessed multi-channel health data obtained to obtain fused feature data.
[0057] In this embodiment, in order to perform a medical health risk assessment on the target object, such as Figure 2As shown, it is necessary to first collect the health data of various channels of the target user, and then preprocess the collected multi-channel health data. Specifically, the health data of the target object from several channels can be summarized through a preset data receiving framework to obtain multi-channel health data corresponding to the target object. The preset data receiving framework is a unified data access framework developed in advance, which is adapted to a variety of medical devices and platform interfaces to realize the integration of user health data such as Electronic Health Records (EHR), laboratory test results, medical imaging data, and wearable device monitoring data. And when data is transmitted, the data needs to be encrypted to ensure data security. The encryption algorithm can use strong encryption algorithms such as AES (Advanced Encryption Standard).
[0058] Furthermore, the multi-channel health data can be filtered at multiple levels to eliminate abnormal data in the multi-channel health data, and data imputation can be performed on the obtained multi-channel health data after elimination to obtain the multi-channel data after imputation. Specifically, when performing data filtering, abnormal data in the multi-channel health data can be first eliminated. For example, the multi-channel health data can be filtered through the preset physiological limit values of several health indicators. For example, if the blood glucose value exceeds 33.3 mmol / L, this data has exceeded the normal physiological limit, so it can be considered that the data abnormality is caused by an abnormal collection instrument. Therefore, this data can be eliminated. And each health indicator corresponds to a corresponding normal data range. Therefore, the corresponding data can be filtered according to the normal data range corresponding to each indicator to obtain the first filtered data. Then, the first abnormal data with contradictory combinations of indicators in the first filtered data can be identified through machine learning algorithms. For example, the contradictory combination of "extremely high blood glucose, but extremely low BMI index" can be identified through the isolation Forest algorithm. This situation may be a data entry error, so it can be directly eliminated. The first abnormal data existing in the first filtered data is eliminated to obtain the second filtered data. Finally, the second abnormal data with logical contradictions in the second filtered data can be screened through the constructed preset knowledge graph. For example, data with logical contradictions such as "the user is five years old, but the prostate specific antigen exceeds the standard" can be screened through the knowledge graph. Then, the second abnormal data with logical contradictions in the second filtered data is eliminated to obtain the multi-channel health data after elimination.
[0059] After obtaining the multi-channel healthy data after culling, it is necessary to perform data imputation on the multi-channel healthy data after culling to obtain the imputed multi-channel data, so as to fill in the missing values therein, improve the integrity and reliability of the data, and then convert the data format of the imputed multi-channel data into a preset data format. For example, convert data from different sources and formats into a unified data model, including the unification of timestamps, the conversion of measurement units, the consistency of coding rules, etc., to ensure the consistency and comparability of the data in subsequent processing. Finally, it is necessary to perform data augmentation on the data. For example, perform data augmentation methods such as rotation and scaling on the image data therein to increase data diversity, and finally obtain the preprocessed multi-channel healthy data. Moreover, the key features in the data can be labeled to ensure the accuracy of feature extraction. It should be noted that after preprocessing, it is necessary to verify the data quality to determine whether the preprocessed multi-channel healthy data meets the data quality conditions. The data integrity and accuracy can be tracked in real time through the established data quality monitoring system to ensure the data quality.
[0060] Furthermore, in order to integrate the key information from different sources and different types of data to provide a more comprehensive and in-depth health risk assessment, it is necessary to perform feature extraction and feature fusion on the data to obtain the fused feature data. It should be noted that feature extraction and feature fusion are processes of determining features. As Figure 3 shown, it includes feature extraction and fusion, feature selection and optimization, feature representation and encoding. Among them, for feature extraction and fusion, it is necessary to identify medical entities in the preprocessed multi-channel healthy data through a preset medical dictionary and regular expressions, and construct several medical entity triples based on the medical entities through a preset entity rule template. Then, perform encoding mapping on several medical entity triples respectively through a preset medical knowledge base to convert several medical entity triples into standardized medical features. Then, perform fusion, feature selection, and semantic enhancement and conversion on the obtained features. By selecting the features that can best reflect the essence of the problem, reduce the influence of noise and redundant information, and thus improve the accuracy of the model. Finally, feature encoding can be performed to make the finally retained features convenient for model processing, and finally obtain the encoded fused feature data.
[0061] In this way, through the preprocessing, feature extraction, and feature fusion of the data, the noise and outliers in the data can be removed, the risk of model overfitting can be reduced, the generalization ability of the model can be improved, and by fusing multiple features together, the data can be analyzed from different perspectives to obtain more comprehensive and accurate results, and multiple features can be transformed into a form that is easier to interpret, thereby improving the interpretability of the model.
[0062] Step S12: Quantitatively analyze the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each indicator in the fused feature data.
[0063] In this embodiment, before quantitatively analyzing through the model, it is necessary to confirm that a preset risk index database has been constructed so that after obtaining the corresponding health risk score, it can be matched in the preset risk index database to obtain the health problems and health development trends corresponding to the target object. The risk index database is used to identify factors related to specific health risks, including physiological and biochemical indicators, lifestyle factors, socioeconomic factors, environmental factors, and mental health status. These indicators are integrated into the risk index database and associated with known health risks and disease outcomes to provide data support for subsequent risk assessments.
[0064] Furthermore, it is necessary to quantitatively analyze the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each indicator in the fused feature data. Specifically, the fused feature data can be input into the preset health risk assessment model to quantitatively analyze the fused feature data through the preset health risk assessment model, obtain the disease probability corresponding to each indicator in the fused feature data, and when generating a health score through quantitative analysis, different formulas need to be used for each different indicator. For example, the ASCVD (Atherosclerotic Cardiovascular Disease) formula can be used to calculate the 10-year arteriosclerosis risk, and the HOMA-IR (Homeostasis Model Assessment-Insulin Resistance Index, HOMA is Homeostasis model assessment) formula can be used when calculating the insulin resistance index. Then, the sigmoid function is used to convert the model output into a probability, and according to the value corresponding to each disease probability, a score is assigned to it in the preset gradient scoring level for each indicator. Finally, the score value assigned to each indicator is multiplied by its corresponding weight to obtain the score value of each indicator. Finally, the scores of each indicator are added together to obtain the final health risk score. In this way, by analyzing the corresponding feature data of the user through the constructed model, the accurate analysis of the user's health can be ensured, and the accuracy of the health score can be guaranteed.
[0065] Step S13: Match according to the health risk score and the user data corresponding to the target object in the preset risk index database to obtain the health problems and health development trends corresponding to the target object.
[0066] In this embodiment, it is necessary to match according to the health risk score and the user data corresponding to the target object in the preset risk index database to obtain the corresponding health development trend of the target object. Specifically, as Figure 4 shown, it is necessary to combine the health risk score output by the model and the user data of the user to generate a health development trend including health suggestions. The health suggestions can be obtained by matching in the preset risk index database. It is possible to match in the preset risk index database according to the user data and the health risk score to match the target disease corresponding to the health risk score corresponding to each index from the preset risk index database, and then perform corresponding analysis in combination with the user data of the user to generate the health development trend of the target object. And the user data includes physiological data, psychological data, and life data, and the life data includes lifestyle and socioeconomic conditions. It should be noted that the latest medical research and clinical guidelines are integrated in the preset risk index database. In this way, after obtaining the health risk score of the user, it is possible to directly match in the preset risk index database to obtain the corresponding health problems and health development trends of the target object, and then directly obtain the health status of the user.
[0067] Step S14: Generate a health assessment report based on the health problems and the health development trend, and convert the professional medical terms in the health assessment report into corresponding target phrases through natural language processing to obtain a target health assessment report.
[0068] In this embodiment, a health assessment report can be generated according to the health problems and health development trend of the user. For example, the health problems are overweight, elevated low-density cholesterol, and accompanied by mild fatty liver, and the corresponding health development trend is that it may develop into hypertension and diabetes. Therefore, corresponding intervention measures and expected health improvement effects can be provided for the user according to the health problems and health development trend of the user. Finally, a health assessment report is generated based on the health problems, health development trend, provided intervention measures, and expected health improvement effects.
[0069] It should be noted that when generating a health assessment report, in order to avoid the inconvenience of professional medical data in the report for users to understand, professional medical terms in the health assessment report can be converted into corresponding target phrases through natural language processing. Specifically, the health assessment report can be segmented through a preset word segmentation tool to obtain several texts to be converted, and then the professional medical terms in the several texts to be converted can be converted into unified target medical terms in the preset medical dictionary to obtain the first target health assessment report. Finally, medical explanation information corresponding to the target medical terms is matched from the preset medical information database and added to the first target health assessment report to obtain the second target health assessment report. Among them, the first target health assessment report can be provided to the corresponding physician for review. All medical terms in the first target health assessment report are replaced with unified medical terms to facilitate the physician's review. The second target health assessment report is added with corresponding medical explanation information to facilitate users without professional medical knowledge to view.
[0070] Furthermore, as Figure 4 shown, the medical and health risk assessment system also provides an interactive user interface, which can provide personalized views, display the patient's health risk assessment results, recommended intervention measures, and expected health improvement effects. At the same time, the interface also includes interactive tools, such as a health risk self-assessment tool, a personalized health plan generator, etc., enabling patients to adjust the recommendations according to their own situations and achieve more effective self-management.
[0071] Furthermore, the current health status and risk assessment evaluation corresponding to the target health assessment report feedback by the target object can be collected and saved as evaluation accuracy data in the local risk assessment database. When the model evaluation time node is reached, the historical evaluation accuracy data for the preset time period corresponding to the model evaluation time node in the risk assessment database is determined. Then, it is determined whether model performance optimization is required based on the historical evaluation accuracy data. If the historical evaluation accuracy data indicates that the user satisfaction reaches the preset satisfaction threshold, no model performance optimization is required. If the historical evaluation accuracy data indicates that the user satisfaction does not reach the preset satisfaction threshold, model performance optimization is required.
[0072] It can be seen that through the method of this application, after receiving the multi-channel health data of the target object, it can be integrated and preprocessed, and feature extraction and feature fusion are performed on the obtained preprocessed multi-channel health data to obtain the fused feature data; subsequently, a preset health risk assessment model is used to perform quantitative analysis on the fused feature data to determine the health risk score corresponding to each index in the fused feature data, so as to match the corresponding health problems and health development trends in the preset risk index database based on the obtained health risk scores and the user's data. Finally, a health assessment report can be generated according to the obtained health problems and health development trends, and in order to facilitate the understanding of users who do not understand medical expertise, the professional medical terms in the health assessment report can be converted into corresponding target phrases through natural language processing. In this way, on the one hand, by analyzing a large amount of medical data through deep learning algorithms, the system can more accurately predict an individual's health risks, and the automated data integration and preprocessing improve the efficiency of data processing, ensuring the quality and consistency of the data. Through standardized processing, the unification of data from different sources and formats is achieved, improving the comparability of data and the accuracy of analysis; on the other hand, through deep learning fusion technology and feature selection and optimization, the expression ability of features and the discrimination of the model are improved, and the model can generate a score for the user's current health status, enabling the user to more intuitively understand the current health situation; on the other hand, through the automated and intelligent assessment process, this system reduces the workload of medical professionals and improves the efficiency of medical services.
[0073] See Figure 5As shown in the figure, the following is the specific process of medical and health risk assessment through the method of this application. First, it is necessary to perform data integration and preprocessing on the multi-channel health data of the target user, and then perform feature extraction and data augmentation on the obtained preprocessed multi-channel health data. If the obtained preprocessed data meets the preset data quality requirements, feature fusion is performed on the data. If the preset data quality requirements cannot be met, the data preprocessing process is returned, and the data preprocessing is continued until the preset data quality requirements are met. When performing feature fusion, it is necessary to perform feature fusion, feature selection, and semantic enhancement and transformation on the features. If the obtained features do not meet the preset validity requirements, the feature fusion process is returned. If the features meet the preset validity requirements, the fused feature data is quantitatively analyzed through a preset health risk assessment model, and it is necessary to ensure that the model has been constructed before performing the quantitative analysis. By quantitatively analyzing the fused feature data through a preset health risk assessment model, the health risk score corresponding to each indicator in the fused feature data can be determined. Finally, based on the obtained health risk score, a match is made in the preset risk indicator database in combination with the user data to obtain the health problems and health development trends corresponding to the target object, and corresponding intervention measures and expected health improvement effects are generated. Finally, a health assessment report is generated based on the health problems, health development trends, provided intervention measures, and expected health improvement effects. And it is possible to regularly determine whether it is necessary to optimize the performance of the model according to the current health status and risk assessment evaluation feedback by the user.
[0074] See Figure 6 As shown in the figure, an embodiment of the present invention discloses a medical and health risk assessment device based on a large model, which is applied to a medical and health risk assessment system and includes:
[0075] A feature processing module 11, configured to perform data integration and preprocessing on the multi-channel health data of the target object, and perform feature extraction and feature fusion on the obtained preprocessed multi-channel health data to obtain fused feature data;
[0076] A risk scoring module 12, configured to quantitatively analyze the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each indicator in the fused feature data;
[0077] A health evaluation module 13, configured to perform a match in a preset risk indicator database according to the health risk score and the user data corresponding to the target object to obtain the health problems and health development trends corresponding to the target object;
[0078] An evaluation report generation module 14, configured to generate a health evaluation report based on the health problems and the health development trend, and convert the professional medical terms in the health evaluation report into corresponding target phrases through natural language processing to obtain a target health evaluation report.
[0079] In this embodiment, the health data of the target object from multiple channels can be integrated and preprocessed, and feature extraction and feature fusion are performed on the preprocessed multi-channel health data to obtain fused feature data; the fused feature data is quantitatively analyzed through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data; according to the health risk score and the user data corresponding to the target object, a match is made in a preset risk index database to obtain the health problems and health development trend corresponding to the target object; a health evaluation report is generated based on the health problems and the health development trend, and the professional medical terms in the health evaluation report are converted into corresponding target phrases through natural language processing to obtain a target health evaluation report. It can be seen that after receiving the multi-channel health data of the target object, it can be integrated and preprocessed, and feature extraction and feature fusion are performed on the preprocessed multi-channel health data to obtain fused feature data; then, the fused feature data is quantitatively analyzed through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data, so as to match the corresponding health problems and health development trend in a preset risk index database based on the obtained health risk score and the user's data. Finally, a health evaluation report can be generated according to the obtained health problems and health development trend, and in order to facilitate the understanding of users who do not understand medical expertise, the professional medical terms in the health evaluation report can be converted into corresponding target phrases through natural language processing. In this way, by applying large model technology, subtle patterns in the data can be captured and learned, so as to more accurately predict the individual's health status and potential risks, and by analyzing the multi-dimensional health data of the individual, customized health risk assessment and intervention suggestions can be provided for each patient, realizing precision medicine and personalized health management.
[0080] In some embodiments, the feature processing module 11 may specifically include:
[0081] A data summarization sub-module, configured to summarize the health data of the target object from several channels through a preset data reception framework to obtain multi-channel health data corresponding to the target object;
[0082] The first data processing sub-module is used to perform multi-layer data filtering on the multi-channel health data to eliminate abnormal data in the multi-channel health data, and perform data interpolation on the obtained multi-channel health data after elimination to obtain interpolated multi-channel data;
[0083] The second data processing sub-module is used to convert the data format of the interpolated multi-channel data into a preset data format, and perform data augmentation on the obtained converted data to obtain preprocessed multi-channel health data;
[0084] The triple construction sub-module is used to identify medical entities in the preprocessed multi-channel health data through a preset medical dictionary and regular expressions, and construct a number of medical entity triples based on the medical entities through a preset entity rule template;
[0085] The feature fusion sub-module is used to perform encoding mapping on the number of medical entity triples through a preset medical knowledge base respectively, so as to convert the number of medical entity triples into standardized medical features, and perform feature fusion on the obtained number of standardized medical features to obtain fused feature data.
[0086] In some embodiments, the first data processing sub-module may specifically include:
[0087] The first data filtering unit is used to perform first data filtering on the multi-channel health data through a preset number of physiological limit values of health indicators to obtain first filtered data;
[0088] The second data filtering unit is used to identify first abnormal data with index contradiction combinations in the first filtered data through a preset machine learning algorithm, and eliminate the first abnormal data to obtain second filtered data;
[0089] The data elimination unit is used to screen out second abnormal data with logical contradictions in the second filtered data by using a preset knowledge graph, and eliminate the second abnormal data to obtain the multi-channel health data after elimination.
[0090] In some embodiments, the risk scoring module 12 may specifically include:
[0091] The quantitative analysis unit is used to input the fused feature data into a preset health risk assessment model, so as to perform quantitative analysis on the fused feature data through the preset health risk assessment model to obtain the disease probability corresponding to each index in the fused feature data;
[0092] The data conversion unit is used to map the disease probability through a preset piecewise linear function to convert the disease probability into a corresponding health risk score.
[0093] In some embodiments, the health evaluation module 13 may specifically include:
[0094] The disease matching unit obtains the user data of the target object, and matches according to the health risk score and the user data in the preset risk index database, so as to match the target diseases corresponding to the health risk score corresponding to each index from the preset risk index database;
[0095] The health trend prediction unit is used to analyze the user data and each target disease to generate the health development trend of the target object;
[0096] Wherein, the user data includes physiological data, psychological data and life data.
[0097] In some embodiments, the evaluation report generation module 14 may specifically include:
[0098] The text splitting unit is used to perform word segmentation processing on the health evaluation report through a preset word segmentation tool to obtain a number of texts to be converted;
[0099] The word replacement unit is used to convert the professional medical terms in the number of texts to be converted into the unified target medical terms in the preset medical dictionary through the preset medical dictionary to obtain the first target health evaluation report;
[0100] The information adding unit is used to match the medical explanation information corresponding to the target medical term from the preset medical information database, and add the medical explanation information to the first target health evaluation report to obtain the second target health evaluation report.
[0101] In some embodiments, the medical and health risk assessment device based on the large model may further include:
[0102] The data storage unit is used to collect the current health status and risk assessment evaluation corresponding to the target health evaluation report feedback by the target object, and store the current health status and risk assessment evaluation as evaluation accuracy data in the local risk assessment database;
[0103] The model optimization unit is used to determine the historical evaluation accuracy data of the preset time period corresponding to the model evaluation time node in the risk assessment database if the model evaluation time node is reached, so as to determine whether model performance optimization is required based on the historical evaluation accuracy data.
[0104] Furthermore, an embodiment of the present application also discloses an electronic device, Figure 7It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment. The content in the figure should not be regarded as any limitation on the scope of use of this application.
[0105] Figure 7 This is a schematic structural diagram of an electronic device 20 provided by an embodiment of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the method for medical and health risk assessment based on a large model disclosed in any of the foregoing embodiments. Additionally, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0106] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of this application, and no specific limitation is imposed on it here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application needs, and no specific limitation is made here.
[0107] In addition, as a carrier for resource storage, the memory 22 may be a read-only memory, a random access memory, a magnetic disk, or an optical disc, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.
[0108] Among them, the operating system 221 is used to manage and control each hardware device on the electronic device 20 and the computer program 222, and it may be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the method for medical and health risk assessment based on a large model executed by the electronic device 20 disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program that can be used to complete other specific tasks.
[0109] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the method for medical and health risk assessment based on a large model disclosed above. For the specific steps of this method, reference may be made to the corresponding content disclosed in the foregoing embodiments, and details will not be repeated here.
[0110] In the present specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.
[0111] Those skilled in the art can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.
[0112] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.
[0113] Finally, it should also be noted that in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.
[0114] The technical solutions provided in this application have been introduced in detail above. Specific examples are used herein to elaborate on the principles and implementation manners of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.
Claims
1. A medical and health risk assessment method based on a large model, characterized in that, Applied to medical and health risk assessment systems, including: Performing data integration and preprocessing on the health data of the target object from multiple channels, and performing feature extraction and feature fusion on the obtained preprocessed multi-channel health data to obtain fused feature data; Performing quantitative analysis on the fused feature data through a preset health risk assessment model to determine a health risk score corresponding to each indicator in the fused feature data; Matching the health risk score and the user data corresponding to the target object in a preset risk indicator database to obtain the health problems and health development trends corresponding to the target object; A health assessment report is generated based on the health problems and the health development trends, and professional medical terms in the health assessment report are converted into corresponding target phrases through natural language processing to obtain a target health assessment report.
2. The method for medical and health risk assessment based on a large model according to claim 1, wherein The data integration and preprocessing of the health data of the target object from multiple channels, and feature extraction and feature fusion of the obtained preprocessed multi-channel health data to obtain fused feature data, include: Aggregating the health data of the target object from several channels through a preset data receiving framework to obtain multi-channel health data corresponding to the target object; Performing multi-layer data filtering on the multi-channel health data to eliminate abnormal data in the multi-channel health data, and performing data interpolation on the obtained multi-channel health data after elimination to obtain interpolated multi-channel data; Converting the data format of the interpolated multi-channel data into a preset data format, and performing data enhancement on the converted data to obtain preprocessed multi-channel health data; Identify medical entities in the preprocessed multi-channel health data through a preset medical dictionary and a regular expression, and construct a plurality of medical entity triples based on the medical entities through a preset entity rule template; The plurality of medical entity triplets are respectively coded and mapped by using a preset medical knowledge base to convert the plurality of medical entity triplets into standardized medical features, and feature fusion is performed on the obtained plurality of standardized medical features to obtain fused feature data.
3. The method for medical and health risk assessment based on large models according to claim 2, wherein The performing multi-layer data filtering on the multi-channel health data to remove abnormal data in the multi-channel health data includes: Performing a first data filtering on the multi-channel health data by using a plurality of preset health indicator physiological limit values to obtain first filtered data; Identifying first abnormal data having a contradictory combination of indicators in the first filtered data by using a preset machine learning algorithm, and removing the first abnormal data to obtain second filtered data; The preset knowledge graph is used to screen the second abnormal data with logical contradictions in the second filtered data, and the second abnormal data is eliminated to obtain the eliminated multi-channel health data.
4. The method for medical and health risk assessment based on a large model according to claim 1, wherein The quantitative analysis of the fused feature data by using a preset health risk assessment model to determine the health risk score corresponding to each indicator in the fused feature data includes: Input the fused feature data into a preset health risk assessment model to perform quantitative analysis on the fused feature data through the preset health risk assessment model, and obtain the disease probability corresponding to each index in the fused feature data; Map the disease probability through a preset piecewise linear function to convert the disease probability into a corresponding health risk score.
5. The method for medical and health risk assessment based on a large model according to claim 1, wherein The matching in the preset risk index database according to the health risk score and the user data corresponding to the target object to obtain the health problems and health development trends corresponding to the target object includes: Obtain the user data of the target object, and perform matching in the preset risk index database according to the health risk score and the user data to match the target diseases corresponding to the health risk scores corresponding to each index from the preset risk index database; Analyze the user data and each of the target diseases to generate the health development trend of the target object; Wherein, the user data includes physiological data, psychological data, and life data.
6. The method for medical and health risk assessment based on a large model according to claim 1, wherein The conversion of the professional medical terms in the health assessment report into corresponding target phrases through natural language processing to obtain a target health assessment report includes: Perform word segmentation processing on the health assessment report through a preset word segmentation tool to obtain several texts to be converted; Convert the professional medical terms in the several texts to be converted into unified target medical terms in the preset medical dictionary through the preset medical dictionary to obtain a first target health assessment report; Match the medical explanation information corresponding to the target medical terms from the preset medical information database, and add the medical explanation information to the first target health assessment report to obtain a second target health assessment report.
7. The method for medical and health risk assessment based on a large model according to any one of claims 1 to 6, characterized in that It also includes: Collect the current health status and risk assessment evaluation corresponding to the target health assessment report feedback by the target object, and save the current health status and risk assessment evaluation as evaluation accuracy data to the local risk assessment database; If the model evaluation time node is reached, determine the historical evaluation accuracy data of the preset time period corresponding to the model evaluation time node in the risk assessment database, and determine whether model performance optimization is required based on the historical evaluation accuracy data.
8. A medical and health risk assessment device based on a large model, characterized in that, Applied to a medical health risk assessment system, including: A feature processing module for performing data integration and preprocessing on the health data of the target object from multiple channels, and performing feature extraction and feature fusion on the obtained preprocessed multi-channel health data to obtain fused feature data; A risk scoring module for performing quantitative analysis on the fused feature data through a preset health risk assessment model to determine the health risk score corresponding to each index in the fused feature data; A health evaluation module for performing matching in the preset risk index database according to the health risk score and the user data corresponding to the target object to obtain the health problems and health development trends corresponding to the target object; An evaluation report generation module, configured to generate a health evaluation report based on the health problems and the health development trend, and convert the professional medical terms in the health evaluation report into corresponding target phrases through natural language processing to obtain a target health evaluation report.
9. An electronic device, characterized in that, Comprising: A memory for storing a computer program; A processor for executing the computer program to implement the large model-based medical and health risk assessment method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, For storing a computer program, wherein when the computer program is executed by a processor, the large model-based medical and health risk assessment method according to any one of claims 1 to 7 is implemented.