Personalized health management system and method based on artificial intelligence

Abstract

The present invention relates to a personalized health management system and method based on artificial intelligence. The method realizes the automatic collection and transcription of health data; uses a large model to automatically summarize, label and generate personalized suggestions for health data; builds a complete health timeline to help doctors quickly and comprehensively understand the health status of patients; and realizes the collaborative sharing of information between patients and doctors through an online consultation platform. The advantages are that it reduces the data management burden of patients, improves the diagnosis and treatment efficiency of doctors, and improves the overall health management effect through intelligent functions.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology applications, and in particular to a personalized health management system and method based on artificial intelligence. Background Art

[0002] With the acceleration of population aging and the transformation of modern lifestyle, the number of patients with chronic diseases continues to grow. Chronic diseases such as diabetes, hypertension and heart disease have the characteristics of long course and many complications, which require patients to carry out continuous self-management and regular medical supervision.

[0003] However, since the health data of patients are scattered in various medical institutions, physical examination centers, community hospitals, and personal testing equipment such as home blood pressure monitors and blood glucose meters, the scattered health data of users has formed a serious data island phenomenon, which has led to the fragmentation of the health timeline, making it difficult for patients to systematically track and analyze their health trends and establish a complete health record. At the same time, with the popularization of wearable devices and Internet of Things technology, the ways of collecting health data are becoming increasingly diverse, but these continuously collected real-time data are often difficult to effectively integrate with traditional medical records, resulting in unsatisfactory health management results.

[0004] From the perspective of medical service providers, doctors face tremendous work pressure and information processing burden in clinical practice. In the process of facing the complicated medical records and test data of each patient, it is difficult for doctors to quickly build a complete health timeline for patients within the limited diagnosis and treatment time, and accurately grasp the law of disease development. The asymmetry of information and the fragmentation of data will seriously affect the doctor's comprehensive understanding and accurate judgment of the patient's condition. Especially for patients with chronic diseases, health status often involves long-term changes in multiple physiological indicators, including clinical test results, imaging data, medication records, lifestyle data, etc. If multi-dimensional information cannot be systematically presented on a unified timeline, doctors will not be able to intuitively understand the characteristics of the patient's disease course evolution and treatment response. At the same time, due to the lack of intelligent auxiliary analysis tools, it is difficult for doctors to quickly extract key clinical clues from massive historical data and discover potential health risk factors, which not only reduces the efficiency of diagnosis and treatment, but also may cause some important medical decision-making basis to be ignored.

[0005] In summary, the current health management methods mainly have the following problems:

[0006] 1. Data interconnection and interoperability issues:

[0007] Because health data is fragmented and divided, and there is a lack of a unified data integration mechanism, it is impossible to build a complete health timeline for patients;

[0008] 2. Insufficient timing analysis:

[0009] Due to the lack of time-series correlation analysis of patients' health data, it is difficult to find the mutual influence relationship between different health parameters, which affects the scientific nature of health management;

[0010] 3. Clinical decision-making efficiency issues:

[0011] Doctors face a large amount of medical records and test data, but lack efficient data analysis and visualization tools, making it difficult to quickly grasp the dynamic changes in patients' health, which affects the accuracy and timeliness of diagnosis and treatment decisions.

[0012] 4. Preventing delayed intervention:

[0013] Since the existing health management model is mainly based on passive treatment and lacks a timeline-based prediction and early warning mechanism, it is impossible to achieve early identification and proactive intervention of health risks;

[0014] 5. Lack of personalized management:

[0015] Health management plans are relatively unified and fail to fully consider individual differences among patients, making it difficult to provide personalized management strategies and intervention recommendations based on patients' historical health trajectories.

[0016] In combination with the above, the related art is implemented through a manual health data management system, as follows:

[0017] Step 1: The patient receives treatment at different medical institutions, generating multiple health reports and records;

[0018] Step 2: Patients manually collect and organize these health data and input them into personal health management applications;

[0019] Step 3: The application provides basic data display and simple statistical functions to help patients view various health indicators;

[0020] Step 4: Based on the data display and statistical results obtained in step 3, the patient determines his / her health status and takes corresponding management measures;

[0021] Step 5: During the diagnosis and treatment process, the doctor tries to understand the patient's health status through the scattered information provided by the patient.

[0022] As can be seen from the above, in the relevant technologies, patients need to manually collect and input health data, which increases the management burden and makes the operation complicated and time-consuming; in the process of manual input and collation, data entry errors are prone to occur, affecting the accuracy of health management and causing data to be prone to errors; the existing health user system only provides basic data display and cannot automatically generate health analysis reports and key opinions, resulting in a lack of automated analysis and summary; and health data lacks effective classification, making it difficult for users and doctors to quickly find and manage specific types of information; it is difficult for doctors to obtain complete and continuous patient health information from scattered data, affecting diagnosis and treatment decisions and making it difficult for doctors to understand the patient's condition; and the existing health user system cannot provide personalized management suggestions and adjustment plans based on the patient's specific health status.

[0023] Currently, no effective solution has been proposed to address the problems existing in existing technologies, such as scattered health data, complex manual management, difficulty for doctors to fully understand the patient's condition, lack of automated summary and inconvenient label management. Summary of the invention

[0024] The purpose of the present invention is to provide a personalized health management system and method based on artificial intelligence to address the deficiencies in the prior art, so as to solve the technical problems existing in the prior art, such as scattered health data, complex manual management, difficulty for doctors to fully understand the patient's condition, lack of automated summary and inconvenient label management.

[0025] To achieve the above object, the technical solution adopted by the present invention is:

[0026] The present invention provides a personalized health management system based on artificial intelligence, including: a data acquisition module, a data sorting module, a knowledge graph construction module, an automatic summary module, a tag management module, a health timeline module, a personalized suggestion module, an interaction module and a reminder module, wherein the data acquisition module is used to obtain the user's health data through image acquisition records, audio acquisition records, local records of the smart terminal, records of the online platform, and / or acquisition records of wearable devices in a smart terminal according to pre-granted permissions; the data sorting module is connected to the data acquisition module and is used to convert the user's health data into target format data and establish a health database based on the target format data; the knowledge graph construction module is connected to the data sorting module and is used to convert the user's health data into target format data and establish a health database based on the target format data. The module is connected to the knowledge graph construction module, and is used to associate the pre-stored medical knowledge with the health database to generate the medical knowledge graph for the corresponding user; the health timeline module is connected to the knowledge graph construction module, and is used to perform semantic annotation of events through the time series in the user's health data based on the medical knowledge graph and the health database to obtain the user's health timeline; the automated summary module is connected to the knowledge graph construction module and the health timeline module, and is used to extract the user's key health indicator data based on the medical knowledge graph, analyze the key health indicator data based on the medical knowledge graph and the health timeline, and obtain preliminary analysis results; and generate a health summary report based on the medical knowledge graph, the health timeline and the preliminary analysis results through the retrieval enhancement generation technology based on the knowledge graph. ; A tag management module is connected to the automated summary module, and is used to generate at least one tag through a large language model based on the health summary report and the health timeline, and assign at least one tag to the health summary report to obtain a health summary report configured with at least one tag; a personalized suggestion module is connected to the tag management module, and is used to analyze the user's health status based on the health summary report and the health timeline; a health management suggestion is generated based on the medical knowledge graph, the health timeline, the health summary report configured with at least one tag, and the health status through the retrieval enhancement generation technology based on the knowledge graph; the interaction module includes: a first user interaction module and a second user interaction module, wherein the first user interaction module is respectively connected to the health timeline module and the personalized suggestion module A personalized suggestion module is connected to the personalized suggestion module, and is used to display the health timeline, at least one tag, a health summary report configured with at least one tag, and an interface for health management suggestions to the user, and to provide an editing function; a second user interaction module is connected to the personalized suggestion module, and is used to display the user's health timeline, at least one tag, a health summary report configured with at least one tag, and an interface for health management suggestions to the medical party, and to generate or adjust a treatment plan based on the health timeline, at least one tag, a health summary report configured with at least one tag, and health management suggestions; a reminder module is connected to the interaction module, and is used to send a health summary report and health management suggestions configured with at least one tag to the user and / or the medical party; and to provide real-time reminders based on the acquired health data of the user.

[0027] Optionally, the data acquisition module is also used to automatically record a video of a first preset length of time for the medical record through a designated APP before the user takes a photo, in the process of acquiring the user's health data through the image acquisition record in the smart terminal according to the pre-granted authority, and to obtain at least three images that meet the preset criteria by prompting the user to stay for a second preset length of time when shooting the medical record, calculate the image sharpness value of each frame from the video of the first preset length of time using an operator, select at least two frames of images with the highest image sharpness values, combine them with at least three images that meet the preset criteria to generate an image set, perform a preliminary scan of the image set through OCR, classify the documents in combination with the user's medical history information, and classify the classified documents into It is assigned to the OCR model corresponding to the specific field for processing, and the image information corresponding to the medical terms and values ​​is extracted. The image information is compared and analyzed with the image set. Through confidence weighting, the weighted comparative analysis results are input into the large language model for fusion to obtain the user's health data; wherein, if the image sharpness value of each frame does not reach the preset threshold, the user is prompted to retake the photo; in the process of assigning the classified documents to the OCR model corresponding to the specific field for processing, if there are specific characters whose character classification probability in the OCR recognition of all pictures is lower than the confidence threshold, and no reasonable results can be obtained in the medical context through the large language model, the specific characters will be marked, and the user will be prompted to check and confirm.

[0028] Optionally, the data collection module is also used to collect the user's voice through an automatic speech recognition model in the process of obtaining the user's health data through the audio collection records in the smart terminal according to the pre-granted authority; when the user's voice meets the preset threshold, it is classified according to the medical field; when the user's voice does not meet the preset threshold, the user is prompted to re-record; after classification according to the medical field, the automatic speech recognition model of the corresponding field is loaded for recognition to obtain medical terms, data corresponding to the medical terms and user-specific information, and the user's health data is generated based on the medical terms, data corresponding to the medical terms and user-specific information; wherein, the medical terms, data corresponding to the medical terms and user-specific information are judged for rationality through a large language model, and if the semantics are incoherent or do not meet the medical reasonable expression, the user is prompted to confirm.

[0029] Optionally, the data collection module is also used to obtain the user's health data through the local records of the smart terminal and the records of the online platform in the smart terminal according to pre-granted permissions. When obtaining the user's health data in the local records of the smart terminal, the locally stored health records of the smart terminal selected by the user are obtained, and the health data of the user is obtained by inputting a large language model for fusion; when obtaining the user's health data in the records of the online platform, the platform is authorized to jump to the online platform, the user's health records are selected, and the health data of the user is obtained by inputting a large language model for fusion.

[0030] Optionally, the data collection module is also used to collect the user's health data in real time during the process of obtaining the user's health data through collection records of smart terminals and / or wearable devices based on pre-granted permissions, wherein the user's health data collected in real time includes at least: heart rate, number of steps, sleep and blood sugar.

[0031] Optionally, the data sorting module is also used to divide the user's health data into a core data layer and an extended data layer; detect whether the data in the core data layer and the data in the extended data layer belong to the standard field range; if the data in the core data layer and the data in the extended data layer belong to the standard field range, perform structured storage to obtain target format data; if the data in the core data layer and the data in the extended data layer do not belong to the standard field range, analyze the nature, importance and frequency of occurrence of the data through a large language model, evaluate whether it is necessary to create a new field based on the nature, importance and frequency of occurrence of the data, and recommend suggestions, display the analysis results to the user and obtain confirmation, create fields based on user confirmation, or expand the supplementary record layer in the data layer; wherein, the large language model is used to count the historical occurrence frequency of similar data, analyze the degree of association with existing medical knowledge and evaluate the clinical application value; and the large language model is used to determine the corresponding classification of the created field, establish the association relationship between the classification and the field, display the classification results to the user and obtain confirmation; medical terminology recognition and standardized mapping are performed on the data in the core data layer and the data in the extended data layer based on the large language model, and new terms are classified and associated, and the terminology knowledge graph of the large language model is dynamically updated.

[0032] Optionally, the data sorting module is also used to perform standard field verification and non-standard data verification on the data of the core data layer and the data of the extended data layer, wherein standard field verification includes: format verification, value range verification, logic verification and timing verification; non-standard data verification includes: rationality verification through a large language model, checking the consistency of the fields, and performing data value assessment; the data sorting module is also used to set an incremental update function, wherein the incremental update function includes: single data update and dynamic field update, and the single data update includes: supporting incremental updates of single inspection data, only processing and storing newly uploaded data items and / or retaining historical data, and constructing a time series; dynamic field updates include: continuously evaluating the usage of newly created fields through a large language model, automatically adjusting the importance of fields based on usage frequency and value analysis, and / or maintaining and optimizing the association between fields; the data sorting module is also used to set a processing feedback optimization function, wherein the processing feedback optimization function includes: collecting usage feedback of the first user interaction module and the second user interaction module, analyzing the feedback through a large language model, automatically optimizing the field classification and association corresponding to the feedback, and updating the data processing rules.

[0033] Optionally, the data sorting module is also used to analyze the user's health data, store the health data according to the classification, and obtain a health database, wherein the storage according to the classification of the health data includes: if the health data includes structured data, the structured data is stored in the structured database; if the health data includes medical text, semantic vectors are extracted from the medical text and stored; if the health data includes medical information and user detection data, it is stored through a knowledge graph.

[0034] Optionally, the knowledge graph construction module is also used to clean the health data in the health database, and to associate the cleaned health data with pre-stored medical knowledge to construct a medical knowledge graph for the corresponding user, and to obtain a verified medical knowledge graph by performing data consistency verification on the constructed medical knowledge graph.

[0035] Optionally, the health timeline module is also used to establish a time association between all health data and corresponding nodes in the medical knowledge graph based on the time of occurrence of events in the health database, semantically annotate key events through a large language model, extract the core concepts and attributes of the events, and link them with the corresponding concepts in the medical knowledge graph to generate an interactive health timeline, wherein the health timeline includes: event nodes, which are associated with the medical knowledge graph to indicate the medical knowledge associated with the event nodes; when new health data is generated, the new health data is updated to the health timeline.

[0036] Optionally, the reminder module is also used to send a health summary report and health management suggestions configured with at least one tag to the user in a specified manner according to a preset time period, and / or push the user's health dynamics to the medical party; the reminder module is also used to perform real-time detection and warning based on the user's health data collected by the data acquisition module; the reminder module is also used to compare the health data collected by the data acquisition module with the health data collected by the wearable device, and issue a warning reminder based on the comparison result; when the data acquisition module is suspended or abnormal, a prompt message is sent to the user, and the prompt message is used to indicate that the data acquisition module upload is abnormal or interrupted, or that the connection status of the wearable device is abnormal or interrupted; when the data acquisition module collects health data that poses a health risk or the data upload interruption time is greater than a preset period, and the collection record cannot be obtained through the wearable device, an emergency notification is made through a preset contact; when the data acquisition module collects health data that poses a health risk or the data upload interruption time is greater than a preset period, and the collection record cannot be obtained through the wearable device, a warning reminder is issued to the user and the medical party, so that the user and the medical party perform emergency operations based on the warning reminder.

[0037] Optionally, the personalized suggestion module is also used to dynamically update the suggestions based on changes in the user's health status.

[0038] Optionally, the personalized health management system based on artificial intelligence also includes: a data security module, wherein the data security module is connected to the reminder module, and is used to keep the user's health data confidential through end-to-end data encryption and preset access control policies, and to process personal data related to health data not related to treatment according to the user's control instructions to obtain processed health data.

[0039] The present invention provides a personalized health management method based on artificial intelligence, which is applied to a personalized health management system based on artificial intelligence, including: collecting health data of at least two data sources of a user; performing format conversion on the health data of at least two data sources to obtain data in a target format, and building a health database based on the data in the target format; generating a medical knowledge graph corresponding to the user by associating the health database with pre-stored medical knowledge; performing event semantic annotation through the time series in the health data of the user based on the medical knowledge graph and the health database to obtain the health timeline of the user; extracting the key health indicator data of the user based on the medical knowledge graph; analyzing the key health indicator data based on the medical knowledge graph and the health timeline to obtain preliminary analysis results; and generating a medical knowledge graph based on the medical knowledge graph through a retrieval enhancement generation technology based on the knowledge graph. , a health summary report is generated based on a health timeline and preliminary analysis results; and at least one tag is generated through a large language model based on the health summary report and the health timeline, and at least one tag is assigned to the health summary report to obtain a health summary report configured with at least one tag; the user's health status is analyzed based on the health summary report and the health timeline; health management suggestions are generated based on the medical knowledge graph, the health timeline, the health summary report configured with at least one tag, and the health status through a retrieval enhancement generation technology based on the knowledge graph; the user's health timeline, at least one tag, the health summary report configured with at least one tag and the health management suggestions, as well as the medical party's diagnosis and treatment plan are displayed to the user and / or the medical party through an interactive module; the health summary report configured with at least one tag and the health management suggestions are sent to the user and / or the medical party.

[0040] The present invention adopts the above technical scheme, by realizing the automatic collection and transcription of health data; using a large model to automatically summarize, label and generate personalized suggestions for health data; constructing a complete health timeline to help doctors quickly and comprehensively understand the patient's health status; and through an online consultation platform, realizing collaborative information sharing between patients and doctors. Compared with the existing technology, it has the following technical effects: reducing the data management burden of patients, improving the diagnosis and treatment efficiency of doctors, and improving the overall health management effect through intelligent functions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a schematic diagram of a personalized health management system based on artificial intelligence according to an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of data flow between modules in an artificial intelligence-based personalized health management system according to an embodiment of the present invention;

[0043] Figure 3It is a flowchart of a personalized health management method based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0045] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0046] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0047] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or units (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" / "several" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0048] Technical terms involved in this application:

[0049] AI: Artificial Intelligence;

[0050] NLP: Natural Language Processing, natural language processing;

[0051] TTS: Text-to-Speech, text to speech;

[0052] ASR: Automatic Speech Recognition, automatic speech recognition;

[0053] OCR: Optical Character Recognition, optical character recognition;

[0054] Health Management System: A system used to monitor and manage personal health data;

[0055] Tags: Tagging, used to classify and organize information;

[0056] Timeline: A sequence of events or data arranged in chronological order.

[0057] GraphMedRAG: Graph-based Medical Retrieval-Augmented Generation, a graph-based medical retrieval-augmented generation technology used to improve the reasoning and generation capabilities of large language models in personalized health management.

[0058] Example 1

[0059] An exemplary embodiment of the present invention is as follows Figure 1 As shown, Figure 1 is a schematic diagram of a personalized health management system based on artificial intelligence according to an embodiment of the present invention; a personalized health management system based on artificial intelligence provided by an embodiment of the present application includes:

[0060] Data collection module 11, data sorting module 12, knowledge graph construction module 13, automated summary module 14, tag management module 15, health timeline module 16, personalized suggestion module 17, interaction module 18 and reminder module 19, wherein the data collection module 11 is used to obtain the user's health data through image acquisition records, audio acquisition records, local records of the smart terminal, records of the online platform, and / or collection records of wearable devices in the smart terminal according to the pre-granted permissions; the data sorting module 12 is connected to the data collection module 11, and is used to convert the user's health data into target format data, and establish a health database based on the target format data; the knowledge graph construction module 13, The data sorting module 12 is connected to the data sorting module 12, and is used to associate the pre-stored medical knowledge with the health database to generate a medical knowledge graph corresponding to the user; the health timeline module 16 is connected to the knowledge graph construction module 13, and is used to perform semantic annotation of events through the time sequence in the user's health data based on the medical knowledge graph and the health database to obtain the user's health timeline; the automatic summary module 14 is respectively connected to the knowledge graph construction module 13 and the health timeline module 16, and is used to extract the user's key health indicator data based on the medical knowledge graph, analyze the key health indicator data based on the medical knowledge graph and the health timeline, and obtain preliminary analysis results; and through the retrieval enhancement generation technology based on the knowledge graph, the user's health key indicator data is analyzed based on the medical knowledge graph and the health timeline to obtain preliminary analysis results. The tag management module 15 is connected to the automatic summary module 14 and is used to generate at least one tag through the large language model according to the health summary report and the health timeline, and assign at least one tag to the health summary report to obtain a health summary report configured with at least one tag; the personalized suggestion module 17 is connected to the tag management module 15 and is used to analyze the user's health status according to the health summary report and the health timeline; and health management suggestions are generated according to the medical knowledge graph, the health timeline, the health summary report configured with at least one tag and the health status through the retrieval enhancement generation technology based on the knowledge graph; the interaction module 18 includes: a first The user interaction module 181 and the second user interaction module 182, wherein the first user interaction module 181 is connected to the health timeline module 16 and the personalized suggestion module 17, respectively, and is used to display the health timeline, at least one tag, a health summary report configured with at least one tag, and an interface of health management suggestions to the user and provide an editing function; the second user interaction module 182 is connected to the personalized suggestion module 17, and is used to display the user's health timeline, at least one tag, a health summary report configured with at least one tag, and an interface of health management suggestions to the medical party, and generate or adjust the diagnosis and treatment plan based on the health timeline, at least one tag, a health summary report configured with at least one tag, and health management suggestions;The reminder module 19 is connected to the interaction module 18 and is used to send a health summary report and health management suggestions configured with at least one tag to the user and / or the medical party; and to make real-time reminders based on the acquired health data of the user. ;

[0061] The personalized health management system based on artificial intelligence provided by the embodiment of the present application is as follows Figure 2 As shown, Figure 2 Schematic diagram of data flow between modules in a personalized health management system based on artificial intelligence according to an embodiment of the present invention; the details are as follows:

[0062] Optionally, the data acquisition module 11 is also used to automatically record a video of a first preset length of time for the medical record through a designated APP before the user takes a photo during the process of acquiring the user's health data through image acquisition records in a smart terminal based on pre-granted permissions, and to obtain at least three images that meet preset standards by prompting the user to stay for a second preset length of time when shooting the medical record, calculate the image sharpness value of each frame from the video of the first preset length of time using an operator, select at least two frames of images with the highest image sharpness values, combine them with at least three images that meet the preset standards, generate an image set, perform a preliminary scan of the image set through OCR, classify the documents in combination with the user's medical history information, and classify the classified documents The OCR model assigned to the corresponding specific field is processed, the image information corresponding to the medical terms and values ​​is extracted, the image information is compared and analyzed with the image set, and the weighted comparative analysis results are input into the large language model for fusion through confidence weighting to obtain the user's health data; wherein, if the image sharpness value of each frame does not reach the preset threshold, the user is prompted to retake the photo; in the process of assigning the classified documents to the OCR model corresponding to the specific field for processing, if there are specific characters whose character classification probability in the OCR recognition of all pictures is lower than the confidence threshold, and no reasonable results can be obtained in the medical context through the large language model, the specific characters will be marked and the user will be prompted to check and confirm.

[0063] Specifically, in the embodiment of the present application, the data acquisition module 11 obtains the use permission of the sensor and / or APP from the smart terminal according to the pre-granted permission. The image acquisition record in the smart terminal is obtained by taking photos and videos by obtaining the permission of the camera, for example, the user's medical records, test reports, diagnosis reports and other written reports provided by medical institutions;

[0064] In the embodiment of the present application, the image acquisition record of the written report obtained by the user in the medical institution is obtained through the intelligent terminal, and the paper report and document obtained by the user from the medical institution is scanned and extracted through the OCR technology; the details are as follows:

[0065] First, the data acquisition module 11 in the embodiment of the present application automatically records a 5-second (i.e., the first preset duration in the embodiment of the present application) video through an App (i.e., the designated App in the embodiment of the present application) before the user takes a photo, and prompts the user to pause for a while (i.e., the second preset duration in the embodiment of the present application) when taking a photo to capture three high-quality pictures (i.e., at least three images that meet the preset standard in the embodiment of the present application). The specific steps are as follows:

[0066] The image sharpness value of each frame is calculated using the Laplacian operator (i.e., the operator in the embodiment of the present application) from the 5 seconds of video before shooting (i.e., the first preset duration in the embodiment of the present application), and the two frames with the highest sharpness values ​​are selected, among which at least two frames of images (wherein, images with image sharpness values ​​exceeding the preset threshold are selected, and two frames of images with image sharpness values ​​TOP2 are selected), and combined with the three pictures at the time of shooting (i.e., at least three images that meet the preset standards in the embodiment of the present application), a total of five images (i.e., the image set in the embodiment of the present application) are generated for OCR processing. If the sharpness values ​​of the video frames do not meet the threshold requirements, the system will prompt the user to reshoot. (That is, if the image sharpness values ​​of each frame in the embodiment of the present application do not meet the preset threshold, the user is prompted to reshoot)

[0067] For these five qualified images, the system first uses a small OCR model to perform a preliminary scan and classifies the documents based on the user's medical history information.

[0068] After classification, the document will be assigned to the corresponding domain-specific OCR model for further processing. The domain-specific model is specially trained for numeric and medical terminology to ensure higher recognition accuracy.

[0069] Finally, the system compares and analyzes the recognition results of the five images, weights them using Laplacian confidence (i.e., confidence in the embodiment of the present application), and feeds them into a large language model for fusion to determine the final recognition result.

[0070] If the character classification probability of a certain character (i.e., a specific character in the embodiment of the present application) under the OCR of all images is lower than the confidence threshold (where 0.97 is an example of the confidence threshold in the embodiment of the present application), and a certain text content (i.e., a specific character in the embodiment of the present application) is judged to be unreasonable in the medical context through the large language model, the system will automatically mark the relevant content for the user to check and confirm.

[0071] Optionally, the data collection module 11 is also used to collect the user's voice through an automatic speech recognition model in the process of obtaining the user's health data through the audio collection records in the smart terminal according to the pre-granted authority, and when the user's voice meets the preset threshold, it is classified according to the medical field; when the user's voice does not meet the preset threshold, the user is prompted to re-record; after classification according to the medical field, the automatic speech recognition model of the corresponding field is loaded for recognition to obtain medical terms, data corresponding to the medical terms and user-specific information, and the user's health data is generated based on the medical terms, data corresponding to the medical terms and user-specific information; wherein, the medical terms, data corresponding to the medical terms and user-specific information are judged for rationality through a large language model, and if the semantics are incoherent or do not meet the medical reasonable expression, the user is prompted to confirm.

[0072] Specifically, in the embodiment of the present application, the data acquisition module 11 obtains the use permission of the sensor and / or APP from the smart terminal according to the pre-granted permission, and the audio acquisition record in the smart terminal is obtained by obtaining the permission of the microphone. For example, the user reads the medical record, test report, diagnosis report and other written reports provided by the medical institution into the microphone;

[0073] In the embodiment of the present application, the audio collection record of the written report obtained by the user in the medical institution is obtained through the intelligent terminal and transcribed through ASR; the details are as follows:

[0074] First, a small ASR model is used for rough recognition. Only when the confidence of the preliminary recognition reaches a preset threshold (i.e., 0.85 is an example of a preset threshold in the embodiment of the present application) (i.e., when the user's voice meets the preset threshold in the embodiment of the present application), the personalized health management system based on artificial intelligence in the embodiment of the present application will perform field classification (i.e., classification according to the medical field in the embodiment of the present application); otherwise, the user will be prompted to re-enter the voice. After determining the field, the system loads the corresponding field-specific ASR model for fine recognition (i.e., the automatic speech recognition model of the corresponding field is loaded for recognition in the embodiment of the present application) to improve the accuracy of speech transcription, especially in the recognition of medical terms and patient-specific information. Similarly, the personalized health management system based on artificial intelligence in the embodiment of the present application will use a large language model to judge the rationality of the transcribed content. If semantic incoherence or medically unreasonable expressions are found, the user will be prompted to confirm.

[0075] Optionally, the data acquisition module 11 is also used to, in the process of obtaining the user's health data through the local records of the smart terminal and the records of the online platform in the smart terminal according to pre-granted permissions, when obtaining the user's health data in the local records of the smart terminal, obtain the health records stored locally in the smart terminal selected by the user, and obtain the user's health data by inputting a large language model for fusion; when obtaining the user's health data in the records of the online platform, jump to the online platform through platform authorization, select the user's health records, and obtain the user's health data by inputting a large language model for fusion.

[0076] Specifically, in the embodiment of the present application, the data acquisition module 11 obtains access rights of the storage medium and / or the online network platform from the smart terminal according to the pre-granted permissions. The local records in the smart terminal and the records of the online platform are obtained by obtaining the storage medium and the online network platform of the smart terminal, for example, the notes and / or remarks stored by the user on the smart terminal; the records of the online network platform may be log posts of social networking platforms;

[0077] In the embodiment of the present application, the notes and / or remarks stored by the user on the smart terminal are obtained through the smart terminal; the records of the online platform can be log posts on the social platform; specifically as follows: the user publishes a health diary fragment similar to Weibo through a mobile device or a web application;

[0078] It should be noted that the smart terminal in the embodiment of the present application may include: a smart phone, a laptop computer or a tablet computer; among them, the smart terminal in the embodiment of the present application takes a smart phone as an example, and the personalized health management system based on artificial intelligence provided in the embodiment of the present application is only illustrated by taking the above example as an example, without specific limitation.

[0079] Optionally, the data collection module 11 is also used to collect the user's health data in real time during the process of obtaining the user's health data through collection records of smart terminals and / or wearable devices based on pre-granted permissions, wherein the user's health data collected in real time includes at least: heart rate, number of steps, sleep and blood sugar.

[0080] Specifically, in the embodiment of the present application, the data acquisition module 11 obtains the access rights of the collection records of the wearable device through the smart terminal and / or the wearable device according to the pre-granted permissions, and the real-time collected user health status obtained by the smart terminal from the wearable device, or the real-time collected user health status obtained by the wearable device;

[0081] In the process of acquiring the user's health data through the collection records of the smart terminal and / or the wearable device in the embodiment of the present application, the user's health data is collected in real time; specifically as follows:

[0082] The wearable devices in the embodiments of the present application may include: wearable devices such as smart bracelets, watches, health applications on smart phones, etc.; the user's health data obtained through collection and recording by smart terminals and / or wearable devices include: heart rate, number of steps, sleep, blood sugar, etc.

[0083] It should be noted that the personalized health management system based on artificial intelligence provided in the embodiment of the present application is only illustrated by taking the above example as an example, and is not specifically limited.

[0084] Optionally, the data sorting module 12 is also used to divide the user's health data into a core data layer and an extended data layer; detect whether the data in the core data layer and the data in the extended data layer belong to the standard field range; if the data in the core data layer and the data in the extended data layer belong to the standard field range, perform structured storage to obtain target format data; if the data in the core data layer and the data in the extended data layer do not belong to the standard field range, analyze the data nature, importance and frequency of occurrence through a large language model, evaluate whether a new field needs to be created based on the data nature, importance and frequency of occurrence, and recommend suggestions, display the analysis results to the user and obtain confirmation, create fields based on user confirmation, or expand the supplementary record layer in the data layer; wherein, the historical frequency of occurrence of similar data is counted through the large language model, the degree of association with existing medical knowledge is analyzed, and the clinical application value is evaluated; and the corresponding classification of the created field is determined through the large language model, and the association relationship between the classification and the field is established, and the classification results are displayed to the user and confirmation is obtained; medical terminology recognition and standardized mapping are performed on the data in the core data layer and the data in the extended data layer based on the large language model, and new terms are classified and associated, and the terminology knowledge graph of the large language model is dynamically updated.

[0085] Specifically, in the data sorting module 12 of the embodiment of the present application, data from different sources and formats are converted into a unified standard format to ensure the compatibility and integration of the data; wherein, the data from different sources and formats are converted into a unified standard format as follows:

[0086] The AI-based personalized health management system provided in the embodiment of the present application adopts an intelligent multi-level data standardization solution based on a large language model to achieve unified data processing and dynamic expansion:

[0087] In the embodiment of the present application, when the data sorting module 12 performs unified data processing, the hierarchical data structure design includes:

[0088] The core data layer (fixed field layer) includes:

[0089] Basic information layer: basic patient information, medical information, medical insurance information, etc.;

[0090] Medical document layer: chief complaint, current medical history, physical examination, auxiliary examination, diagnosis information, treatment plan, doctor's order record, etc.;

[0091] Inspection data layer: routine inspection, special inspection, microbiological inspection, pathological inspection, etc.;

[0092] Image data layer: original image examination data, structured description reports, and doctor’s diagnosis opinions;

[0093] Health log layer: user subjective records, wearable device data, and lifestyle records.

[0094] The extended data layer (dynamic field layer) includes:

[0095] New inspection item layer: stores new inspection indicators other than standard inspection items;

[0096] Supplementary record layer: stores information that cannot be classified into existing fields;

[0097] Special instructions layer: stores non-routine information that requires special attention.

[0098] Intelligent parsing and dynamic expansion based on a large language model in the embodiments of the present application:

[0099] Data type identification and processing:

[0100] Identify whether the input data belongs to the standard field range;

[0101] For data within the standard field range, structured storage is directly performed (i.e., structured storage is performed in the embodiment of the present application to obtain target format data);

[0102] For data outside the scope of standard fields: the large language model analyzes the nature, importance, and frequency of occurrence of the data; evaluates whether a new field needs to be created and gives recommendations; displays the analysis results to the user and obtains confirmation; decides to create a new field or store it in a supplementary record layer based on user confirmation.

[0103] In the embodiment of the present application, dynamic field management includes:

[0104] Automatic evaluation mechanism in the embodiment of the present application: the large language model counts the historical frequency of similar data; analyzes the degree of association with existing medical knowledge; and evaluates the clinical application value (i.e., the large language model counts the historical frequency of similar data, analyzes the degree of association with existing medical knowledge, and evaluates the clinical application value in the embodiment of the present application);

[0105] The intelligent classification in the embodiment of the present application includes: the large language model automatically determines the most suitable classification for the new field; establishes an association relationship with related fields; displays the classification result to the user and obtains confirmation (that is, in the embodiment of the present application, the corresponding classification of the created field is determined by the large language model, the association relationship between the classification and the field is established, and the classification result is displayed to the user and obtains confirmation);

[0106] Intelligent terminology standardization in the embodiment of the present application includes: using a large language model to perform medical terminology recognition and standardized mapping; classifying and analyzing the association of new terms; and maintaining a dynamically updated terminology knowledge graph (i.e., in the embodiment of the present application, medical terminology recognition and standardized mapping are performed on the data of the core data layer and the data of the extended data layer based on the large language model, and new terms are classified and analyzed for association, and the terminology knowledge graph of the large language model is dynamically updated).

[0107] Optionally, the data sorting module 12 is also used to perform standard field verification and non-standard data verification on the data of the core data layer and the data of the extended data layer, wherein standard field verification includes: format verification, value range verification, logic verification and timing verification; non-standard data verification includes: rationality verification through a large language model, checking the consistency of the fields, and performing data value assessment; the data sorting module 12 is also used to set an incremental update function, wherein the incremental update function includes: single data update and dynamic field update, and the single data update includes: supporting incremental updates of single inspection data, only processing and storing newly uploaded data items and / or retaining historical data, and constructing a time series; dynamic field updates include: continuously evaluating the usage of newly created fields through a large language model, automatically adjusting the importance of fields based on usage frequency and value analysis, and / or maintaining and optimizing the association between fields; the data sorting module 12 is also used to set a processing feedback optimization function, wherein the processing feedback optimization function includes: collecting usage feedback of the first user interaction module 18 and the second user interaction module 18, analyzing the feedback through a large language model, automatically optimizing the field classification and association corresponding to the feedback, and updating data processing rules.

[0108] Specifically, in the embodiment of the present application, the data quality control in the data sorting module 12 is as follows:

[0109] In the embodiment of the present application, standard field verification includes:

[0110] Format validation: ensuring that data conforms to predefined format specifications;

[0111] Value range validation: check whether the data is within a reasonable range;

[0112] Logical verification: verify the logical relationship between data;

[0113] Timing verification: Ensure the temporal continuity of data.

[0114] The non-standard data validation in the embodiment of the present application includes: rationality verification through a large language model, consistency check with related fields; data value assessment (that is, rationality verification is performed through a large language model in the embodiment of the present application, consistency check of fields is performed, and data value assessment is performed).

[0115] The incremental update mechanism in the embodiment of the present application includes: single data update, wherein the single data update includes: supporting incremental update of single inspection data; only processing and storing newly uploaded data items; retaining historical data, and building a complete time series (that is, supporting incremental update of single inspection data, only processing and storing newly uploaded data items and / or retaining historical data, and building a time series in the embodiment of the present application);

[0116] In the embodiment of the present application, dynamic field updating includes: a large language model continuously evaluating the usage of newly created fields; automatically adjusting the importance of fields based on usage frequency and value analysis; maintaining and optimizing the association between fields (that is, in the embodiment of the present application, a large language model continuously evaluates the usage of newly created fields, automatically adjusts the importance of fields based on usage frequency and value analysis, and / or maintains and optimizes the association between fields).

[0117] The feedback optimization mechanism in the embodiment of the present application includes: user feedback processing: collecting usage feedback from doctors and patients; analyzing the feedback content with a large language model, and automatically optimizing field classifications and associations; and continuously updating data processing rules (i.e., collecting usage feedback from the first user interaction module 18 and the second user interaction module 18 in the embodiment of the present application, analyzing the feedback through a large language model, automatically optimizing field classifications and associations corresponding to the feedback, and updating data processing rules).

[0118] Optionally, the data sorting module 12 is also used to analyze the user's health data, store the health data according to the classification, and obtain a health database, wherein the storage according to the classification of the health data includes: if the health data includes structured data, the structured data is stored in the structured database; if the health data includes medical text, semantic vectors are extracted from the medical text and stored; if the health data includes medical information and user detection data, it is stored through a knowledge graph.

[0119] Specifically, the data sorting module 12 in the embodiment of the present application stores the standardized data in a centralized health database, supporting fast retrieval and efficient analysis.

[0120] In the embodiment of the present application, a hybrid medical knowledge graph storage solution for retrieval-enhanced generation (RAG) is used to improve existing database technology, achieve efficient storage, retrieval and analysis, and thus better support personalized health management. In the embodiment of the present application, a relational database, a vector database and a graph database are combined, and optimized for the characteristics of medical data.

[0121] Specifically, the personalized health management system based on artificial intelligence provided in the embodiment of the present application adopts the following architecture:

[0122] Structured data storage: Use PostgreSQL to store standardized medical data, such as medical records, test results, etc. PostgreSQL's JSONB field type is used to store semi-structured data to accommodate the diversity of medical data.

[0123] Semantic vector storage: Milvus vector database is used to store semantic vectors extracted from medical texts. These vectors are generated by pre-trained large language models in the medical field, capturing the semantic information of the text and supporting fast semantic-based retrieval.

[0124] Knowledge graph storage: A medical knowledge graph is constructed using the Neo4j graph database. The nodes in the graph represent medical entities (such as diseases, symptoms, and drugs), and the edges represent the relationships between entities. The knowledge graph not only stores prior knowledge in the medical field, but also integrates personalized information extracted from user data.

[0125] The advantages of hybrid storage solutions are:

[0126] Efficient retrieval: Vector databases support efficient similarity search and can quickly retrieve relevant information based on user input. Graph databases support complex graph traversal operations and can be used to explore the associations between different medical entities.

[0127] Knowledge reasoning: Knowledge graphs provide a basis for rule-based reasoning and knowledge discovery, and can be used for auxiliary diagnosis, personalized treatment plan recommendations, etc.

[0128] Support for RAG: This storage solution is deeply integrated with GraphMedRAG technology to ensure efficient knowledge retrieval and generation. The stored semantic vectors and knowledge graphs can be directly used to enhance the prompt of LLM, thereby generating more accurate and comprehensive health analysis reports and personalized recommendations.

[0129] Dynamic expansion: The solution can be flexibly expanded to adapt to new data types and new medical knowledge.

[0130] Compared with traditional database technology, the improvements of this solution are:

[0131] Richer semantic expression: Vector databases and graph databases can capture the semantic information of medical data, while traditional databases mainly focus on the structured storage of data.

[0132] More powerful retrieval and analysis capabilities: Vector databases support semantic retrieval, and graph databases support graph traversal and reasoning, which are difficult to achieve with traditional databases.

[0133] More flexible scalability: The hybrid storage architecture can be flexibly expanded according to actual needs to adapt to the ever-changing medical data and knowledge.

[0134] Optionally, the knowledge graph construction module 13 is also used to clean the health data in the health database, and to associate the cleaned health data with pre-stored medical knowledge to construct a medical knowledge graph for the corresponding user, and to obtain a verified medical knowledge graph by performing data consistency verification on the constructed medical knowledge graph.

[0135] Specifically, the knowledge graph construction module 13 in the embodiment of the present application constructs and maintains a medical knowledge graph based on medical data and domain knowledge to provide knowledge support for personalized health management.

[0136] Among them, the OCR scan results, ASR transcription texts and fragmented health diaries uploaded by users are integrated in real time and directly stored in the hybrid medical knowledge graph database:

[0137] (1) Data cleaning: remove duplicates, process missing values, and detect outliers in the collected data. Use statistical methods and pre-trained medical big language models to identify and process abnormal and erroneous data.

[0138] (2) Knowledge graph construction: The cleaned data is directly used to build and update the user's personalized medical knowledge graph. The knowledge graph construction process includes: extracting medical entities and relationships from text data; storing the extracted information in the Neo4j graph database; storing the semantic vector of the text in the Milvus vector database to support semantic-based retrieval.

[0139] (3) Data consistency verification: Use a large language model to perform consistency verification on the integrated knowledge graph (i.e., the medical knowledge graph of the corresponding user in the embodiment of the present application), for example, to check whether there are inconsistencies or conflicts between different data sources.

[0140] Through the above steps, the artificial intelligence-based personalized health management system provided in the embodiment of the present application can integrate the multi-source heterogeneous data uploaded by users into the hybrid medical knowledge graph database (i.e., the health database in the embodiment of the present application) in real time, providing a high-quality, structured data foundation for subsequent health analysis and personalized recommendation generation.

[0141] Optionally, the health timeline module 16 is also used to establish a time association between all health data and corresponding nodes in the medical knowledge graph based on the time of occurrence of events in the health database, semantically annotate key events through a large language model, extract core concepts and attributes of the events, and link them with corresponding concepts in the medical knowledge graph to generate an interactive health timeline, wherein the health timeline includes: event nodes, which are associated with the medical knowledge graph to indicate medical knowledge associated with the event nodes; when new health data is generated, the new health data is updated to the health timeline.

[0142] Specifically, the process of generating the health timeline by the health timeline module 16 in the embodiment of the present application is as follows:

[0143] Input: integrated health data and event information (i.e., the health database in the embodiment of the present application), and the user's personalized medical knowledge graph (i.e., the medical knowledge graph in the embodiment of the present application).

[0144] Processing process:

[0145] Time series enhancement: All health data, important events (such as surgery, hospitalization), medical records, etc. are sorted according to the time of event occurrence, and time association is established with relevant nodes in the knowledge graph. For example, diagnosis events are associated with disease nodes, and medication events are associated with drug nodes.

[0146] Event semantic annotation: Use the large language model to semantically annotate key events, extract the core concepts and attributes of the events, and link them with relevant concepts in the knowledge graph (i.e., the medical knowledge graph in the embodiment of this application). For example, the event "diagnosed with type 2 diabetes" will be annotated as "Disease: type 2 diabetes" and linked to the "type 2 diabetes" node in the knowledge graph.

[0147] Interactive timeline visualization: Generate an interactive health timeline that supports functions such as zooming in, zooming out, and filtering specific events. The event nodes on the timeline are associated with the knowledge graph (i.e., the medical knowledge graph in the embodiment of the present application), and you can click on the event node to view related medical knowledge.

[0148] Real-time synchronous updates: New health data and events are updated to the timeline and knowledge graph in real time to ensure the integrity and timeliness of the information.

[0149] Output result: A visualized health timeline showing the patient's complete health history for viewing by the patient (ie, the user in the embodiment of the present application) and the doctor.

[0150] In the embodiment of the present application, the automated summary module 14 automatically generates a health summary report based on GraphMedRAG; specifically, as follows:

[0151] Input: User’s personalized medical knowledge graph and health timeline data.

[0152] Processing process: Extraction of key indicators based on the graph (i.e., the medical knowledge graph in the embodiment of the present application): Using the graph structure, extract key indicators closely related to the user's health status (i.e., the health key indicator data in the embodiment of the present application), such as blood sugar, blood pressure, weight, etc. Trend analysis based on the graph: Based on the health timeline and knowledge graph, analyze the changing trends and potential risks of key indicators (i.e., the health key indicator data is analyzed based on the medical knowledge graph and the health timeline in the embodiment of the present application to obtain preliminary analysis results). For example, the system can predict the risk of diabetes based on the user's blood sugar change trend and family history. Natural language generation based on GraphMedRAG: Using GraphMedRAG technology, combined with knowledge graph and timeline information, a more comprehensive and accurate health summary report is generated (i.e., the health summary report is generated based on the medical knowledge graph, the health timeline and the preliminary analysis results through the retrieval enhancement generation technology based on the knowledge graph in the embodiment of the present application). GraphMedRAG enhances the prompt of LLM through the graph structure, enabling it to better understand the contextual information of medical data and generate more insightful summaries. Multi-dimensional summary classification: Generate corresponding health summary content according to different time periods (daily, weekly, monthly) and different health indicators.

[0153] Output: Automated daily and weekly health summary reports with key indicators, trend analysis and health insights.

[0154] The process of performing intelligent tag management by the tag management module 15 in the embodiment of the present application is specifically as follows:

[0155] Input: Automatically generated health summary report and health timeline data (ie, the health timeline in the embodiment of the present application).

[0156] Processing process: Label generation: Use the large language model to analyze the summary report content and timeline events, and automatically generate relevant labels (that is, generate at least one label through the large language model based on the health summary report and health timeline in the embodiment of the present application), such as "hyperglycemia", "successful weight loss", "drug adjustment", etc. Label allocation: Assign the generated labels to the corresponding health data and summary reports, and support multi-dimensional classification (that is, assign at least one label to the health summary report in the embodiment of the present application). Label optimization: Dynamically adjust the label generation model based on feedback from users and doctors to improve the accuracy and relevance of labels.

[0157] Output results: health data, summary reports and timelines with smart labels, supporting efficient data retrieval and classification analysis.

[0158] Optionally, the personalized suggestion module 17 is also used to dynamically update the suggestions according to changes in the user's health status.

[0159] The process of generating personalized health advice based on the knowledge graph and GraphMedRAG by the personalized advice module 17 in the embodiment of the present application is as follows:

[0160] Input: User’s personalized medical knowledge graph, health timeline data, and automated health summary report.

[0161] Processing: Graph-based health status analysis: Based on the knowledge graph and health timeline, analyze the user's health status, potential risks, and the effectiveness of past treatment plans. GraphMedRAG-based personalized recommendation generation: Using GraphMedRAG technology, combined with knowledge graphs, timelines, and health summaries, more accurate and personalized health management recommendations are generated. For example, the system can recommend appropriate diet, exercise, and drug adjustment plans based on the user's medical history, medication history, lifestyle, and current health status.

[0162] In the embodiment of the present application, the personalized suggestion module 17 generates personalized health management suggestions based on the AI ​​analysis results, including diet adjustment, exercise plan, drug management, etc. Dynamic suggestion updates are provided to respond to changes in the patient's health status at any time.

[0163] The interaction module 18 in the embodiment of the present application is used to provide interaction functions for users and doctors, wherein the first user interaction module 181 is an interaction module applicable to users; the second user interaction module 182 is an interaction module applicable to doctors or medical institutions;

[0164] Among them, the first user interaction module 181 provides a user-friendly interface to support patients to view health summaries, manage tags, edit health diaries and receive health advice. Similar to social media platforms, it allows users to publish fragmented health diaries to enhance user interactivity and participation.

[0165] The second user interaction module 182 provides a dedicated interface for doctors to display the patient's complete health timeline, key health summary and smart tags, supporting doctors to view the patient's health recommendations and adjust the diagnosis and treatment plan.

[0166] Specifically, in the embodiment of the present application, the interaction and feedback between the user and the doctor are as follows:

[0167] Enter: Health Summary Report, Smart Tags, Health Timeline, and Health Recommendations.

[0168] Processing process:

[0169] User interface display: In the patient-side application, health summaries, labels, timelines, and health recommendations are displayed, and visualization tools are provided (i.e., in the embodiment of the present application, a health timeline, at least one label, a health summary report configured with at least one label, and an interface for health management recommendations are displayed to the user, and editing functions are provided).

[0170] Doctor-side interface: provides a dedicated interface for doctors to display the patient's complete health timeline, key health summaries and smart tags, supporting quick browsing and in-depth analysis (i.e., in the embodiment of the present application, the interface displays the user's health timeline, at least one tag, a health summary report configured with at least one tag, and health management suggestions to the medical party, and generates or adjusts the diagnosis and treatment plan based on the health timeline, at least one tag, a health summary report configured with at least one tag, and health management suggestions).

[0171] Interactive operations: Patients can edit their health diaries, add notes, and mark key events; doctors can view and provide feedback on patients’ health conditions and suggestions.

[0172] Feedback mechanism: Collect feedback from patients and doctors on health advice and system functions for continuous improvement and optimization of system performance.

[0173] Output: Enhanced user and physician experience, facilitating communication and collaboration between patients and physicians.

[0174] Optionally, the reminder module 19 is also used to send a health summary report and health management suggestions configured with at least one tag to the user in a specified manner according to a preset time period, and / or push the user's health dynamics to the medical party; the reminder module 19 is also used to perform real-time detection and warning based on the user's health data collected by the data acquisition module 11; the reminder module 19 is also used to compare the health data collected by the data acquisition module 11 with the health data collected by the wearable device, and to issue a warning reminder based on the comparison result; when the data acquisition module 11 is suspended or abnormal, a prompt message is sent to the user, and the prompt message is used to indicate that the data acquisition module 11 uploads abnormally or is interrupted, or that the connection status of the wearable device is abnormal or interrupted; when the data acquisition module 11 collects health data with health risks or the data upload interruption time is greater than the preset period, and the collection record cannot be obtained through the wearable device, an emergency notification is made through a preset contact; when the data acquisition module 11 collects health data with health risks or the data upload interruption time is greater than the preset period, and the collection record cannot be obtained through the wearable device, a warning reminder is issued to the user and the medical party, so that the user and the medical party perform emergency operations based on the warning reminder.

[0175] Specifically, in the embodiment of the present application, the reminder module 19 sends regular health summary reports and personalized suggestions to patients and doctors, and reminds important health indicators and events through push notifications, emails or text messages, etc. It supports real-time health warnings, such as high blood sugar risks, medication reminders, etc.

[0176] In the embodiment of the present application, the reminder module 19 performs as follows:

[0177] Input: daily and weekly (i.e., preset time periods in the embodiment of the present application) health summary reports (i.e., health summary reports configured with at least one label in the embodiment of the present application), personalized health advice (i.e., health management advice in the embodiment of the present application), changes in user health data (i.e., the health dynamics of the user in the embodiment of the present application), and real-time health data from associated wearable devices.

[0178] Processing process:

[0179] Regular notifications and reminders: The AI-based personalized health management system in the embodiment of the present application sends health summary reports and suggestions (i.e., health summary reports and health management suggestions configured with at least one label in the embodiment of the present application) to patients (i.e., users in the embodiment of the present application) through push notifications, emails, or text messages (i.e., the specified methods in the embodiment of the present application) according to a preset time period (i.e., the preset time period in the embodiment of the present application), and pushes the patient's key health dynamics (i.e., the health dynamics in the embodiment of the present application) to doctors.

[0180] Real-time reminders and warnings: Combine the health data uploaded by the user and the real-time data from the associated wearable devices (such as smart bracelets, watches, mobile health applications) to conduct more comprehensive health monitoring (i.e., the collected health data of the user in the embodiment of this application is detected and warned in real time). For example, the system can issue more timely health warnings based on the user's real-time heart rate, exercise status and blood sugar data.

[0181] Data interruption plan: Multi-source data cross-validation: The system cross-validates the data uploaded by the user with the data from the wearable device to improve data reliability and identify potential data upload interruptions or abnormal situations. For example, if the user stops uploading blood sugar data, but the smartwatch continues to monitor abnormal blood sugar values, the system will issue an alert (i.e., the health data collected by the data acquisition module 11 in the embodiment of the present application is compared with the health data collected by the wearable device, and an early warning reminder is issued based on the comparison results).

[0182] Active care reminder: When the system detects an interruption or abnormal mode in data upload, it will actively send a care reminder to the user, ask about his or her health status, and encourage the user to continue uploading data or check the connection status of the wearable device (that is, in the embodiment of the present application, when the data acquisition module 11 is suspended or abnormal, a prompt message is sent to the user, and the prompt message is used to indicate that the data acquisition module 11 is uploading abnormally or interrupted, or that the connection status of the wearable device is abnormal or interrupted).

[0183] Emergency contact notification: Users can set emergency contacts in advance. When the system detects serious health risks or the data interruption time is too long and valid data cannot be obtained through the wearable device, the emergency contact can be notified (that is, in the embodiment of the present application, when the data acquisition module 11 collects health data that poses a health risk or the data upload interruption time is greater than the preset period, and the collection record cannot be obtained through the wearable device, emergency notification is made through the preset contact).

[0184] Collaborative reminder: When the patient's health condition is abnormal or data upload is interrupted, the system notifies the patient and the doctor at the same time to facilitate timely measures (that is, in the embodiment of the present application, when the data acquisition module 11 collects health data that poses a health risk or the data upload interruption time is greater than the preset period, and the collection record cannot be obtained through the wearable device, an early warning reminder is issued to the user and the medical party, so that the user and the medical party can perform emergency operations based on the early warning reminder).

[0185] Optionally, the personalized health management system based on artificial intelligence in the embodiment of the present application also includes: a data security module 10, wherein the data security module 10 is connected to the reminder module 19, and is used to keep the user's health data confidential through end-to-end data encryption and preset access control policies, and process personal data related to health data not related to treatment according to the user's control instructions to obtain processed health data.

[0186] Specifically, end-to-end data encryption and strict access control policies are used to ensure the security and privacy of patients' health data. Patients are supported to mark the information they want doctors to view through options, enabling selective sharing of content.

[0187] In the embodiment of the present application, the data security module 10 performs security and privacy protection as follows:

[0188] Input: All patient health data and interaction records.

[0189] Processing process:

[0190] Data encryption: End-to-end encryption of stored and transmitted health data ensures data confidentiality and integrity.

[0191] For example, using the industry-standard AES-256 encryption algorithm to encrypt all sensitive data, including data in static storage and during transmission, can effectively ensure data security.

[0192] Fine-grained access control:

[0193] Tag-based access control: Users can set data access permissions based on predefined tag categories (e.g., “physical examination data,” “medical record information,” “lifestyle data,” etc.). Doctors can access data in a specific tag category only with user authorization.

[0194] Time range control: Users can set the time range for data sharing, for example, only allowing doctors to view data from the most recent year.

[0195] Separate authorization for sensitive data: The system uses a large language model and pre-set rules to identify potentially sensitive data, such as mental illness diagnosis, genetic information, etc. These sensitive data require separate authorization from the user before they can be shared with doctors.

[0196] Privacy protection: The system is designed to comply with GDPR and other relevant laws and regulations, and provides data anonymization and de-identification functions to protect user privacy to the greatest extent.

[0197] In the embodiment of the present application, the client local storage encryption is as follows:

[0198] AES-256 encryption algorithm is used to encrypt sensitive data. The key is generated by the password set by the user through the PBKDF2 key derivation function, ensuring that even if the password is leaked, it is difficult for attackers to crack the encrypted data.

[0199] In the embodiment of the present application, the server storage and access control are as follows:

[0200] De-identification and anonymization: Data uploaded to the server is de-identified and anonymized, and the original identifier is encrypted using AES-256 and stored in an independent secure database. Role-based access control (RBAC) is used, and only authorized personnel can access it.

[0201] Attribute-based encryption (ABE): Attribute-based encryption (ABE) technology is used for users' health data. ABE allows access policies to be set based on users' attributes (e.g. age, gender, disease type). Only doctors who meet the access policy can decrypt and view the data.

[0202] Access log recording: The system records all data access operations, including visitor identity, access time, access data, etc., for auditing and tracking purposes.

[0203] Data recovery: Users need to provide multi-factor authentication (such as password + SMS verification code) to decrypt and restore the original data.

[0204] Emergency: Users can pre-set emergency contacts and authorize them to access certain critical data in an emergency. Emergency contacts need to provide authentication information and proof of their pre-set relationship with the user to gain access.

[0205] Data backup and recovery: Back up health data regularly to ensure quick recovery in case of data loss or system failure.

[0206] Output: A safe and reliable health management system that provides refined data sharing control and ensures the security and privacy of patient data.

[0207] Examples of user and doctor operations and system feedback in the artificial intelligence-based personalized health management system provided in the embodiments of the present application are as follows:

[0208] User Action 1: The patient takes a photo and uploads a paper health report obtained from the hospital.

[0209] System feedback 1:

[0210] OCR processing: The system uses OCR technology to scan and extract text content in the report.

[0211] Data integration: The extracted information is standardized and stored in health databases.

[0212] Health Timeline Update: Add new health events to the timeline.

[0213] Page display changes 1:

[0214] Notification reminder: The system notifies the patient that the report has been successfully uploaded and processed.

[0215] Timeline Updates: New health events are displayed in chronological order in the timeline.

[0216] User operation 2: The patient records the day's exercise and diet through voice input.

[0217] System feedback 2:

[0218] ASR transcription: The system transcribes speech content using ASR technology.

[0219] Data integration: Transcribed text information is standardized and stored in health databases.

[0220] Automatic tag addition: The system automatically analyzes and adds relevant tags, such as "exercise" and "healthy diet".

[0221] Page display change 2:

[0222] Real-time updates: New health records appear instantly in the health diary and timeline.

[0223] Tag display: Automatically added tags are displayed next to related records for easy classification and retrieval.

[0224] Doctor operation 1: During the online consultation, the doctor views the patient's health timeline and automatically generated health summary.

[0225] System feedback 1:

[0226] Timeline display: Doctors view the patient's complete health timeline in a dedicated interface to quickly understand the patient's health history and key events.

[0227] Health Summary and Labels: The system displays daily and weekly health summary reports and smart labels to help doctors comprehensively assess the patient's health status.

[0228] Page display changes 1:

[0229] Professional interface: The doctor-side interface displays a detailed health timeline, charts of changes in key health indicators, and smart labels.

[0230] Auxiliary suggestions: The system provides auxiliary diagnosis and treatment suggestions based on health data for doctors' reference.

[0231] User operation 3: The patient participates in the online consultation and selects the health information he or she wants the doctor to view.

[0232] System feedback 3:

[0233] Content selection: The system provides options that allow patients to check the health data and summary reports they want their doctors to view.

[0234] Privacy protection: Only the content selected by the patient is shared, ensuring the privacy of other sensitive information.

[0235] Physician notification: The system notifies the physician and provides the health information that the patient has authorized to view.

[0236] Page display changes 3:

[0237] Selection interface: Patients see a simple selection interface in the health management application, which makes it easy to check the information to be shared.

[0238] Doctor-side update: The doctor-side interface only displays the health information that the patient chooses to share, optimizing the doctor's viewing efficiency.

[0239] User operation 4: Patients view the weekly AI-generated health summary to understand their own health trends.

[0240] System feedback 4:

[0241] Health summary display: The system displays a detailed weekly health summary report, including trend analysis of key indicators and health insights.

[0242] Personalized recommendations: The system generates personalized health recommendations based on summary reports to help patients adjust their health management strategies.

[0243] Page display changes 4:

[0244] Summary report: A health summary report with pictures and text, combined with trend charts, intuitively displays health changes.

[0245] Recommendations module: Personalized health recommendations are displayed below the summary report, providing actionable insights.

[0246] The personalized health management system based on artificial intelligence provided in the embodiment of the present application provides a comprehensive solution to the challenges faced by patients and doctors in health management and medical services. By integrating OCR and ASR technologies, the automatic collection and transcription of health data can be realized; large models are used to automatically summarize, label and generate personalized suggestions for health data; a complete health timeline is constructed to help doctors quickly and comprehensively understand the health status of patients; and through the online consultation platform, information collaborative sharing between patients and doctors is achieved. The system not only reduces the data management burden of patients and improves the diagnosis and treatment efficiency of doctors, but also improves the overall health management effect and user experience through intelligent functions.

[0247] The personalized health management system based on artificial intelligence provided by the embodiment of the present application is in the process of intelligent data collection and processing: through OCR and ASR technology, the automatic collection and transcription of health data is realized, manual input is reduced, and data integration efficiency is improved. Multi-source heterogeneous medical data integration: the system can integrate multi-source heterogeneous data such as OCR, ASR, wearable devices, and use large language models for standardization and structured processing to build a complete user health portrait. In the construction and application of medical knowledge graphs: the system builds a personalized medical knowledge graph and integrates timeline information into it, providing a rich knowledge basis for health analysis and suggestion generation. Through GraphMedRAG, the reasoning ability of LLM is enhanced. Using GraphMedRAG technology, combined with knowledge graph and timeline information, the reasoning ability of LLM is enhanced, enabling it to generate more accurate and personalized health summaries and suggestions. Automated health summaries and suggestions use large models to generate daily and weekly health summary reports, extract key health indicators and trends, and assist patients and doctors in making scientific health management and diagnosis and treatment decisions. Intelligent label management automatically adds relevant labels to health data and summary reports, supports multi-dimensional data classification, and facilitates efficient retrieval and analysis. By constructing a complete health timeline, the patient's health data and events are displayed in chronological order, helping doctors to quickly and comprehensively understand the patient's health history, thereby improving the diagnosis and treatment effect. The collaborative platform between patients and doctors integrates online consultations with the health management system, allowing patients to select and authorize doctors to view specific health information, realize selective sharing of information, protect privacy and improve the doctor's diagnosis and treatment efficiency. Personalized health advice is generated based on AI analysis results. The personalized health management system based on artificial intelligence provided in the embodiment of the present application provides personalized health advice, helps patients to conduct scientific and effective self-management, and assists doctors in formulating accurate diagnosis and treatment plans. Advanced data encryption and access control strategies are used in the security and privacy protection mechanism to ensure the security and privacy protection of patient data, comply with relevant laws and regulations, and enhance user trust. The application of large models in NLP, image recognition and generation in multifunctional intelligent applications realizes the automatic processing, summary generation, label management and personalized suggestions of health data, and improves the intelligence level and user experience of the system.

[0248] In the personalized health management system based on artificial intelligence provided by the embodiment of the present application, patients can participate in health management more actively and proactively, and doctors can provide personalized medical services more efficiently, thereby improving the service level and user experience in the overall medical and health field.

[0249] Among them, GraphMedRAG in the embodiment of the present application significantly improves the reasoning and generation capabilities of the large language model (LLM) by integrating multi-source health data (such as electronic medical records, wearable device data, user health diaries) and dynamically constructing a medical knowledge graph.

[0250] Different from the traditional RAG method, GraphMedRAG adopts a hierarchical knowledge graph structure combined with semantic vector storage to achieve more efficient semantic retrieval and information fusion. GraphMedRAG technology ensures that the generated health analysis and recommendations are not only based on individualized data, but also highly accurate and interpretable, thus providing safe, reliable and intelligent support for personalized health management systems.

[0251] The technical solutions are as follows:

[0252] GraphMedRAG: Retrieval enhancement generation technology based on knowledge graph;

[0253] GraphMedRAG technology is an innovative retrieval enhancement generation framework designed for personalized health management. Its core goal is to improve the performance and accuracy of large language models in processing complex medical data through semantic understanding of knowledge graphs and efficient retrieval of semantic vectors. GraphMedRAG technology includes the following key steps:

[0254] Dynamic knowledge graph construction:

[0255] Multi-source data integration: The system collects health data in real time from multiple data sources (such as electronic medical records, wearable devices, and user health diaries), and uses OCR (optical character recognition) and ASR (automatic speech recognition) technologies to convert unstructured data into structured information.

[0256] Entity and relationship extraction: Natural language processing (NLP) and deep learning technologies are used to automatically extract key medical entities (such as diseases, symptoms, drugs) and their relationships from structured and semi-structured data to build a preliminary medical knowledge graph.

[0257] Hierarchical graph architecture: Establish a three-level knowledge graph structure:

[0258] User Layer: Contains health data and history of individual users.

[0259] Domain Layer: Integrates data from medical literature, textbooks, and standard medical dictionaries to provide domain knowledge support.

[0260] Foundation Layer: References authoritative resources such as the Unified Medical Language System (UMLS) to ensure standardization of terminology and relationships.

[0261] Semantic vector storage and indexing:

[0262] Vectorization processing: Convert text information in the knowledge graph into high-dimensional semantic vectors, store them in an efficient vector database (such as Milvus), and support fast semantic similarity retrieval.

[0263] Graph database integration: Use graph databases (such as Neo4j) to store and manage the constructed knowledge graph, supporting complex graph traversal and semantic retrieval operations.

[0264] Enhanced semantic retrieval and generation:

[0265] Semantic matching: Through semantic vector matching technology, quickly locate the medical entities and information most relevant to the user's query.

[0266] Graph-based retrieval enhancement: GraphMedRAG knowledge graph is used in combination with semantic vector retrieval results to provide rich background knowledge for large language models and enhance the accuracy and contextual relevance of generated content.

[0267] Multimodal data fusion: The system integrates multimodal data such as text, numerical data and images, and uses LLM's multimodal analysis capabilities to provide comprehensive health insights and trend analysis.

[0268] Verification and optimization of generated content:

[0269] Explainability and Verification: Ensure that the generated health recommendations and analyses are based on reliable information in the graph, provide references to source data, and make the output results highly explainable and verifiable.

[0270] Feedback-driven optimization: Through feedback from users and doctors, the knowledge graph structure and retrieval algorithm are continuously optimized to improve the overall performance of the system.

[0271] The personalized health management system based on artificial intelligence provided by the embodiment of the present application can perform efficient semantic retrieval through GraphMedRAG: combining semantic vector storage and knowledge graph to support efficient response to complex health queries. Deep information fusion: multimodal data integration to achieve a comprehensive understanding and analysis of the user's health status. Personalization and explainability: Generate explainable and accurate health recommendations based on individualized knowledge graphs to enhance user trust. Dynamic update and scalability: Real-time update and expansion of knowledge graphs to ensure that the health management system is always based on the latest medical knowledge and user data. Secure and reliable data processing: Industry-standard encryption and access control technologies are used to ensure the security and privacy of user health data.

[0272] GraphMedRAG technology provides powerful technical support for personalized health management systems through its unique knowledge graph construction and semantic retrieval capabilities, significantly improving the system's performance in data processing, information retrieval and content generation.

[0273] The dynamic hierarchical knowledge graph structure in the personalized health management system based on artificial intelligence provided by the embodiment of the present application seamlessly connects the user's personalized health data with industry standard knowledge and basic medical terminology through a unique three-level construction method to form a comprehensive and dynamically updated knowledge graph. The GraphMedRAG technology integrates semantic vector storage and graph database, and achieves high-precision response to complex health queries through GraphMedRAG enhanced semantic retrieval and information fusion. Multimodal data fusion and real-time processing can integrate multiple data types such as text, numerical values, and images, and use the stream processing framework to support real-time collection and analysis of wearable device data, thereby improving the immediacy and accuracy of health monitoring. Intelligent data standardization and dynamic expansion can automatically identify and convert data formats using large language models, and flexibly adapt to emerging health data types through dynamic field management systems to ensure the long-term scalability of the system. In addition, advanced fairness verification and content verification enhance the verifiability of generated content through graph structures, ensuring that all generated health analyses and recommendations are based on reliable data sources, reducing information illusions and erroneous reasoning.

[0274] The present invention adopts the above technical scheme, by realizing the automatic collection and transcription of health data; using a large model to automatically summarize, label and generate personalized suggestions for health data; constructing a complete health timeline to help doctors quickly and comprehensively understand the patient's health status; and through an online consultation platform, realizing collaborative information sharing between patients and doctors. Compared with the existing technology, it has the following technical effects: reducing the data management burden of patients, improving the diagnosis and treatment efficiency of doctors, and improving the overall health management effect through intelligent functions.

[0275] Example 2

[0276] An exemplary embodiment of the present invention is as follows Figure 3 As shown, Figure 3 : is a flow chart of a personalized health management method based on artificial intelligence according to an embodiment of the present invention, which is applied to the personalized health management system based on artificial intelligence in Example 1. The personalized health management method based on artificial intelligence provided in the embodiment of the present application includes:

[0277] Step S301, collecting health data of the user from at least two data sources;

[0278] Step S302, converting the format of health data from at least two data sources to obtain data in a target format, and building a health database based on the data in the target format;

[0279] Step S303, generating a medical knowledge graph corresponding to the user by associating the health database with the pre-stored medical knowledge;

[0280] Step S304, performing semantic annotation of events through the time sequence in the user's health data according to the medical knowledge graph and the health database to obtain the user's health timeline;

[0281] Step S305, extracting the user's key health indicator data based on the medical knowledge graph;

[0282] Step S306, analyzing the key health indicator data according to the medical knowledge graph and the health timeline to obtain a preliminary analysis result;

[0283] Step S307, generating a health summary report based on the medical knowledge graph, the health timeline and the preliminary analysis results through the retrieval enhancement generation technology based on the knowledge graph; generating at least one tag through the large language model based on the health summary report and the health timeline, and assigning the at least one tag to the health summary report to obtain a health summary report configured with at least one tag;

[0284] Step S308, analyzing the user's health status based on the health summary report and the health timeline; generating health management suggestions based on the medical knowledge graph, the health timeline, the health summary report and the health status configured with at least one tag through the retrieval enhancement generation technology based on the knowledge graph;

[0285] Step S309, displaying the user's health timeline, at least one tag, a health summary report configured with at least one tag and health management suggestions, as well as the medical party's diagnosis and treatment plan to the user and / or medical party through the interactive module;

[0286] Step S310: Send a health summary report and health management suggestions configured with at least one tag to the user and / or medical party.

[0287] The present invention adopts the above technical scheme, by realizing the automatic collection and transcription of health data; using a large model to automatically summarize, label and generate personalized suggestions for health data; constructing a complete health timeline to help doctors quickly and comprehensively understand the patient's health status; and through an online consultation platform, realizing collaborative information sharing between patients and doctors. Compared with the existing technology, it has the following technical effects: reducing the data management burden of patients, improving the diagnosis and treatment efficiency of doctors, and improving the overall health management effect through intelligent functions.

[0288] The above description is only a preferred embodiment of the present invention, and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A personalized health management system based on artificial intelligence, characterized in that: include: Data collection module, data sorting module, knowledge graph construction module, automated summary module, tag management module, health timeline module, personalized suggestion module, interaction module and reminder module, among which, The data acquisition module is used to obtain the user's health data through image acquisition records, audio acquisition records in the smart terminal, local records of the smart terminal, records of the online platform, and / or collection records of the wearable device according to the pre-granted permissions; The data sorting module is connected to the data acquisition module and is used to convert the user's health data into target format data and establish a health database based on the target format data; The knowledge graph construction module is connected to the data sorting module and is used to associate the health database with the pre-stored medical knowledge to generate a medical knowledge graph corresponding to the user; The health timeline module is connected to the knowledge graph construction module and is used to perform semantic annotation of events through the time sequence in the user's health data according to the medical knowledge graph and the health database to obtain the user's health timeline; The automated summary module is connected to the knowledge graph construction module and the health timeline module, respectively, and is used to extract the key health indicator data of the user based on the medical knowledge graph, analyze the key health indicator data based on the medical knowledge graph and the health timeline, and obtain a preliminary analysis result; and generate a health summary report based on the medical knowledge graph, the health timeline and the preliminary analysis result through a retrieval enhancement generation technology based on the knowledge graph; The tag management module is connected to the automatic summary module, and is used to generate at least one tag through a large language model according to the health summary report and the health timeline, and assign the at least one tag to the health summary report to obtain the health summary report configured with the at least one tag; The personalized suggestion module is connected to the tag management module, and is used to analyze the health status of the user according to the health summary report and the health timeline; generate health management suggestions according to the medical knowledge graph, the health timeline, the health summary report configured with the at least one tag, and the health status through the knowledge graph-based retrieval enhancement generation technology; The interaction module includes: a first user interaction module and a second user interaction module, wherein the first user interaction module is connected to the health timeline module and the personalized suggestion module, respectively, and is used to display the health timeline, the at least one tag, the health summary report configured with the at least one tag, and the interface of the health management suggestion to the user and provide an editing function; the second user interaction module is connected to the personalized suggestion module, and is used to display the health timeline, the at least one tag, the health summary report configured with the at least one tag, and the interface of the health management suggestion of the user to the medical party, and generate or adjust the diagnosis and treatment plan according to the health timeline, the at least one tag, the health summary report configured with the at least one tag, and the health management suggestion; The reminder module is connected to the interaction module, and is used to send the health summary report and the health management suggestions configured with the at least one tag to the user and / or the medical party; and to provide real-time reminders based on the acquired health data of the user.

2. The personalized health management system based on artificial intelligence according to claim 1 is characterized in that: The data acquisition module is also used to, in the process of acquiring the health data of the user through the image acquisition record in the smart terminal according to the pre-granted authority, automatically record a video of a first preset length of time for the medical record through a designated APP before the user takes a photo, and obtain at least three images that meet the preset standard by prompting the user to stay for a second preset length of time when shooting the medical record, use an operator to calculate the image sharpness value of each frame from the video of the first preset length of time, select at least two frames of images with the highest image sharpness values, and combine them with the at least three images that meet the preset standard to generate an image set, perform a preliminary scan of the image set through OCR, classify the documents in combination with the user's medical history information, assign the classified documents to the OCR model corresponding to the specific field for processing, extract image information corresponding to medical terms and numerical values, compare and analyze the image information with the image set, and input the weighted comparative analysis results into the large language model for fusion through confidence weighting to obtain the health data of the user; If the image sharpness value of each frame does not reach a preset threshold, the user is prompted to reshoot; In the process of assigning the classified documents to the OCR model corresponding to the specific field for processing, if there is a specific character whose character classification probability in the OCR recognition of all images is lower than the confidence threshold, and no reasonable result can be obtained in the medical context through the large language model, the specific character will be marked and the user will be prompted to check and confirm.

3. The personalized health management system based on artificial intelligence according to claim 2 is characterized in that: The data collection module is further used to collect the user's voice through the automatic speech recognition model in the process of obtaining the user's health data through the audio collection record in the smart terminal according to the pre-granted authority, and when the user's voice meets the preset threshold, classify it according to the medical field; when the user's voice does not meet the preset threshold, prompt the user to re-record; After classification according to the medical field, an automatic speech recognition model of the corresponding field is loaded for recognition to obtain medical terms, data corresponding to the medical terms and specific information of the user, and health data of the user is generated based on the medical terms, data corresponding to the medical terms and specific information of the user; The large language model is used to judge the rationality of the medical term, the data corresponding to the medical term and the user's specific information. If the semantics are incoherent or do not meet the medical rationale, the user is prompted to confirm.

4. The personalized health management system based on artificial intelligence according to claim 3 is characterized in that: The data acquisition module is also used for obtaining the health data of the user through the local records of the smart terminal and the records of the online platform in the smart terminal according to the pre-granted permissions. When obtaining the health data of the user in the local records of the smart terminal, the locally stored health records of the smart terminal selected by the user are obtained, and the health data of the user is obtained by inputting the large language model for fusion; when obtaining the health data of the user in the records of the online platform, the user is jumped to the online platform through platform authorization, the health records of the user are selected, and the health data of the user is obtained by inputting the large language model for fusion.

5. The personalized health management system based on artificial intelligence according to claim 4 is characterized in that: The data collection module is also used to collect the user's health data in real time during the process of obtaining the user's health data through the collection records of the smart terminal and / or the wearable device based on the pre-granted permissions, wherein the real-time collection of the user's health data includes at least: heart rate, number of steps, sleep and blood sugar.

6. The personalized health management system based on artificial intelligence according to claim 5 is characterized in that: The data sorting module is further used to divide the user's health data into a core data layer and an extended data layer; detect whether the data of the core data layer and the data of the extended data layer belong to the standard field range; if the data of the core data layer and the data of the extended data layer belong to the standard field range, perform structured storage to obtain the target format data; If the data in the core data layer and the data in the extended data layer do not belong to the standard field range, the data properties, importance and frequency of occurrence are analyzed by the large language model, and whether a new field needs to be created is evaluated based on the data properties, importance and frequency of occurrence, and recommendations are made, the analysis results are displayed to the user and confirmation is obtained, and a field or a supplementary record layer in the extended data layer is created based on the user's confirmation; wherein, the historical frequency of occurrence of similar data is counted by the large language model, the degree of association with existing medical knowledge is analyzed, and the clinical application value is evaluated; The corresponding classification of the created field is determined through the large language model, an association relationship between the classification and the field is established, the classification result is displayed to the user and a confirmation is obtained; medical terminology recognition and standardized mapping are performed on the data of the core data layer and the data of the extended data layer according to the large language model, and new terms are classified and associated analyzed, and the terminology knowledge graph of the large language model is dynamically updated.

7. The personalized health management system based on artificial intelligence according to claim 6, characterized in that: The data sorting module is further used to perform standard field verification and non-standard data verification on the data of the core data layer and the data of the extended data layer, wherein the standard field verification includes: format verification, value range verification, logic verification and timing verification; the non-standard data verification includes: rationality verification through the large language model, checking the consistency of the field, and performing data value assessment; The data sorting module is also used to set an incremental update function, wherein the incremental update function includes: single data update and dynamic field update, the single data update includes: supporting incremental update of single inspection data, only processing and storing newly uploaded data items and / or retaining historical data, and building a time series; the dynamic field update includes: continuously evaluating the use of newly created fields through the large language model, automatically adjusting the importance of fields based on usage frequency and value analysis, and / or maintaining and optimizing the association between fields; The data sorting module is also used to set up a processing feedback optimization function, wherein the processing feedback optimization function includes: using feedback of the first user interaction module and the second user interaction module of the mobile phone, analyzing the feedback through the large language model, automatically optimizing the field classification and association relationship corresponding to the feedback, and updating the data processing rules.

8. The personalized health management system based on artificial intelligence according to claim 7, characterized in that: The data sorting module is also used to analyze the health data of the user, store the health data according to the classification of the health data, and obtain the health database, wherein the storage according to the classification of the health data includes: if the health data includes structured data, the structured data is stored in a structured database; if the health data includes medical text, semantic vectors are extracted from the medical text and the semantic vectors are stored; if the health data includes medical information and user detection data, the health data is stored through a knowledge graph.

9. The personalized health management system based on artificial intelligence according to claim 8, characterized in that: The knowledge graph construction module is also used to clean the health data in the health database, and to associate the cleaned health data with pre-stored medical knowledge to construct a medical knowledge graph corresponding to the user, and to obtain the verified medical knowledge graph by performing data consistency verification on the constructed medical knowledge graph.

10. The personalized health management system based on artificial intelligence according to claim 1, characterized in that: The health timeline module is also used to establish a time association between all health data and corresponding nodes in the medical knowledge graph based on the time of event occurrence in the health database, semantically annotate key events through the large language model, extract core concepts and attributes of events, and link them with corresponding concepts in the medical knowledge graph to generate an interactive health timeline, wherein the health timeline includes: event nodes, which are associated with the medical knowledge graph and are used to indicate medical knowledge associated with the event nodes; when new health data is generated, the new health data is updated to the health timeline.

11. The personalized health management system based on artificial intelligence according to claim 1, characterized in that: The reminder module is further configured to send the health summary report and the health management suggestions configured with the at least one tag to the user in a specified manner according to a preset time period, and / or push the user's health dynamics to the medical party; The reminder module is also used to perform real-time detection and warning according to the health data of the user collected by the data collection module; The reminder module is further used to compare the health data collected by the data collection module with the health data collected by the wearable device, and to issue an early warning reminder according to the comparison result; When the data collection module is suspended or abnormal, a prompt message is sent to the user, and the prompt message is used to indicate that the data collection module is abnormal or interrupted in uploading, or the connection status of the wearable device is abnormal or interrupted; when the data collection module collects health data with health risks or the data upload interruption time is greater than the preset period, and the collection record cannot be obtained through the wearable device, an emergency notification is made through the preset contact; When the data collection module collects health data that poses a health risk or the data upload interruption time is greater than a preset period, and the collection record cannot be obtained through the wearable device, an early warning reminder is issued to the user and the medical party so that the user and the medical party can perform emergency operations based on the early warning reminder.

12. The personalized health management system based on artificial intelligence according to claim 1, characterized in that: The personalized suggestion module is also used to dynamically update the suggestions according to changes in the health status of the user.

13. The personalized health management system based on artificial intelligence according to claim 1, characterized in that: The personalized health management system based on artificial intelligence also includes: a data security module, wherein: The data security module is connected to the reminder module, and is used to keep the user's health data confidential through end-to-end data encryption and preset access control policies, and to process personal data related to the health data that is not related to treatment according to the user's control instructions to obtain the processed health data.

14. A personalized health management method based on artificial intelligence, characterized in that: Applied to personalized health management systems based on artificial intelligence, including: Collect health data of the user from at least two data sources; Convert the health data of the at least two data sources into a format to obtain data in a target format, and establish a health database based on the data in the target format; Generate a medical knowledge graph corresponding to the user by associating the health database with pre-stored medical knowledge; Performing semantic annotation of events through the time sequence in the user's health data according to the medical knowledge graph and the health database to obtain the user's health timeline; Extracting the user's key health indicator data based on the medical knowledge graph; Analyze the key health indicator data according to the medical knowledge graph and the health timeline to obtain preliminary analysis results; Generate a health summary report based on the medical knowledge graph, the health timeline and the preliminary analysis result through a retrieval enhancement generation technology based on a knowledge graph; and generate at least one tag through a large language model based on the health summary report and the health timeline, and assign the at least one tag to the health summary report to obtain the health summary report configured with the at least one tag; Analyze the health status of the user according to the health summary report and the health timeline; generate health management suggestions according to the medical knowledge graph, the health timeline, the health summary report configured with the at least one tag, and the health status through the knowledge graph-based retrieval enhancement generation technology; Displaying the user's health timeline, the at least one tag, the health summary report configured with the at least one tag, the health management suggestions, and the diagnosis and treatment plan of the medical party to the user and / or the medical party through an interactive module; The health summary report and the health management suggestions configured with the at least one tag are sent to the user and / or the medical party.

Images