Intelligent treatment system based on health medical big data

Abstract

The invention relates to the field of health medical big data, and particularly discloses an intelligent treatment system based on health medical big data, which comprises the following steps of: firstly, acquiring an electronic health record of a user, and physiological monitoring data and social environment factor data of the user acquired by wearable equipment; and feature extraction and correlation analysis are carried out on the three parts, and finally, personalized regulation and control suggestions for health of the user are generated through a generator, so that the user is helped to more effectively adjust the lifestyle and reduce the risk of chronic diseases, a health management strategy with prevention as a main part and intervention as an auxiliary part is realized, and the health treatment effect is improved.

Description

Technical Field

[0001] The present application relates to the field of health and medical big data, and more specifically, to an intelligent governance system based on health and medical big data. Background Art

[0002] Health management aims to prevent and control the occurrence and development of diseases, reduce medical expenses, and improve the quality of life. Chronic diseases are one of the management systems. Chronic diseases are a very large clinical concept. They cover almost all the most important types of diseases such as the cardiovascular system, respiratory system, endocrine system, and digestive system, and involve the widest range of patients. Such diseases usually have an insidious onset and a long progression process after onset. There may be correlations between various chronic diseases.

[0003] However, the traditional medical model is often unable to effectively achieve early warning and personalized intervention for chronic diseases due to its lag.

[0004] Therefore, an intelligent governance system based on health and medical big data is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an intelligent governance system based on health and medical big data, which first obtains the user's electronic health records, the user's physiological monitoring data collected by wearable devices, and the user's social environmental factor data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three, and finally generates personalized health regulation suggestions for users through a generator, thereby helping users to adjust their lifestyles more effectively and reduce the risk of chronic diseases, thereby realizing a health management strategy of prevention first and intervention second, and improving the effect of health governance.

[0006] According to one aspect of the present application, there is provided an intelligent governance system based on health and medical big data, which includes:

[0007] Health and medical data collection module, used to obtain user electronic health records, user physiological monitoring data collected by wearable devices, and user social environment factor data;

[0008] A health and medical data processing module, used to extract a user health global correlation feature vector and a user social environment factor semantic correlation feature vector from the user electronic health record, the user physiological monitoring data collected by the wearable device and the user social environment factor data;

[0009] The user health regulation suggestion generating module is used to generate a personalized regulation suggestion for the user health based on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector.

[0010] Compared with the existing technology, the present application provides an intelligent governance system based on health and medical big data, which first obtains the user's electronic health records, the user's physiological monitoring data collected by wearable devices, and the user's social environmental factor data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a generator is used to generate personalized health regulation suggestions for the user, thereby helping users to adjust their lifestyles more effectively and reduce the risk of chronic diseases, thereby realizing a health management strategy of prevention first and intervention second, and improving the effectiveness of health governance. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0012] Figure 1 This is a block diagram of an intelligent governance system based on health and medical big data according to an embodiment of the present application.

[0013] Figure 2 This is a block diagram of a health and medical data processing module in an intelligent governance system based on health and medical big data according to an embodiment of the present application.

[0014] Figure 3 This is a block diagram of a user physiological monitoring feature extraction unit in an intelligent management system based on health and medical big data according to an embodiment of the present application.

[0015] Figure 4 This is a block diagram of a user health regulation suggestion generation module in an intelligent management system based on health and medical big data according to an embodiment of the present application.

[0016] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0018] Figure 1 Schematic diagram of the intelligent management system based on health and medical big data in the embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the intelligent governance system 100 based on health and medical big data includes: a health and medical data acquisition module 110, which is used to obtain user electronic health records, user physiological monitoring data collected by wearable devices, and user social environment factor data; a health and medical data processing module 120, which is used to extract user health global correlation feature vectors and user social environment factor semantic correlation feature vectors from the user electronic health records, the user physiological monitoring data collected by wearable devices, and the user social environment factor data; a user health regulation suggestion generation module 130, which is used to generate user health personalized regulation suggestions based on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector.

[0019] In the above-mentioned intelligent governance system 100 based on health and medical big data, the health and medical data acquisition module 110 is used to obtain user electronic health records, user physiological monitoring data collected by wearable devices, and user social environmental factor data. It should be understood that health management aims to prevent and control diseases, reduce medical expenses, and improve the quality of life. In this system, chronic diseases are an important field, covering major diseases of multiple systems such as cardiovascular, respiratory, endocrine, and digestive systems, involving a wide range of patient groups. These diseases usually have an insidious onset, a long course of disease, and there may be correlations between different chronic diseases. However, traditional medical models are often delayed in response, making it difficult to effectively achieve early warning and personalized intervention of chronic diseases. Therefore, in the technical solution of the present application, by obtaining user electronic health records, user physiological monitoring data collected by wearable devices, and user social environmental factor data, and combining deep learning technology, personalized health regulation suggestions are generated to help users adjust their lifestyles more effectively and reduce the risk of chronic diseases. This method implements a health management strategy with prevention as the main and intervention as the auxiliary, thereby improving the effect of health governance.

[0020] Specifically, the user's electronic health record contains a wealth of medical information, such as medical history, diagnosis results, medication use, and examination reports. These records can provide users with past and current health conditions and disease history, providing basic data for the health management system. Physiological monitoring data collected by wearable devices, including heart rate, blood pressure, number of steps, sleep patterns, etc., provide real-time health information in users' daily lives. These data are highly real-time and continuous, and can help monitor users' physical conditions, detect abnormal changes, and respond to health problems in a timely manner. The user's social environmental factor data includes living environment, socioeconomic status, psychological state, etc., which have an important impact on health. These data help the system understand the user's life background, so that these external factors can be taken into account when formulating health management plans, and more comprehensive intervention measures can be provided. The combined use of these data can obtain a panoramic view of the user's health status, in which electronic health records provide historical and diagnostic information, physiological monitoring data provide real-time dynamic changes, and social environmental factor data supplement external factors that affect health. By integrating this information, the health management system can more accurately assess the user's health status, discover potential health risks, provide personalized health advice, and warn before the disease occurs, thereby significantly improving the effectiveness and efficiency of health management.

[0021] In the above-mentioned intelligent governance system 100 based on health and medical big data, the health and medical data processing module 120 is used to extract the user's health global association feature vector and the user's social environment factor semantic association feature vector from the user's electronic health record, the user's physiological monitoring data collected by the wearable device, and the user's social environment factor data. It should be understood that extracting the health global association feature vector and the social environment factor semantic association feature vector enables the health management system to comprehensively consider the individual's health status and the external environment, and provide more comprehensive and personalized health interventions. This method not only improves the accuracy of health management, but also helps users detect health risks at an early stage and take appropriate preventive measures, thereby improving overall health levels and quality of life.

[0022] Figure 2 FIG. 1 is a block diagram of a health and medical data processing module in an intelligent management system based on health and medical big data according to an embodiment of the present application. Figure 2As shown, in a specific embodiment of the present application, the health and medical data processing module 120 includes: a user electronic health record feature extraction unit 121, used to perform feature extraction on the user electronic health record to obtain a user electronic health record semantic feature vector; a user physiological monitoring feature extraction unit 122, used to perform feature extraction on the user physiological monitoring data collected by the wearable device to obtain a user physiological monitoring data text local association feature vector; a user health global feature association unit 123, used to associate the user electronic health record semantic feature vector with the user physiological monitoring data text local association feature vector to obtain the user health global association feature vector; a user social environment factor feature extraction unit 124, used to perform semantic encoding on the user social environment factor data to obtain the user social environment factor semantic association feature vector.

[0023] It should be understood that electronic health records (EHRs) contain the user's medical history, diagnosis results, treatment plans, and medication information. These data are the basis for understanding the user's health status, but the original records usually contain a large amount of complex information, which needs to be converted into structured feature vectors through feature extraction for further analysis and processing. The purpose of feature extraction is to convert the unstructured and semi-structured information in electronic health records into semantically meaningful feature vectors, so that they can be efficiently processed and analyzed in computer systems. That is, considering that electronic health records usually exist in text form, this information has a variety of expressions and complex contextual relationships, and it is very difficult to directly use these text data for analysis. Therefore, feature extraction methods are needed to simplify and standardize the data in order to extract useful information.

[0024] Furthermore, feature extraction of user physiological monitoring data collected by wearable devices is a key link in improving the accuracy and real-time performance of health management. These data usually include physiological indicators such as heart rate, number of steps, blood pressure, body temperature, etc., which can reflect the user's daily health status. However, the original data is often high-dimensional and time-series, and it may be very complicated to directly process and analyze these data. Therefore, it is necessary to convert them into meaningful feature vectors through feature extraction. These data often have time dependence and local correlation, that is, changes in data within a specific time period or under certain specific conditions will affect the health status. Through feature extraction, these local correlation features can be effectively captured, allowing the system to identify important health signals and trends.

[0025] Furthermore, the semantic feature vector of the electronic health record contains detailed information about the user's historical health status, such as medical history, diagnosis records, treatment plans, etc. This information can reveal the user's long-term health trends and potential disease risks. The local association feature vector of the physiological monitoring data text reflects the changes in the user's physiological state in a specific time period, such as heart rate fluctuations, changes in the number of steps, etc. These local features can capture the user's current health dynamics and monitor short-term changes in health status in real time. The process of associating these two feature vectors aims to integrate the user's long-term health history and current physiological state to generate a comprehensive health global association feature vector. Combine the long-term health information in the electronic health record with the short-term physiological changes in the physiological monitoring data. This fusion can provide a comprehensive perspective on the user's health status. For example, by combining long-term medical history and treatment records with current physiological data, the progression of the disease and the effectiveness of interventions can be more accurately assessed.

[0026] In particular, semantic coding of user social environmental factor data is a key step in achieving comprehensive personalized analysis and intervention in a health management system. Social environmental factors, such as place of residence, work environment, social support network, and economic status, have an important impact on personal health. By semantically coding these data, their potential impact on health can be effectively extracted and quantified, thereby providing more accurate health management strategies. The complexity of social environmental factors lies in that they usually involve multiple dimensions and levels. For example, living in an area with severe air pollution may have a negative impact on respiratory health, while the strength of the social support network may affect mental health. The original social environmental data usually exists in an unstructured or semi-structured form, such as free text descriptions, questionnaire results, or statistical data. The processing and analysis of these data need to be converted into structured semantic features for further analysis and application by the computer system. In a specific embodiment of the present application, the user social environmental factor data includes eating habits, exercise, smoking and drinking habits, air quality, climate change, and residential area.

[0027] In a specific embodiment of the present application, the user electronic health record feature extraction unit 121 includes: passing the user electronic health record through a user electronic health record long link text semantic understanding model to obtain a user electronic health record semantic sequence; passing the user electronic health record semantic sequence through a converter-based user electronic health record context long link encoder to obtain the user electronic health record semantic feature vector.

[0028] It should be understood that electronic health records (EHRs) contain a large amount of unstructured and structured data, such as medical records, diagnostic information, treatment records, and drug prescriptions. This information usually exists in the form of text, which contains rich context and semantic relationships. Long-link text semantic understanding models can effectively capture the long-distance dependencies and complex semantic information of text when processing these electronic health records. This model can convert electronic health records into semantic sequences by deeply understanding the contextual relationships in the text, so that the representation of information not only considers single data points, but also combines the coherence and context between data. Among them, the specific steps may include: 1. Data preprocessing: First, the text data in the electronic health records is preprocessed, including cleaning, labeling, and standardization. During the processing, redundant information and unstructured data in the original records may be sorted out so that the model can effectively process them. 2. Semantic understanding: Use long-link text semantic understanding models (such as BERT, GPT, etc.) to perform semantic analysis on the preprocessed data. These models can recognize and understand complex language patterns, contextual information, and implicit relationships in the text, thereby extracting semantic features. 3. Generate semantic sequences: Convert the semantic features extracted by the model into semantic sequences. These sequences are high-level representations of the content of the user's health record, which can capture the details and contextual relationships of the user's health information. 4. Context integration: Use semantic sequences for further analysis and decision support. This sequence not only contains static information, but also combines changes in time and context to help the system fully understand the user's health status and dynamic changes. In this way, deeper semantic information can be extracted from health records to identify potential health patterns and risks.

[0029] Furthermore, electronic health records contain a large amount of medical information, such as medical records, diagnoses, treatment plans, and drug prescriptions, which usually have complex contexts and long-range dependencies. Transformer-based encoders (such as the BERT series) can deeply understand the semantics and context of the text when processing this data. The advantage of the transformer model is that it can handle complex dependencies in long texts, and capture the connections between different information fragments through a multi-layer self-attention mechanism, thereby understanding the contextual information of the entire document. Among them, the multi-layer self-attention mechanism processing process includes: 1. Semantic sequence input: First, the semantic sequences extracted from the electronic health record are used as input. These sequences contain key information and contextual relationships in the user's health record. The input semantic sequence is usually the result of preprocessing and semantic analysis, which can better express the details of health information. 2. Context modeling: The transformer-based encoder will perform context modeling on the input semantic sequence. The transformer encodes each word in the sequence through a multi-head self-attention mechanism to capture the relationship and semantic information between different words. For example, it can understand the relationship between a symptom and a diagnosis result, or the connection between a treatment plan and a patient's response. 3. Generate feature vectors: After context modeling, the encoder generates a semantic feature vector. This vector is a highly compressed representation of the semantic information of the user's health record, which contains comprehensive features extracted from the text. These feature vectors not only reflect the information of a single record, but also take into account the context and long-term dependencies between records. In this way, the complex information in the electronic health record can be deeply understood and analyzed to reveal potential health problems and trends. Specifically, the embedding layer of the converter-based user electronic health record context long link encoder is used to convert the user electronic health record semantic sequence into an embedding vector to obtain a sequence of user electronic health record embedding vectors; the converter-based Bert model of the converter-based user electronic health record context long link encoder is used to perform global context semantic encoding on the sequence of user electronic health record embedding vectors to obtain multiple user electronic health record feature vectors; and the multiple user electronic health record feature vectors are cascaded to obtain the user electronic health record semantic feature vector.

[0030] Figure 3 FIG. 1 is a block diagram of a user physiological monitoring feature extraction unit in an intelligent management system based on health and medical big data according to an embodiment of the present application. Figure 3As shown, in a specific embodiment of the present application, the user physiological monitoring feature extraction unit 122 includes: a user physiological monitoring embedding coding subunit 1221, which is used to perform user physiological monitoring data embedding coding on the user physiological monitoring data collected by the wearable device to obtain a user physiological monitoring data word vector sequence; a user physiological monitoring two-dimensional arrangement subunit 1222, which is used to arrange the user physiological monitoring data word vector sequence in two dimensions into a user physiological monitoring data word embedding matrix; a user physiological monitoring convolutional coding subunit 1223, which is used to pass the user physiological monitoring data word embedding matrix through a user physiological monitoring data text convolutional neural network as a feature extractor to obtain a local correlation feature vector of the user physiological monitoring data text.

[0031] It should be understood that wearable devices usually provide a variety of physiological data including heart rate, number of steps, body temperature, sleep patterns, etc. These data are often highly structured and large, but for in-depth analysis and modeling, these data need to be converted into a form that can capture their potential patterns and relationships. Among them, the process of embedding coding of physiological monitoring data includes: 1. Data preprocessing: First, the original physiological data collected by the wearable device is cleaned and preprocessed. This includes removing noise, processing missing values, and standardizing the data format to make it suitable for embedding coding processing. 2. Data embedding: The key to embedding coding is to convert physiological monitoring data into "word vectors", which is a way to map data to a high-dimensional space so that the data can be effectively processed by computer models. Similar to word embedding in natural language processing (such as Word2Vec, GloVe), physiological data embedding can capture the potential relationships and features between data by mapping each data point (such as heart rate, number of steps, body temperature, etc.) into a vector space. This mapping not only retains the characteristics of the original data, but also converts it into a form that can be understood by the computer model. 3. Generate word vector sequence: Through embedding coding, physiological data is converted into a series of word vectors. Each word vector is a point in a high-dimensional space that represents the characteristics and context of a specific data point. For example, continuous heart rate data and step data will be mapped to adjacent word vectors, thereby preserving the continuity and change pattern of the time series data. 4. Application analysis and modeling: The generated word vector sequence can be used for various machine learning and data analysis tasks. For example, by analyzing the word vector sequence, the system can identify the user's health patterns, predict potential health risks, or optimize personalized health recommendations. Because word vectors can effectively represent complex relationships in the data, they can improve the performance and accuracy of the model. Through embedded coding, various physiological monitoring data are unified into a common vector space, which helps to identify associations and patterns between data.

[0032] Furthermore, user physiological monitoring data is usually continuous time series data, such as heart rate, number of steps, body temperature, etc. The data at each time point is converted into a word vector through embedding coding and represented as a high-dimensional vector. These word vectors are arranged in chronological order to form a one-dimensional word vector sequence. Converting these one-dimensional word vector sequences into a two-dimensional matrix usually involves mapping the time series and different data features into the rows and columns of the matrix. For example, each row of the matrix can represent a data vector at a time point, while each column can represent a feature dimension. This two-dimensional arrangement not only retains the sequential information in the time series, but also more clearly shows the relationship between different features in the matrix. By arranging the word vector sequence into a matrix, the model can use matrix operations (such as convolution, pooling, etc.) to capture patterns and changes in time series data. This matrix representation can better capture the temporal and spatial characteristics in the data, allowing the model to more effectively identify potential trends and anomalies when performing analysis.

[0033] Furthermore, the word embedding matrix of the user's physiological monitoring data is subjected to feature extraction through a convolutional neural network (CNN) to obtain local correlation feature vectors in order to efficiently mine and extract important patterns and features from the data. Convolutional neural networks are particularly suitable for processing matrix-structured data, such as two-dimensional embedding matrices, because they can automatically learn and extract local features in the data and capture spatial and local relationships in the data. The operation process of the convolutional neural network is as follows: 1. Data representation: After word embedding encoding, the user's physiological monitoring data is converted into a two-dimensional matrix. In this matrix, each row represents the data feature at a time point, and each column corresponds to a different dimension of the data. In this way, each element of the matrix is ​​part of the word vector and contains information about the time series and feature space. 2. Convolution operation: The convolutional neural network operates on the two-dimensional matrix by applying filters (or convolution kernels) through convolutional layers. The filter slides on the matrix and performs local convolution operations. This operation can extract local features from the matrix, such as short-term patterns of time series or local relationships between features. The data features learned by each filter are called convolution feature maps. 3. Local correlation extraction: The filters of the convolutional neural network can automatically identify and extract important local patterns in the matrix. For example, for user physiological monitoring data, these patterns may include sudden changes in heart rate, fluctuating patterns in the number of steps, etc. The convolution operation can capture these local features and further transform them nonlinearly through activation functions (such as ReLU) to highlight important features and suppress irrelevant information. 4. Feature aggregation: The feature map processed by the convolution layer will pass through the pooling layer (such as maximum pooling or average pooling) to further reduce the data dimension and retain the most significant features. The pooling operation allows the network to pay more attention to important local features when extracting features, while ignoring detailed noise, enhancing the generalization ability of the model. 5. Generate feature vectors: The feature map obtained by convolution and pooling operations will eventually be flattened and converted into a local correlation feature vector through a fully connected layer. The convolution operation focuses on the local area of ​​the data and can effectively capture important changes and patterns in the short term. Specifically, each layer of the user physiological monitoring data text convolutional neural network used as a feature extractor performs convolution processing, mean pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward pass of the layer, so that the last layer of the user physiological monitoring data text convolutional neural network used as a feature extractor outputs the local correlation feature vector of the user physiological monitoring data text, wherein the input of the user physiological monitoring data text convolutional neural network used as a feature extractor is the user physiological monitoring data word embedding matrix.

[0034] In a specific embodiment of the present application, the user health global feature association unit 123 includes: using std::vector data structure to define and store user electronic health record semantic feature vectors and user physiological monitoring data text local association feature vectors; calculating the transposed vector of the user physiological monitoring data text local association feature vector; using the matrixMultiply matrix multiplication function to perform a matrix product operation on the electronic health record semantic feature vector and the transposed vector; generating and outputting the user health global association feature vector.

[0035] The specific code is shown below.

[0036]

[0037]

[0038]

[0039] It should be understood that, first, the above code uses std::vector to define and store two main feature vectors, namely semanticVector and localFeatureVector. Among them, semanticVector represents the semantic features of the user's electronic health record, while localFeatureVector represents the local correlation features in the user's physiological monitoring data. The storage method of these two feature vectors ensures that the data can be operated and calculated in an efficient form. Then, the code handles the transposition of the local correlation feature vector of the user's physiological monitoring data text during the calculation process. In practical applications, if the local correlation feature vector of the user's physiological monitoring data text needs to participate in the operation in the form of a row vector, the transposition operation is necessary. Next, the above code uses a C++ function called matrixMultiply to perform a matrix multiplication operation. The core function of this function is to perform matrix multiplication calculation on the semantic feature vector of the user's electronic health record and the local correlation feature vector of the user's physiological monitoring data text. Matrix multiplication is a common operation in linear algebra, which is used to multiply and sum the elements of two matrices one by one to generate a new matrix. In this code, the result of the matrix multiplication is the user's health global correlation feature vector, which reflects the overall correlation of the health data. Finally, the calculated global correlation feature vector of user health is printed to the console through standard output. This step makes the calculation results directly viewable, which facilitates further data analysis and verification. This output method is particularly useful for debugging and result verification because it allows developers to quickly check the correctness of the calculation and the accuracy of the output data. The above code implements the global correlation feature calculation of user health data through efficient vector storage, transposition processing, matrix multiplication operations and result output. It provides a basis for further data analysis and health management, and can help researchers and developers gain in-depth understanding of the key correlation features in user health data.

[0040] In a specific embodiment of the present application, the user social environment factor feature extraction unit 124 includes: passing the user social environment factor data through a user social environment factor semantic encoder to obtain a plurality of user social environment factor feature vectors; and cascading the plurality of user social environment factor feature vectors into the user social environment factor semantic association feature vector.

[0041] It should be understood that social environmental factors, such as family background, occupation, education level and geographical location, have complex semantic information, which can significantly affect the user's decision-making and behavior patterns. The main function of the semantic encoder is to convert these complex, usually unstructured social environmental data into a format that can be effectively processed by the machine learning model, thereby improving the performance and analysis capabilities of the model. Specifically, the user social environmental factor data is segmented to obtain a user social environmental factor word sequence; the embedding layer of the user social environmental factor semantic encoder is used to map each user social environmental factor word in the user social environmental factor word sequence to a word embedding vector to obtain a sequence of user social environmental factor word embedding vectors; the converter-based Bert model of the user social environmental factor semantic encoder is used to perform global context semantic encoding on the sequence of user social environmental factor word embedding vectors to obtain multiple user social environmental factor feature vectors.

[0042] Furthermore, the social environment factors of users usually involve multiple aspects, such as family background, education level, occupation, and geographical location. These factors each contain unique information and features. By cascading different feature vectors, these independent social environment factors can be integrated together to form a comprehensive feature representation. This comprehensive information helps to more comprehensively capture the impact of users' social background and environment on their behavior and decision-making.

[0043] In the above-mentioned intelligent governance system 100 based on health and medical big data, the user health regulation suggestion generation module 130 is used to generate personalized regulation suggestions for user health based on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector. It should be understood that each user's health status and social environment are unique. The global correlation feature vector reveals the user's health trends and existing problems, while the semantic correlation feature vector shows the potential impact of the social environment on health. Combining these two types of features, the user's personalized needs can be accurately identified, such as whether they need to increase exercise, adjust their diet, improve their mental health, etc. Personalized health regulation suggestions can help users better adapt to and adjust to various challenges in the social environment while improving their health.

[0044] Figure 4 FIG. 1 is a block diagram of a user health regulation suggestion generation module in an intelligent management system based on health and medical big data according to an embodiment of the present application. Figure 4As shown, in a specific embodiment of the present application, the user health regulation suggestion generation module 130 includes: a user monitoring status feature fusion unit 131, which is used to perform back-propagation compensation based on feature fine-grained dimension offset on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector to obtain the user monitoring status feature vector; a user health personalized regulation unit 132, which is used to pass the user monitoring status feature vector through a generator to generate a user health personalized regulation suggestion.

[0045] It should be understood that by combining health data and social environment information, potential health risks can be identified more effectively. The fused feature vector can reveal these potential health risks and provide more accurate monitoring data so that preventive measures can be taken early.

[0046] In particular, in the technical solution of the present application, considering the different acquisition backgrounds and feature processing methods of the user health global correlation feature vector and the user social environmental factor semantic correlation feature vector, the feature representations of the user health global correlation feature vector and the user social environmental factor semantic correlation feature vector may have different scales, structures and semantic dimensions. For example, physiological monitoring data may reflect more specific physiological states, while social environmental data may reflect potential patterns that affect health. The difference in distribution between the two in the feature space leads to the offset of the fine-grained dimension of the feature, that is, in the fusion stage, the dimension and representation of the feature are not completely matched, which makes it difficult to directly perform effective fusion. In this way, the simple fusion method fails to fully consider the heterogeneity between different data sources, which may cause the local structure of the feature to be distorted. Specifically, if a simple weighting or splicing method is used in feature fusion, and the relative importance of each feature and its contribution to the health status are not fully considered, the generated user monitoring state feature vector may not effectively represent the true state of the user's health, resulting in local structural collapse, that is, the local correlation relationship of the feature fails or cannot accurately reflect the actual health status. Therefore, in the technical solution of the present application, a user monitoring state feature vector is obtained by performing back-propagation compensation based on feature fine-grained dimension shift on the user health global association feature vector and the user social environment factor semantic association feature vector.

[0047] Among them, the user health global correlation feature vector and the user social environment factor semantic correlation feature vector are subjected to back propagation compensation based on feature fine-grained dimension offset to obtain the user monitoring state feature vector, including: calculating the position difference between the user health global correlation feature vector and the user social environment factor semantic correlation feature vector to obtain the dimension offset differential feature vector; calculating the position point multiplication between the user health global correlation feature vector and the user social environment factor semantic correlation feature vector to obtain the dimension offset position point multiplication feature vector; calculating the position addition between the user health global correlation feature vector and the user social environment factor semantic correlation feature vector to obtain the dimension offset position addition feature vector; adding the dimension offset differential feature vector, The dimensional offset position dot product feature vector and the dimensional offset position addition feature vector are input into a one-dimensional convolution layer and maximum pooling is performed to obtain a dimensional offset convolutional coding pooling feature vector; the square root of the two norm of the dimensional offset convolutional coding pooling feature vector is weighted and added to the one norm of the dimensional offset convolutional coding pooling feature vector to obtain a dimensional offset factor; the square root of the dimensional offset position dot product feature vector and the length of the user health global association feature vector are added by position to obtain a dimensional offset intermediate vector; using the dimensional offset factor as the first weighting coefficient and the second weighted hyperparameter as the second weighting coefficient, the position-weighted sum of the dimensional offset position addition feature vector and the dimensional offset intermediate vector is calculated to obtain a user monitoring status feature vector.

[0048] Among them, the user health global correlation feature vector and the user social environment factor semantic correlation feature vector are back-propagated and compensated based on the feature fine-grained dimension offset to obtain the user monitoring state feature vector, including:

[0049]

[0050]

[0051]

[0052] Among them, V1 represents the global correlation feature vector of user health, V2 represents the semantic correlation feature vector of user social environment factors, represents positional subtraction, ⊙ represents positional multiplication, represents positional addition, conv1D represents a one-dimensional convolutional layer, MaxPool represents maximum pooling, V μ represents the dimension-shifted convolutional coding pooling feature vector, ||·||1 represents the first norm, ||·||2 represents the second norm, α represents the first weighted hyperparameter, τ represents the dimension-shift factor, L represents the length of the user health global association feature vector, β represents the second weighted hyperparameter, V fRepresents the user monitoring state feature vector.

[0053] In the technical solution of the present application, in the process of fusing the user health global correlation feature vector and the user social environment factor semantic correlation feature vector, the feature fine-grained dimension offset between the user health global correlation feature vector and the user social environment factor semantic correlation feature vector is taken into account, which may cause the fused user monitoring state feature vector to have local structural collapse. Based on this, in the technical solution of the present application, the user health global correlation feature vector and the user social environment factor semantic correlation feature vector are back-propagated and compensated based on the feature fine-grained dimension offset.

[0054] Specifically, based on the restricted expression of the foreground-background structure distinction of the high-dimensional manifold of the user health global association feature vector and the user social environment factor semantic association feature vector by the amplitude of the semantic difference vector between the user health global association feature vector and the user social environment factor semantic association feature vector, the fine-grained correspondence of key features between the user health global association feature vector and the user social environment factor semantic association feature vector is modeled, and then active fine-grained correspondence adjustment propagation is performed through the back-propagation displacement compensation information between the corresponding eigenvalues ​​of the user health global association feature vector and the user social environment factor semantic association feature vector, so as to avoid the imbalance at the feature vector level of the user health global association feature vector and the user social environment factor semantic association feature vector in a focused manner, thereby synchronizing the micro dimensions of the two feature vectors, ensuring that in the fusion process, the information between the feature vectors can be effectively aligned and combined, thereby avoiding abnormal accumulation of local structures.

[0055] Furthermore, the user monitoring state feature vector integrates the user's health data and social environment information to provide a comprehensive view of health status. However, these data are usually of high complexity and dimensionality and are not easy to understand or apply directly. The generator is able to convert these complex feature vectors into specific health recommendations, so that users can directly obtain clear and practical interventions, making it easier to take action. Using the user monitoring state feature vector as input, the generator can generate tailored recommendations based on the user's unique health status and social environment factors. Such personalized recommendations are based on the user's actual data, can more accurately meet their needs, and improve the effectiveness of interventions.

[0056] In summary, the embodiment of the present application first obtains the user's electronic health records, the user's physiological monitoring data collected by wearable devices, and the user's social environmental factor data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three, and finally uses a generator to generate personalized health regulation suggestions for the user, thereby helping the user to adjust their lifestyle more effectively and reduce the risk of chronic diseases, thereby realizing a health management strategy of prevention first and intervention second, and improving the effectiveness of health management.

[0057] As described above, the intelligent governance system 100 based on health and medical big data according to the embodiment of the present application can be implemented in various terminal devices. In one example, the intelligent governance system 100 based on health and medical big data can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent governance system 100 based on health and medical big data can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the intelligent governance system 100 based on health and medical big data can also be one of the many hardware modules of the terminal device.

[0058] Alternatively, in another example, the intelligent governance system 100 based on health and medical big data and the terminal device may also be separate devices, and the intelligent governance system 100 based on health and medical big data may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0059] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application.

[0060] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. Figure 5 , which shows a schematic diagram of the structure of an electronic device suitable for implementing a terminal device or a server in an embodiment of the present application. Figure 5As shown, the electronic device includes one or more processors, a communication unit, etc. The one or more processors are, for example: one or more central processing units (CPUs), and / or one or more graphics processing units (GPUs), etc. The processor can perform various appropriate actions and processes according to executable instructions stored in a read-only memory (ROM) or executable instructions loaded from a storage part to a random access memory (RAM). The communication unit may include but is not limited to a network card, and the network card may include but is not limited to an IB (Infiniband) network card. The processor may communicate with a read-only memory and / or a random access memory to execute executable instructions, and is connected to the communication unit through a bus and communicates with other target devices through the communication unit, thereby completing the corresponding operation of any system provided in the embodiments of the present application, for example, the primary shard manager shards the features in the feature queue; for any shard, the primary shard manager stores the any shard in the corresponding shard feature library, and records the mapping relationship between the any shard and the corresponding shard feature library in a mapping table; in response to a failure of the primary shard manager, a shard manager is reselected from at least one slave shard manager as the new primary shard manager; the new primary shard manager obtains shard information of the shard currently being written from all shard feature libraries, and recreates the mapping relationship of the shard currently being written based on the shard information of the shard currently being written in all shard feature libraries.

[0061] In addition, various programs and data required for the operation of the device can also be stored in the RAM. The CPU, ROM and RAM are connected to each other through a bus. In the case of RAM, the ROM is an optional module. The RAM stores executable instructions, or writes executable instructions into the ROM at runtime, and the executable instructions enable the processor to perform operations corresponding to any of the above-mentioned systems of the present invention. The input / output (I / O) interface is also connected to the bus. The communication unit can be integrated or can be configured to have multiple submodules (e.g., multiple IB network cards) and be linked on the bus.

[0062] The following components are connected to the I / O interface: an input section including a keyboard, a mouse, etc.; an output section including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive as needed so that a computer program read therefrom is installed into the storage section as needed.

[0063] If necessary, Figure 5The architecture shown is only an optional implementation method. In the specific practice, the above can be modified according to actual needs. Figure 5 The number and type of components can be selected, deleted, added or replaced; different functional components can be set up by separate settings or integrated settings, for example, the GPU and CPU can be set up separately or the GPU can be integrated on the CPU, the communication unit can be set up separately or integrated on the CPU or GPU, etc. These replaceable implementations all fall within the protection scope disclosed in the present invention.

[0064] In addition, an embodiment of the present invention further provides a computer program, including computer-readable codes. When the computer-readable codes are executed on a device, a processor in the device executes instructions for implementing each step in a data processing system of any of the above embodiments of the present invention.

[0065] In addition, an embodiment of the present invention further provides a computer-readable storage medium for storing computer-readable instructions, which, when executed, implement the operations of each step in the data processing system of any of the above embodiments of the present invention.

[0066] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the system embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the system embodiment.

[0067] The system and apparatus of the present invention may be implemented in many ways. For example, the system and apparatus of the present invention may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above sequence of steps for the system is for illustration only, and the steps of the system of the present invention are not limited to the sequence specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the system of the present invention. Thus, the present invention also covers a recording medium storing a program for executing the system of the present invention.

[0068] The description of the present invention is given for the purpose of illustration and description, and is not intended to be exhaustive or to limit the invention to the disclosed form. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for particular uses.

Claims

1. An intelligent governance system based on health and medical big data, characterized in that: include: Health and medical data collection module, used to obtain user electronic health records, user physiological monitoring data collected by wearable devices, and user social environment factor data; A health and medical data processing module, used to extract a user health global correlation feature vector and a user social environment factor semantic correlation feature vector from the user electronic health record, the user physiological monitoring data collected by the wearable device and the user social environment factor data; The user health regulation suggestion generating module is used to generate a personalized regulation suggestion for the user health based on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector.

2. The intelligent management system based on health and medical big data according to claim 1 is characterized in that: The health and medical data processing module includes: A user electronic health record feature extraction unit, configured to extract features from the user electronic health record to obtain a semantic feature vector of the user electronic health record; A user physiological monitoring feature extraction unit, configured to extract features from the user physiological monitoring data collected by the wearable device to obtain a local correlation feature vector of the user physiological monitoring data text; A user health global feature association unit, used for associating the user electronic health record semantic feature vector with the user physiological monitoring data text local association feature vector to obtain the user health global association feature vector; The user social environment factor feature extraction unit is used to perform semantic encoding on the user social environment factor data to obtain the user social environment factor semantic association feature vector.

3. The intelligent management system based on health and medical big data according to claim 2 is characterized in that: The user electronic health record feature extraction unit comprises: The user's electronic health record is passed through a user's electronic health record long link text semantic understanding model to obtain a user's electronic health record semantic sequence; The user electronic health record semantic sequence is passed through a converter-based user electronic health record context long-link encoder to obtain the user electronic health record semantic feature vector.

4. The intelligent governance system based on health and medical big data according to claim 3 is characterized in that: The user physiological monitoring feature extraction unit comprises: A user physiological monitoring embedding coding subunit, used for embedding the user physiological monitoring data collected by the wearable device to obtain a user physiological monitoring data word vector sequence; A user physiological monitoring two-dimensional arrangement subunit, used for two-dimensionally arranging the user physiological monitoring data word vector sequence into a user physiological monitoring data word embedding matrix; The user physiological monitoring convolutional encoding subunit is used to embed the user physiological monitoring data words into a matrix through a user physiological monitoring data text convolutional neural network as a feature extractor to obtain a local correlation feature vector of the user physiological monitoring data text.

5. The intelligent governance system based on health and medical big data according to claim 4 is characterized in that: The user health global feature association unit includes: Use std::vector data structure to define and store user electronic health record semantic feature vectors and user physiological monitoring data text local association feature vectors; Calculating a transposed vector of a local correlation feature vector of the user physiological monitoring data text; Using the matrixMultiply matrix multiplication function to perform a matrix product operation on the electronic health record semantic feature vector and the transposed vector; Generate and output the user health global association feature vector.

6. The intelligent governance system based on health and medical big data according to claim 5 is characterized in that: The user social environment factor feature extraction unit comprises: Passing the user social environment factor data through a user social environment factor semantic encoder to obtain a plurality of user social environment factor feature vectors; The multiple user social environment factor feature vectors are cascaded into the user social environment factor semantic association feature vector.

7. The intelligent governance system based on health and medical big data according to claim 6 is characterized in that: The user health regulation suggestion generating module includes: A user monitoring state feature fusion unit is used to perform back propagation compensation based on feature fine-grained dimension offset on the user health global correlation feature vector and the user social environment factor semantic correlation feature vector to obtain the user monitoring state feature vector; The user health personalized regulation unit is used to pass the user monitoring state feature vector through a generator to generate a user health personalized regulation suggestion.

8. The intelligent governance system based on health and medical big data according to claim 7 is characterized in that: The user monitoring state feature fusion unit includes: Calculating the position difference between the user health global association feature vector and the user social environment factor semantic association feature vector to obtain a dimension shift difference feature vector; Calculate the position point multiplication between the user health global association feature vector and the user social environment factor semantic association feature vector to obtain a dimension shift position point multiplication feature vector; Calculate the positional addition between the user health global association feature vector and the user social environment factor semantic association feature vector to obtain a dimension shift positional addition feature vector; Input the dimension offset differential feature vector, the dimension offset position point product feature vector and the dimension offset position addition feature vector into a one-dimensional convolution layer and perform maximum pooling to obtain a dimension offset convolutional coding pooling feature vector; The square root value of the two-norm of the dimension-shifted convolutional coding pooling feature vector is weighted and added to the one-norm of the dimension-shifted convolutional coding pooling feature vector to obtain a dimension-shift factor; The dimension offset is added by position to obtain a dimension offset intermediate vector by multiplying the feature vector by position and the square root of the length of the user health global associated feature vector; Using the dimension shift factor as the first weighting coefficient and the second weighted hyperparameter as the second weighting coefficient, the position-weighted sum of the dimension shift position-added feature vector and the dimension shift intermediate vector is calculated to obtain the user monitoring state feature vector.

Images