Health data evolution analysis method and device, equipment and medium

Abstract

The invention relates to the technical field of intelligent decision making, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses a health data evolution analysis method, device and equipment and a medium, and the method comprises the steps: obtaining user historical data of a user, and constructing an individual health map of the user according to the user historical data; generating health analysis suggestion data of the user according to the individual health map, and collecting multi-modal health data according to the health analysis suggestion data; extracting multi-modal features in the multi-modal health data, and constructing a health evolution graph according to the multi-modal features; calculating an analysis candidate set of the user according to the health evolution graph; and generating a health data evolution visualization report of the user according to the analysis candidate set. The accuracy of health data evolution analysis can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, and in particular to a method, apparatus, device, and medium for health data evolution analysis. Background Technology

[0002] In the field of healthcare, in recent years, with the rapid development of technologies such as big data and artificial intelligence, some online medical consultation platforms have emerged. Online healthcare can conduct health data analysis, obtain visual reports on the evolution of health data, provide doctors with preliminary analysis suggestions, shorten patients' waiting time, and improve doctors' consultation efficiency.

[0003] Meanwhile, in the fintech field, with the rapid development of technologies such as big data and artificial intelligence in recent years, insurance companies typically utilize intelligent assessment systems when reviewing user applications. These systems involve inquiring about the insured's health status and lifestyle habits to obtain visualized reports on the evolution of health data, thereby quickly identifying high-risk individuals and improving underwriting efficiency and accuracy. Banks and other financial institutions can also leverage intelligent assessment systems to provide health consultations and risk assessments when recommending and processing long-term financial products (such as wealth management products and loan products). This allows them to offer more comprehensive and accurate business advice, for example, recommending flexible wealth management products for high-risk clients with underlying health conditions to prepare for unforeseen financial needs.

[0004] However, existing methods for analyzing the evolution of health data have the following problems: 1. Existing analyses are mostly static, i.e., they are mostly based on pre-set question-answering trees or keyword matching, which cannot flexibly cope with the complexity and evolution of patients' symptoms; 2. In continuous dialogue scenarios, it is difficult to understand the context, which can easily lead to logical breaks or incorrect recommendations; 3. Inability to distinguish individual user differences: The analysis lacks in-depth learning of users' historical health data and patterns of health problem progression, resulting in general and unpersonalized analysis recommendations.

[0005] 4. Simple analysis process: Most of them adopt a linear questioning method, which cannot achieve parallel processing of multiple concurrent symptoms and difficult diagnosis.

[0006] This has resulted in poor accuracy in existing health data evolution analysis. Therefore, improving the accuracy of health data evolution analysis has become an urgent problem to be solved. Summary of the Invention

[0007] This invention provides a method, apparatus, device, and medium for health data evolution analysis to address the problem of poor accuracy in existing health data evolution analysis, thereby improving the accuracy of health data evolution analysis.

[0008] Firstly, a method for analyzing the evolution of health data is provided, including: Obtain the user's historical data and construct the user's individual health profile based on the historical data; Based on the individual health profile, health analysis suggestion data for the user is generated, and multimodal health data is collected based on the health analysis suggestion data; Extract multimodal features from the multimodal health data, and construct the user's health evolution map based on the multimodal features; An analysis candidate set for the user is generated based on the health evolution map, and a health data evolution visualization report for the user is generated based on the analysis candidate set.

[0009] Secondly, a health data evolution analysis device is provided, comprising: The individual health map construction module is used to acquire the user's historical data and construct the user's individual health map based on the historical data. The data acquisition module is used to generate health analysis suggestion data for the user based on the individual health profile, and to collect multimodal health data based on the health analysis suggestion data; A health evolution map construction module is used to extract multimodal features from the multimodal health data and construct the user's health evolution map based on the multimodal features; The report generation module is used to generate an analysis candidate set for the user based on the health evolution map, and to generate a health data evolution visualization report for the user based on the analysis candidate set.

[0010] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned health data evolution analysis method.

[0011] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned health data evolution analysis method.

[0012] In the aforementioned scheme implemented by a health data evolution analysis method, apparatus, device, and medium, by acquiring a user's historical data and constructing an individual health map based on this data, a user's health status and evolution process can be simulated holographically, dynamically, and personally, providing a foundation for subsequently generating health analysis suggestion data for the user. By generating health analysis suggestion data based on the individual health map and collecting multimodal health data based on this data, collaborative reasoning using multimodal information can solve the problem of single information sources, thus enabling more accurate health analysis. Extracting multimodal features from the multimodal health data and constructing a health evolution map based on these features allows for a more comprehensive understanding of the patient's health problem development process, uncovering potential health problem clues, and effectively improving the accuracy of health analysis. Calculating the user's analysis candidate set based on the health evolution map allows for more accurate calculation of the analysis candidate set using contextual information within the health evolution map, generating a visualized report of the user's health data evolution, and performing precise and visualized health data evolution analysis, further improving the accuracy of health data evolution analysis. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of an application environment for a health data evolution analysis method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a health data evolution analysis method according to an embodiment of the present invention; Figure 3 yes Figure 2 A flowchart illustrating a specific implementation method of step S1; Figure 4 yes Figure 2 A flowchart illustrating a specific implementation method of step S3; Figure 5 This is a schematic diagram of a health data evolution analysis device according to one embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 7 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] The health data evolution analysis method provided in this embodiment of the invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can obtain the user's historical data, construct the user's individual health profile based on the historical data, generate health analysis suggestion data for the user based on the individual health profile, collect multimodal health data based on the health analysis suggestion data, extract multimodal features from the multimodal health data, construct the user's health evolution profile based on the multimodal features, generate the user's analysis candidate set based on the health evolution profile, and generate a health data evolution visualization report for the user based on the analysis candidate set. This invention provides a health data evolution analysis device. For target result businesses, by collecting multimodal health data, it can perform collaborative reasoning through multimodal information, solving the problem of single information sources, thereby conducting more accurate health analysis; by constructing a health evolution profile, it can more comprehensively understand the development process of the patient's health problems, discover potential health problem clues, and effectively improve the accuracy of health analysis; by calculating the user's analysis candidate set based on the health evolution profile, it can utilize the contextual information in the health evolution profile to calculate the analysis candidate set more accurately, further improving the accuracy of health analysis. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0017] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a health data evolution analysis method provided in an embodiment of the present invention includes the following steps: S1. Obtain the user's historical data and construct the user's individual health profile based on the historical data.

[0018] In this embodiment of the invention, user historical data is multi-dimensional data that reflects the user's physical condition. For example, it may include electronic medical record data, physical examination data, data detected by wearable devices, user lifestyle and environmental data, and user's subjective complaints during health analysis.

[0019] Specifically, an individual health graph is a graph structure built based on a user's historical data. It uses elements from that data as nodes, such as health problems, symptoms, medications, and genes. Edges are established based on the relationships between the data, such as a patient having a certain health problem, a health problem having specific symptoms, or a medication used to treat a certain health problem.

[0020] In the embodiments of the present invention, see Figure 3 As shown, constructing the user's individual health profile based on the user's historical data includes: S31. Perform data format unification processing on the user historical data to obtain target user data; S32. Extract the user's health nodes and the connection relationships between each health node based on the target user data; S33. Construct a graph structure based on the health nodes and the connection relationships to obtain the user's individual health graph.

[0021] In this embodiment of the invention, data format unification processing involves converting data from different sources in user historical data into a unified format and encoding standard, such as a unified encoding system for health problem analysis and units for examination and testing indicators.

[0022] Specifically, health nodes can be extracted using preset health node types, such as user, health problem, symptoms, examination and test indicators, medication, and medical institution. For example, a user node includes basic attributes such as name, gender, and age; a health problem node includes attributes such as health problem name, category, and code.

[0023] Furthermore, the connection relationship is the type of relationship between health nodes. For example, the "analysis" relationship connects individual nodes and health problem nodes, the "accompaniment" relationship connects symptom nodes and health problem nodes, and the "medication" relationship connects individual nodes and drug nodes. Based on the connection relationship, corresponding edges can be established between health nodes to connect them and form a graph structure, thus obtaining the user's individual health graph.

[0024] In detail, potential relationships between nodes can be mined based on semantic associations and professional knowledge in the medical field. For example, there is a companion relationship between "analyzing symptoms" and "health problems," where certain symptoms often accompany specific health problems, so a "companion" edge can be defined to connect them. Similarly, there may be a relationship between "anatomical location" and "health problems," where a certain health problem usually occurs in a specific anatomical location, in which case a "lesion location" edge can be established.

[0025] In this embodiment of the invention, by constructing a user's individual health profile, a user's health status and evolution process can be simulated holographically, dynamically, and in a personalized manner, including physical examination reports, medical records, medication history, lifestyle habits, etc., providing a foundation for generating health analysis and suggestion data for the user in the future.

[0026] S2. Generate health analysis suggestion data for the user based on the individual health profile, and collect multimodal health data based on the health analysis suggestion data.

[0027] In this embodiment of the invention, the health analysis suggestion data can be dynamically adjusted according to the user's individual health profile, enabling the health analysis system to have the ability to make "personalized" judgments. The health analysis system includes a graph neural network (GNN), which uses the graph neural network to generate the user's health analysis suggestion data based on the individual's health profile.

[0028] The core idea of ​​graph neural networks is to update the representation of the current node by aggregating the feature information of neighboring nodes in an individual's health graph, thereby capturing the dependencies and topological features between nodes in the graph. The key to graph neural networks lies in the message passing mechanism, which uses information flow between nodes to pass the features of neighboring nodes to the target node to update the target node's representation.

[0029] Specifically, the basic process of a graph neural network includes three steps: message generation, message aggregation, and node update. First, the neighboring nodes of the target node generate messages based on their features; then, the messages from all neighboring nodes are aggregated into a single aggregated message; finally, the aggregated message is combined with the original features of the target node to generate a new node representation.

[0030] In this embodiment of the invention, generating health analysis suggestion data for the user based on the individual health profile includes: Calculate the initial features of each node in the individual health map; Based on the initial features, generate neighbor node messages for each node; The neighbor node messages are aggregated to obtain the target node message; The initial features are updated based on the target node message to obtain the target node features; Health problem prediction is performed on the features of the target node to obtain the initial candidate problem probabilities; Based on the initial candidate question probabilities, health analysis suggestion data for the user is generated.

[0031] Specifically, pre-trained Graph Convolutional Networks (GCNs) can be used to generate health analysis suggestion data. A GCN is a neural network model based on graph-structured data, consisting of an input layer, hidden layers, and an output layer. The input layer receives the initial features of each node and the feature matrices of its neighboring nodes. The hidden layer consists of multiple graph convolutional layers, each of which introduces non-linearity through an activation function, such as ReLU activation, to increase the model's expressive power. The output layer can be designed according to the specific task. The output layer is usually a graph convolutional layer followed by a softmax activation function to output the probability distribution of each category, thereby obtaining health problem analysis suggestions corresponding to the individual's health map.

[0032] In this embodiment of the invention, the Word2Vec algorithm can be used to learn the embedding representation of each node, obtaining the initial features of each node. Then, messages are generated based on the features of the target node's neighboring nodes. For example, in GCN, the features of neighboring nodes can be multiplied by the preset weight matrix of each layer to generate neighboring node messages, where a neighboring node is a node that has a direct edge connection with the target node.

[0033] Furthermore, message aggregation is performed in each graph convolutional layer, where all neighbor messages of a node are aggregated into a single vector. This can be achieved through summation aggregation, mean aggregation, pooling aggregation, attention aggregation, etc. Then, a new node representation is generated by combining the initial feature representation of the current node with the aggregated target node message. The initial features are then updated to obtain the target node features.

[0034] In detail, the initial features of the nodes can be updated according to a preset update function, which usually includes nonlinear transformations, to obtain the features of the target node. Alternatively, a gating mechanism (such as GraphSAGE's gated update) can be used to control the fusion ratio of the initial features and the target node messages through a gating vector.

[0035] In this system, the hidden layers of the graph convolutional network consist of multiple graph convolutional layers. Through multi-layer graph convolutional computation, semantic features from local to global can be learned hierarchically, capturing multi-scale feature information, enhancing the richness of the feature representation of target nodes, and thus improving the accuracy of health analysis suggestion data generation.

[0036] In this embodiment of the invention, classification prediction involves aggregating the features of target nodes to obtain a graph-level feature representation. Based on the graph-level feature representation, the predicted probabilities of multiple initial candidate questions are calculated. For example, the probability of influenza is 33%, and the probability of rhinitis is 30%. Health analysis suggestion data is generated by calling a preset consultation template based on the initial candidate question probabilities.

[0037] The probability distribution can be obtained based on the probability of each initial candidate question. Then, the information entropy of the probability distribution can be calculated to obtain the uncertainty result of the initial candidate questions. For example, if the probability of a health problem in the initial candidate questions is much higher (e.g., influenza 90%, others much lower), the information entropy is small, indicating "high certainty". If the distributions of multiple candidate health problems are similar (e.g., influenza 33%, rhinitis 30%, allergies 28%), the entropy is high, indicating "high uncertainty". By normalizing the information entropy, the uncertainty score of the probability distribution of the initial candidate questions is obtained. When the uncertainty score is high, the corresponding consultation template is called to obtain health analysis suggestion data.

[0038] For example, if the uncertainty score is ranked as 1. Alcoholic cerebellar lesion; 2. Meniere's syndrome; 3. Hypertensive encephalopathy, then health analysis suggestions will be obtained for questions such as "Does it have tinnitus?", "Has your sleep quality recently declined?", and "Have you ever had a history of syncope?".

[0039] In this embodiment of the invention, health analysis and suggestion data is fed back to the user, and data fed back by the user based on the health analysis and suggestion data is collected to obtain multimodal health data. The multimodal health data includes text data input by the user, image data related to the consultation (such as photos of skin lesions), and voice data spoken by the user.

[0040] In this embodiment of the invention, multimodal health data can be used for collaborative reasoning through multimodal information, solving the problem of a single information source and thus enabling more accurate health analysis.

[0041] S3. Extract multimodal features from the multimodal health data, and construct the user's health evolution map based on the multimodal features.

[0042] In this embodiment of the invention, multimodal features are feature sets of different modalities in multimodal health data, such as features of text data, features of image data, and features of voice data.

[0043] Specifically, see Figure 4 As shown, the extraction of multimodal features from the multimodal health data includes: S41. Extract text semantics from the text data in the multimodal health data to obtain text features; S42. Perform multi-layer convolution processing on the images in the multimodal health data to obtain image features; S43. Perform text recognition on the speech data in the multimodal health data to obtain speech text, and perform text semantic extraction on the speech text to obtain speech semantic features; S44. Aggregate the text features, the image features, and the speech semantic features to obtain the multimodal features in the multimodal health data.

[0044] Specifically, the text features are the text semantic features in the text data, which convert unstructured or semi-structured medical texts into computable semantic information to assist in health problem analysis and treatment decision-making. Specifically, through text semantic extraction, medical entities and the associations between entities in the text data can be identified. For example, from "The user takes amlodipine due to hypertension", the relationship "hypertension - treatment - amlodipine" can be extracted, and professional terms in the text data can also be understood.

[0045] Specifically, pre-built pre-trained models in the medical field (such as Chinese-MedBERT, BioBERT) can be used for text semantic extraction to understand the interrogation semantics in the text data. Among them, the Chinese-MedBERT model is a pre-trained language model specifically designed for the Chinese medical field. Usually, based on the BERT (Bidirectional Encoder Representations from Transformers) model architecture, it is optimized for the medical field. For example, full entity masking is adopted, and words identified by the knowledge graph and named entity recognition are used as masking units. For example, both characters of "abdominal pain" are masked simultaneously, and domain knowledge entities such as symptoms, health problems, examinations, treatments, and drugs are introduced.

[0046] The core architecture of BioBERT is basically the same as BERT. First, the text data is segmented into word segments, and each word segment is indexed and encoded. The segmented word segments are converted into corresponding word embeddings. At the same time, position embedding is added to capture the order information of the text, and sentence embedding is added to distinguish different sentences. Special markers are used, such as the [CLS] marker placed at the beginning of the sequence to represent the classification information of the entire sequence; the [SEP] marker is used to separate different sentences. The multi-layer Transformer encoder captures the contextual relationships in different subspaces, which can simultaneously focus on different positions of the input sequence to obtain global dependencies. The encoder also includes a feedforward fully connected layer to perform non-linear transformation on the features at each position in the sequence to further extract and transform features. Residual connections and layer normalization are used after each encoder sub-layer to help alleviate the training difficulties of deep networks, accelerate network convergence, and improve the performance and stability of the model. By extracting the embedding features of all encoding layers, and then performing weighted or averaged processing to obtain richer semantic information, the text features are obtained.

[0047] Furthermore, Automatic Speech Recognition (ASR) technology can be used to convert speech data into text, resulting in speech-text.

[0048] Specifically, speech data can be segmented and windowed to reduce discontinuities between frames. Feature parameters, such as Mel-frequency cepstral coefficients, can then be extracted from the speech data. Finally, a pre-defined speech recognition model can be used for text recognition to obtain the speech text. However, using the above method for extracting text semantics from text data to extract text semantics from speech text does not yield speech semantic features.

[0049] In this embodiment of the invention, the images in the multimodal health data are real-time images provided by users during consultations, such as images of users measuring blood pressure and body temperature, photos of patients showing skin problems such as rashes, erythema, and acne, images of body pain, and images of examination reports. Then, a convolutional neural network pre-trained in the medical field is used to perform multi-layer convolution processing on the images to obtain the image features of each image.

[0050] In detail, by aggregating text features, image features, and speech semantic features from the multimodal health data, more comprehensive analytical information can be provided, making up for the shortcomings of a single modality and improving the reliability of subsequent candidate set calculations.

[0051] Specifically, the health evolution map records and displays the changes in a user's health status over time based on multimodal health data from user feedback. It provides an intuitive understanding of health dynamics, focuses on the changing trends of a user's health status over time, records long-term changes in physiological indicators, the gradual impact of changes in lifestyle habits on health, and fluctuations in psychological state at different stages, forming a continuous and dynamic health evolution record.

[0052] In this embodiment of the invention, constructing the user's health evolution map based on the multimodal features includes: The multimodal features are aligned to obtain aligned features, and the multimodal features are fused based on the aligned features to obtain the target features. Based on the target features, health problems are predicted to obtain candidate problem probabilities; The health analysis suggestion data is iteratively optimized based on the candidate problem probabilities to obtain iterative analysis suggestion data. Based on the iterative analysis recommendations, the user's iterative multimodal data is collected. Generate the user's prediction question sequence based on the iterative multimodal data; Construct the user's health evolution map based on the predicted question sequence.

[0053] In this embodiment of the invention, feature alignment processing maps features from different modalities (such as images, text, audio, etc.) to a common feature space for comparison and feature fusion. Specifically, linear transformations (such as matrix multiplication) can be used to project features from different modalities into a shared subspace. Common linear transformation methods include principal component analysis (PCA) and linear discriminant analysis (LDA); or a unified architecture (such as Multimodal Transformer) can be used to embed each modality into a unified semantic representation to obtain aligned features.

[0054] Furthermore, aligned features can be concatenated to obtain target features, forming a richer and more discriminative feature space. For example, images can provide intuitive visual semantic support for text, while text can further explain user actions and relationships in images, making semantic understanding richer and more accurate, and effectively improving the accuracy of health evolution map construction.

[0055] In detail, a pre-built medical large language model, such as a localized model based on the existing large model Med-PaLM or Chinese Medical BERT, can be used to predict health problems of the target features, obtaining the predicted problem probabilities. Then, following the steps described above for generating health analysis suggestion data for the user, suggestion data corresponding to the candidate problem probabilities is generated, thereby updating the health analysis suggestion data to obtain iterative analysis suggestion data. Each update receives multimodal health data feedback from the user, resulting in iterative multimodal data. Specifically, iterative optimization of the health analysis suggestion data can be stopped when the user does not provide feedback or when the number of iterations meets a preset threshold, thus collecting iterative multimodal data.

[0056] Furthermore, health problem prediction is performed on the multimodal data collected in each iteration of the multimodal data, obtaining the probability of candidate problems for each step. Based on the predicted probability analysis, each symptom of the user is sorted by time to obtain a predicted problem sequence, which can reflect the evolution of health problems. Then, various health problems at the same time in the predicted problem sequence, such as cough, fever, and headache, are used as nodes. Based on the association and evolution relationship of each node in the predicted problem sequence—for example, a cough may develop into phlegm, and a fever may be accompanied by a headache—the predicted problem sequence is correlated to obtain the user's health evolution map.

[0057] Furthermore, conditional random field models can be used to capture the transition probabilities between sequences of predicted questions, or attention-based models can be used to highlight important symptoms and key evolution paths, thereby associating predicted questions in the sequence of predicted questions.

[0058] Preferably, each node in the health evolution map can also include information such as the name, occurrence time, duration, and severity of each symptom in the predicted problem probability, thereby gaining a more comprehensive understanding of the patient's health problem development process, discovering potential health problem clues, and effectively improving the accuracy of health analysis.

[0059] S4. Generate an analysis candidate set for the user based on the health evolution map, and generate a health data evolution visualization report for the user based on the analysis candidate set.

[0060] In this embodiment of the invention, the analysis candidate set refers to a set of analysis results of possible health problems of the user inferred based on the health evolution map, so that doctors can conduct further evaluation and decision-making.

[0061] In this embodiment of the invention, generating the analysis candidate set for the user based on the health evolution map includes: Calculate the node embedding features of each node in the health evolution map; Calculate the normalized attention coefficient for each node based on the node embedding features; The node embedding features are updated based on the normalized attention coefficients to obtain the node target features; A fully connected prediction is performed on the target features of the nodes to obtain the analysis candidate set of the user.

[0062] In detail, a feature vector is initialized for each node in the health evolution graph to obtain node embedding features. These can be vectors obtained by encoding the node's attribute information (such as the name of the health problem, the description of symptoms, etc.) or randomly initialized embedding vectors. Then, an attention mechanism is used to calculate the attention coefficient. Specifically, the feature vectors of each node and its neighboring nodes are subjected to a linear transformation, and then the attention coefficient of each node to each neighboring node is calculated using an attention function (such as dot product, concatenation followed by a fully connected layer, etc.). The calculated attention coefficients are normalized using the softmax function to obtain normalized coefficients, such that the sum of the attention coefficients of all neighboring nodes is 1. The normalized attention coefficient deficit represents the importance of neighboring nodes to the central node.

[0063] Furthermore, based on the normalized attention coefficients, the feature vectors of neighboring nodes are weighted and summed to obtain an aggregated representation of the neighboring node features. The original feature vector of the node is then concatenated or summed with the aggregated feature vectors of the neighboring nodes to obtain the new feature vector of the node, which is the target feature of the node, thus completing the update of the node features.

[0064] Furthermore, the node target features are input into a pre-constructed fully connected layer and mapped to a preset feature space. This allows for the prediction of the target node features, obtaining the probability of each candidate analysis. The candidate analysis with the higher probability is selected to obtain the analysis candidate set.

[0065] In this embodiment of the invention, multiple GAT (Graph Attention Network) layers can be stacked, and the attention calculation and feature update process described above can be repeated in each layer. The output of each layer serves as the input of the next layer, which can capture node dependencies at greater distances and richer semantic information. At the same time, the contextual information in the healthy evolution graph (such as the type, attributes, and associations with other nodes of the node) can be used to better focus on useful information and more accurately calculate and analyze the candidate set.

[0066] In this embodiment of the invention, the health data evolution visualization report is a structured summary of the user's consultation process and generates a visual report that supports doctors' review and feedback.

[0067] In this embodiment of the invention, generating a visualization report on the user's health data evolution based on the analysis candidate set includes: An analysis suggestion set corresponding to the analysis candidate set is generated using a pre-set medical suggestion database; The user's health analysis path is constructed based on the analysis suggestion set and the health evolution map; The health analysis path is formatted to generate a health analysis summary; The health analysis summary is added to a pre-built report template to obtain a visualization report of the user's health data evolution.

[0068] In this embodiment of the invention, the medical advice database contains a large number of authoritative treatment guidelines and standard procedures, which can provide scientifically sound treatment plans for analyzing candidate sets, thereby improving the accuracy and safety of diagnosis and treatment. Furthermore, it provides abundant case data for doctors to compare similar cases, learn from treatment experience, and find the best treatment plan.

[0069] Furthermore, the health analysis path guides users step-by-step through consultations based on symptoms from each user's multimodal health data feedback and symptoms predicted from that data. For example, it includes the symptom evolution process in the health evolution map, as well as information such as the name, occurrence time, duration, and severity of each symptom. It can also include health analysis suggestions for each consultation, thus creating a visualized path of "symptoms → predicted health problems → health problem suggestions → suggested actions," which supports doctors in reviewing and providing feedback.

[0070] Specifically, the formatting process involves segmenting and dividing the health analysis path into sections or columns according to the order of user information - chief complaint - symptom characteristics - analysis candidate set (sorted by probability) - analysis suggestion set, thus obtaining a summary of the health analysis. This makes the analysis easier to read and allows users or subsequent medical personnel to quickly understand the core content of the health analysis, including key information such as the user's chief complaint, main symptoms, candidate analysis results, and analysis suggestions.

[0071] Specifically, the report template is a predefined format and structure for creating standardized reports. It can include user information, main symptoms, chief complaint medical history, examination results, analysis candidate set, and corresponding analysis suggestion set. The content in the health analysis summary is added to the report template to visualize the user's consultation information, which helps doctors to conduct subsequent analysis.

[0072] In this embodiment of the invention, in the financial field, health data evolution visualization reports can be used for credit assessment and risk prediction, provide financial product information and investment advice, and help financial institutions prevent financial fraud and ensure compliant operation. For example, financial institutions can provide personalized investment advice and financial planning based on health data evolution visualization reports, generate personalized investment advice reports, recommend suitable financial products and services to customers, and improve user experience and satisfaction.

[0073] As can be seen, in the above solution, for the target result business, the user's historical data is acquired, and an individual health profile of the user is constructed based on the historical data; health analysis suggestion data for the user is generated based on the individual health profile, and multimodal health data is collected based on the health analysis suggestion data; multimodal features are extracted from the multimodal health data, and a health evolution profile of the user is constructed based on the multimodal features; an analysis candidate set for the user is generated based on the health evolution profile; and a health data evolution visualization report for the user is generated based on the analysis candidate set. By collecting multimodal health data, collaborative reasoning can be performed through multimodal information, solving the problem of single information sources, thereby enabling more accurate health analysis; further constructing a health evolution profile allows for a more comprehensive understanding of the patient's health problem development process, discovering potential health problem clues, and effectively improving the accuracy of health analysis; calculating the user's analysis candidate set based on the health evolution profile utilizes the contextual information in the health evolution profile to more accurately calculate the analysis candidate set, further improving the accuracy of health data evolution analysis.

[0074] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0075] In one embodiment, a health data evolution analysis device is provided, which corresponds one-to-one with the health data evolution analysis method described in the above embodiments. For example... Figure 5 As shown, this health data evolution analysis device includes an individual health atlas construction module 101, a data acquisition module 102, a health evolution atlas construction module 103, and a report generation module 104. Detailed descriptions of each functional module are as follows: The individual health map construction module 101 is used to acquire the user's historical data and construct the user's individual health map based on the user's historical data. Data acquisition module 102 is used to generate health analysis suggestion data for the user based on the individual health profile, and to collect multimodal health data based on the health analysis suggestion data; The health evolution map construction module 103 is used to extract multimodal features from the multimodal health data and construct the user's health evolution map based on the multimodal features; The report generation module 104 is used to generate an analysis candidate set for the user based on the health evolution map, and to generate a health data evolution visualization report for the user based on the analysis candidate set.

[0076] In one embodiment, the individual health map construction module 101, in constructing the user's individual health map based on the user's historical data, is used for: The user historical data is processed to unify the data format to obtain the target user data; Extract the user's health nodes and the connection relationships between each health node based on the target user data; Based on the health nodes and the connection relationships, a graph structure is constructed to obtain the user's individual health graph.

[0077] In one embodiment, the data acquisition module 102 generates health analysis suggestion data for the user based on the individual health profile, for the following purposes: Calculate the initial features of each node in the individual health map; Based on the initial features, generate neighbor node messages for each node; The neighbor node messages are aggregated to obtain the target node message; The initial features are updated based on the target node message to obtain the target node features; Health problem prediction is performed on the features of the target node to obtain the initial candidate problem probabilities; Based on the initial candidate question probabilities, health analysis suggestion data for the user is generated.

[0078] In one embodiment, the health evolution map construction module 103, after extracting multimodal features from the multimodal health data, is used for: Text semantic extraction is performed on the text data in the multimodal health data to obtain text features; The images in the multimodal health data are subjected to multi-layer convolution processing to obtain image features; Text recognition is performed on the speech data in the multimodal health data to obtain speech text, and text semantic extraction is performed on the speech text to obtain speech semantic features; The text features, image features, and speech semantic features are combined to obtain the multimodal features in the multimodal health data.

[0079] In one embodiment, the health evolution map construction module 103, in constructing the user's health evolution map based on the multimodal features, is further configured to: The multimodal features are aligned to obtain aligned features, and the multimodal features are fused based on the aligned features to obtain the target features. Based on the target features, health problems are predicted to obtain candidate problem probabilities; The health analysis suggestion data is iteratively optimized based on the candidate problem probabilities to obtain iterative analysis suggestion data. Based on the iterative analysis recommendations, the user's iterative multimodal data is collected. Generate the user's prediction question sequence based on the iterative multimodal data; Construct the user's health evolution map based on the predicted question sequence.

[0080] In one embodiment, the report generation module 104 generates an analysis candidate set for the user based on the health evolution map, for the following purposes: Calculate the node embedding features of each node in the health evolution map; Calculate the normalized attention coefficient for each node based on the node embedding features; The node embedding features are updated based on the normalized attention coefficients to obtain the node target features; A fully connected prediction is performed on the target features of the nodes to obtain the analysis candidate set of the user.

[0081] In one embodiment, the report generation module 104 generates a visualization report on the evolution of the user's health data based on the analysis candidate set, for the following purposes: An analysis suggestion set corresponding to the analysis candidate set is generated using a pre-set medical suggestion database; The user's health analysis path is constructed based on the analysis suggestion set and the health evolution map; The health analysis path is formatted to generate a health analysis summary; The health analysis summary is added to a pre-built report template to obtain a visualization report of the user's health data evolution.

[0082] This invention provides a health data evolution analysis device. For a target outcome business, it acquires a user's historical data and constructs an individual health map of the user based on the historical data. It then generates health analysis suggestion data for the user based on the individual health map, collects multimodal health data based on the health analysis suggestion data, extracts multimodal features from the multimodal health data, constructs a health evolution map of the user based on the multimodal features, generates an analysis candidate set for the user based on the health evolution map, and generates a health data evolution visualization report for the user based on the analysis candidate set. By collecting multimodal health data, collaborative reasoning can be performed using multimodal information, solving the problem of single information sources and thus enabling more accurate health analysis. Furthermore, constructing a health evolution map allows for a more comprehensive understanding of the patient's health problem development process, discovering potential health problem clues, and effectively improving the accuracy of health analysis. Calculating the user's analysis candidate set based on the health evolution map utilizes the contextual information within the health evolution map to more accurately calculate the analysis candidate set, further improving the accuracy of health data evolution analysis.

[0083] For specific limitations regarding a health data evolution analysis device, please refer to the limitations of a health data evolution analysis method described above, which will not be repeated here. Each module in the aforementioned health data evolution analysis device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0084] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a health data evolution analysis method on the server side.

[0085] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a health data evolution analysis method on the client side. In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain the user's historical data and construct the user's individual health profile based on the historical data; Based on the individual health profile, health analysis suggestion data for the user is generated, and multimodal health data is collected based on the health analysis suggestion data; Extract multimodal features from the multimodal health data, and construct the user's health evolution map based on the multimodal features; An analysis candidate set for the user is generated based on the health evolution map, and a health data evolution visualization report for the user is generated based on the analysis candidate set.

[0086] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain the user's historical data and construct the user's individual health profile based on the historical data; Based on the individual health profile, health analysis suggestion data for the user is generated, and multimodal health data is collected based on the health analysis suggestion data; Extract multimodal features from the multimodal health data, and construct the user's health evolution map based on the multimodal features; An analysis candidate set for the user is generated based on the health evolution map, and a health data evolution visualization report for the user is generated based on the analysis candidate set.

[0087] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0089] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0090] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

[0091] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for analyzing the evolution of health data, characterized in that, include: Obtain the user's historical data and construct the user's individual health profile based on the historical data; Based on the individual health profile, health analysis suggestion data for the user is generated, and multimodal health data is collected based on the health analysis suggestion data; Extract multimodal features from the multimodal health data, and construct the user's health evolution map based on the multimodal features; An analysis candidate set for the user is generated based on the health evolution map, and a health data evolution visualization report for the user is generated based on the analysis candidate set.

2. The health data evolution analysis method as described in claim 1, characterized in that, The step of constructing the user's individual health profile based on the user's historical data includes: The user historical data is processed to unify the data format to obtain the target user data; Extract the user's health nodes and the connection relationships between each health node based on the target user data; Based on the health nodes and the connection relationships, a graph structure is constructed to obtain the user's individual health graph.

3. The health data evolution analysis method as described in claim 1, characterized in that, The step of generating health analysis suggestion data for the user based on the individual health profile includes: Calculate the initial features of each node in the individual health map; Based on the initial features, generate neighbor node messages for each node; The neighbor node messages are aggregated to obtain the target node message; The initial features are updated based on the target node message to obtain the target node features; Health problem prediction is performed on the features of the target node to obtain the initial candidate problem probabilities; Based on the initial candidate question probabilities, health analysis suggestion data for the user is generated.

4. The health data evolution analysis method as described in claim 1, characterized in that, The step of constructing the user's health evolution map based on the multimodal features includes: The multimodal features are aligned to obtain aligned features, and the multimodal features are fused based on the aligned features to obtain the target features. Based on the target features, health problems are predicted to obtain candidate problem probabilities; The health analysis suggestion data is iteratively optimized based on the candidate problem probabilities to obtain iterative analysis suggestion data. Based on the iterative analysis recommendations, the user's iterative multimodal data is collected. Generate the user's prediction question sequence based on the iterative multimodal data; Construct the user's health evolution map based on the predicted question sequence.

5. The health data evolution analysis method as described in claim 1, characterized in that, The step of generating the analysis candidate set for the user based on the health evolution map includes: Calculate the node embedding features of each node in the health evolution map; Calculate the normalized attention coefficient for each node based on the node embedding features; The node embedding features are updated based on the normalized attention coefficients to obtain the node target features; A fully connected prediction is performed on the target features of the nodes to obtain the analysis candidate set of the user.

6. The health data evolution analysis method as described in claim 1, characterized in that, The step of generating a visualization report on the evolution of the user's health data based on the analysis candidate set includes: An analysis suggestion set corresponding to the analysis candidate set is generated using a pre-set medical suggestion database; The user's health analysis path is constructed based on the analysis suggestion set and the health evolution map; The health analysis path is formatted to generate a health analysis summary; The health analysis summary is added to a pre-built report template to obtain a visualization report of the user's health data evolution.

7. The health data evolution analysis method as described in claim 1, characterized in that, The extraction of multimodal features from the multimodal health data includes: Text semantic extraction is performed on the text data in the multimodal health data to obtain text features; The images in the multimodal health data are subjected to multi-layer convolution processing to obtain image features; Text recognition is performed on the speech data in the multimodal health data to obtain speech text, and text semantic extraction is performed on the speech text to obtain speech semantic features; The text features, image features, and speech semantic features are combined to obtain the multimodal features in the multimodal health data.

8. A health data evolution analysis device, characterized in that, include: The individual health map construction module is used to acquire the user's historical data and construct the user's individual health map based on the historical data. The data acquisition module is used to generate health analysis suggestion data for the user based on the individual health profile, and to collect multimodal health data based on the health analysis suggestion data; A health evolution map construction module is used to extract multimodal features from the multimodal health data and construct the user's health evolution map based on the multimodal features; The report generation module is used to generate an analysis candidate set for the user based on the health evolution map, and to generate a health data evolution visualization report for the user based on the analysis candidate set.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the health data evolution analysis method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the health data evolution analysis method as described in any one of claims 1 to 7.

Images