Household health early warning method based on artificial intelligence and related equipment

Abstract

The invention provides a home health early warning method based on artificial intelligence and related equipment, and is applied to the technical field of medical data processing. The method comprises the following steps: preprocessing patient bracelet monitoring data, medical text data and environment data to generate a target physiological data sequence, medical text features, environment features and climate data features; processing the environment characteristics and the climate data characteristics to generate infectious disease monitoring characteristics; processing literatures in the knowledge base data based on a literature evidence intensity grading system to generate updated knowledge base data; processing the target physiological data sequence, the medical text features and the infectious disease monitoring features based on a target health early warning model to generate target health risk features; and processing the target health risk features based on the target health early warning model and the updated knowledge base data to generate patient health early warning information.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a home health warning method based on artificial intelligence and related devices. Background Art

[0002] In terms of health guidance, the targeted aging population and those with chronic diseases, especially the four major chronic diseases that require regular reexaminations or even daily monitoring, are characterized by being at high risk, having mobility difficulties but needing to frequently go to the hospital for screening or examinations, or needing to daily monitor their physical conditions or disease progress. For example, patients with hypertension, diabetes, COPD, asthma, and cancer. These patients are characterized by persistent diseases that need to be monitored and may lead to serious adverse consequences.

[0003] In the prior art, data processing and integration are difficult, various data formats are not unified, making it difficult to comprehensively analyze, and there are also conflicts in medical knowledge. The effects of disease warning and health management are not good. It is impossible to accurately warn of infectious disease risks using multi-source data, and personalized and real-time services cannot be provided for chronic disease patients. Moreover, medical knowledge is updated rapidly, but the existing systems cannot keep up in time, resulting in lagging health information. These problems seriously affect the home health management experience of patients, may delay the condition, increase medical costs, and also cause unreasonable allocation of medical resources, reducing the overall efficiency and quality of medical services. There is an urgent need for new technical solutions to solve these problems in order to improve the level of home health management.

[0004] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of the present disclosure, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of this application is to provide an artificial intelligence-based home health warning method and related devices, which at least overcome the problems existing in the prior art to a certain extent, collect patient bracelet monitoring data, medical text data, environmental data and knowledge base data, and perform preprocessing and feature extraction. For example, perform format conversion and key information extraction on medical text data, and screen meteorological and infectious disease information from environmental data. Process the knowledge base data through a literature evidence strength grading system to ensure the accuracy and timeliness of knowledge. In terms of model construction and application, use models such as SEIR to process infectious disease-related data and evaluate the transmission risk. At the same time, train an initial health warning model by obtaining other patients' data, and optimize it to obtain a target model for analyzing patients' health risks. For example, combine the patient's physiological abnormal indicators, medical symptoms and infectious disease risk levels to evaluate the health risk factors and degrees. Based on the target health warning model and the updated knowledge base, process the patient's health risk characteristics, generate a health risk correlation evaluation value, adjust the internal decision vector to obtain a deviation vector, and then convert it into intuitive health warning information to provide disease risk tips and prevention and treatment suggestions for patients, achieve efficient home health management, improve the patient's quality of life and save medical resources.

[0006] Other features and advantages of this application will become apparent from the following detailed description, or be learned in part through the practice of the present invention.

[0007] According to one aspect of this application, there is provided an artificial intelligence-based home health warning method, including: obtaining patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information; preprocessing the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features; processing the environmental features and climate data features to generate infectious disease monitoring features; processing the literature in the knowledge base data based on a literature evidence strength grading system to generate updated knowledge base data; processing the target physiological data sequence, medical text features, and infectious disease monitoring features based on a target health warning model to generate target health risk features; processing the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information.

[0008] Another aspect of the present application is an artificial intelligence-based home health warning device, which is characterized by including: an acquisition module for acquiring patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information; a processing module for preprocessing the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features; processing the environmental features and climate data features to generate infectious disease monitoring features; processing the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data; processing the target physiological data sequence, medical text features, and infectious disease monitoring features based on the target health warning model to generate target health risk features; and processing the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information.

[0009] According to still another aspect of the present application, an electronic device is characterized by including: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the artificial intelligence-based home health warning method described above by executing the executable instructions.

[0010] According to yet another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a second processor, the artificial intelligence-based home health warning method described above is implemented.

[0011] The artificial intelligence-based home health warning method and related devices provided by the present application collect patient bracelet monitoring data, medical text data, environmental data, and knowledge base data by a server, and perform preprocessing and feature extraction. For example, format conversion and key information extraction are performed on the medical text data, and meteorological and infectious disease information is screened from the environmental data. The knowledge base data is processed through the literature evidence strength grading system to ensure the accuracy and timeliness of the knowledge. In terms of model construction and application, models such as SEIR are used to process infectious disease-related data and evaluate the transmission risk. At the same time, the initial health warning model is trained by obtaining other patient data, and the target model is obtained through optimization for analyzing the health risks of patients. For example, by combining the patient's physiological abnormal indicators, medical symptoms, and infectious disease risk levels, the health risk factors and degrees are evaluated. Based on the target health warning model and the updated knowledge base, the patient health risk features are processed to generate a health risk correlation evaluation value, the internal decision vector is adjusted to obtain a deviation vector, and then it is converted into intuitive health warning information, providing disease risk prompts and prevention and treatment suggestions for patients, realizing efficient home health management, improving the quality of life of patients, and saving medical resources.

[0012] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and do not limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flowchart showing a method for home health warning based on artificial intelligence provided by an embodiment of the present application;

[0014] Figure 2 A schematic structural diagram showing a device for home health warning based on artificial intelligence provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for explaining and understanding the present invention, and are not used to limit the present invention.

[0016] The following is combined with Figure 1 to describe a method for home health warning based on artificial intelligence according to an exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown for the convenience of understanding the spirit and principle of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0017] In one embodiment, the present application also proposes a method for home health warning based on artificial intelligence and related devices. Figure 1 Schematically shown is a flowchart of a method for home health warning based on an embodiment of the present application. As Figure 1 shown, the method is applied to a server and includes:

[0018] S101, obtaining patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information.

[0019] In one implementation, the patient wears a smart bracelet daily. Through built-in sensors such as a heart rate sensor, an acceleration sensor, a sleep monitoring sensor, etc., the bracelet collects the patient's physiological data in real time. These data include the patient's heart rate change data throughout the day, such as resting heart rate, peak heart rate during exercise, etc.; step count data, recording the patient's daily activity steps and reflecting their daily exercise amount; sleep data, covering sleep onset time, number of awakenings, deep sleep and light sleep durations, etc. These data are stored locally on the bracelet in the form of a time series or transmitted via Bluetooth to a mobile application bound to it. The system regularly obtains these data through an interface docking with the mobile application, providing a basis for subsequent analysis of the patient's health status. During the patient's past medical treatment processes, a large amount of medical text materials have been accumulated. The hospital's electronic medical record system stores the patient's physical examination reports, including the test results of various physiological indicators such as height, weight, blood pressure, blood sugar, blood lipids, etc., as well as the doctor's analysis and diagnosis opinions on these indicators; there are also hospitalization records and discharge summaries, which detail information such as the patient's disease diagnosis, treatment process, medication use, and rehabilitation status at the time of discharge. The system obtains these medical text data through secure data interaction with the hospital's medical information system and performs subsequent processing and analysis.

[0020] Environmental data mainly comes from official channels such as professional meteorological monitoring agencies and disease prevention and control centers. The system obtains the real-time climate data of the patient's location and the areas to be visited subsequently through the data interface with the meteorological department, such as the daily maximum temperature, minimum temperature, average humidity, air quality index (AQI), etc. These data reflect the current climate conditions. At the same time, it obtains infectious disease monitoring information from the official database of the disease prevention and control center, including the incidence, epidemic trend, and spread range of various infectious diseases in the region recently. These environmental data are closely related to the patient's health status. For example, cold weather may induce blood pressure fluctuations in hypertensive patients, and the prevalence of infectious diseases may increase the patient's infection risk, thus affecting their health management strategy.

[0021] Knowledge base data covers medical knowledge resources in multiple aspects. It obtains the latest scientific research achievement literature from professional medical databases, which contains the latest research results in the medical field, such as the research progress of new hypertension treatment drugs, the association research between hypertension and other diseases, etc.; obtains the electronic versions of authoritative medical textbooks, which contain medical theories and diagnostic and treatment standards verified through long-term practice; and there are also various drug instructions, which detail information such as the drug's ingredients, indications, usage and dosage, adverse reactions, etc. In addition, it collects medical information, expert opinions, etc. published on the official websites of various medical institutions. The system integrates these knowledge base data from different sources to build a comprehensive medical knowledge base, providing knowledge support for subsequent health warnings and treatment suggestions.

[0022] Medical knowledge is constantly evolving and updating. To ensure the accuracy and timeliness of the health warnings and diagnosis and treatment suggestions provided by the system, it is necessary to obtain information on the medical knowledge update mechanism. The system subscribes to the push services of professional medical journals to keep abreast of the latest research findings and updates to clinical guidelines in a timely manner; it pays attention to the dynamics of medical academic conferences to obtain the cutting-edge research and expert consensus released at the conferences. At the same time, it uses automated literature monitoring tools to regularly scan major medical databases, screen and evaluate newly published literature, and determine its credibility and influence according to the literature evidence strength grading system. In addition, communication channels with medical expert teams have been established, and the expert teams regularly review and update the data in the knowledge base of the system to ensure the accuracy and practicality of the knowledge. Through these methods, the system obtains information on the medical knowledge update mechanism, can update the knowledge base data in a timely manner, and provide better health services for patients.

[0023] S102, preprocess the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features.

[0024] In one implementation, data cleaning is performed on the patient bracelet monitoring data to remove outliers and missing values, generating the preliminarily processed bracelet physiological data. The patient wears a smart bracelet daily, and the bracelet continuously collects their physiological data. On a certain day, there was a record of a heart rate of 30 beats per minute (far below the normal range) in the data recorded by the bracelet, and there was also a situation where the step count data was missing for some time periods. The system performs data cleaning on this data, determines the outlier such as a heart rate of 30 beats per minute as invalid data, and fills in the missing step count data through comparison with the heart rate data in the previous and subsequent time periods and reasonable estimation, thereby generating the preliminarily processed bracelet physiological data.

[0025] Feature extraction is performed on the preliminarily processed bracelet physiological data to generate a target physiological data sequence. Among them, the target physiological data sequence includes a heart rate feature sequence, a step count feature sequence, and a sleep duration feature sequence. After completing the data cleaning, the system performs feature extraction on the preliminarily processed bracelet physiological data. From the heart rate data, the heart rate values per minute are extracted in chronological order to form a heart rate feature sequence, such as [75, 78, 80,...], which can reflect the changes in the patient's heart rate throughout the day. Similarly, from the step count data, the step counts are statistically calculated per hour or half-day to generate a step count feature sequence, such as [500, 800, 1200,...], which is used to evaluate the patient's daily activity level. For sleep data, the system extracts information such as the sleep start time, number of awakenings, deep sleep and light sleep durations, and constructs a sleep duration feature sequence, like [22:30, 2, 180, 240], where 22:30 represents the sleep start time, 2 represents the number of awakenings, and 180 and 240 respectively represent the deep sleep and light sleep durations (in minutes).

[0026] Perform format conversion processing on medical text data to generate medical text in the target format, and process the medical text in the target format to generate medical text features. Among them, the medical text features include symptom judgment features and symptom intervention features. After a patient visits the hospital, the hospital's electronic medical record system stores medical text data such as their physical examination reports, hospitalization records, and discharge summaries. The initial formats of these data may vary, for example, the physical examination report is in PDF format and the hospitalization record is in Word format. The system performs format conversion processing on these medical text data, uniformly converting all data into XML or JSON format that is convenient for computer processing to generate medical text in the target format. For the medical text after format conversion, the system further processes it to generate medical text features. In the physical examination report, the system uses natural language processing technology to identify descriptions such as "high blood pressure" and "unstable blood sugar", and uses them as symptom judgment features. At the same time, the suggestions given by the doctor in the report are extracted, such as "need to take antihypertensive drugs on time" and "regularly monitor blood sugar and adjust diet", which are used as symptom intervention features to provide a basis for subsequent health assessment and early warning.

[0027] Perform screening and feature extraction processing on meteorological data in environmental data to generate climate data features. Among them, the climate data features include temperature features, humidity features, and air quality features. The system obtains the meteorological data of the patient's location from the official website or data interface of the meteorological department. The data contains information such as the daily maximum temperature, minimum temperature, average humidity, and air quality index (AQI). When processing these meteorological data, the system performs screening and feature extraction processing, taking the daily average temperature as the temperature feature. For example, the temperature feature for a certain week is [25, 26, 24, 23, 22, 21, 20] (unit: °C). From the humidity data, the daily average humidity is extracted as the humidity feature, such as [60%, 55%, 62%, 58%, 65%, 70%, 72%]. For the air quality index, it is directly used as the air quality feature, for example, [50, 45, 48, 52, 55, 60, 65]. These features can reflect the possible impact of the local climate environment on the patient's health.

[0028] Screen and extract features from non-meteorological data in environmental data to generate environmental features, where the environmental features include infectious disease information for different regions and times. The system obtains the non-meteorological data in environmental data, that is, infectious disease monitoring information, from the database of the Center for Disease Control and Prevention. These information include the incidence and epidemic trends of infectious diseases in different regions at different times. The system screens and extracts features from these data. For example, when influenza is prevalent in the region where the patient is located, the system extracts information such as the number of influenza cases, the transmission speed, and the infection situation in surrounding areas in this region as the infectious disease information in the environmental features. These information can be used to evaluate the potential threat of infectious diseases to the patient's health, and combined with the patient's own health status, provide a more comprehensive basis for health warnings.

[0029] S103. Process the environmental features and climate data features to generate infectious disease monitoring features.

[0030] In one implementation, extract and screen the basic infectious disease information in the environmental features and the meteorological impact correlation information in the climate data features to generate potential risk factors for infectious disease transmission, meteorological factor correlation data, and basic information on the regional epidemic transmission trend. Assume that the patient lives in a northern city. During the high-incidence season of influenza in winter, the system obtains the environmental features and climate data features of the region where the patient is located. In the environmental features, the basic infectious disease information shows that the number of recently confirmed influenza cases in this region has been gradually increasing, the proportion of influenza patients visiting the hospital outpatient department has increased, and the influenza transmission speed in surrounding cities is relatively fast. In the climate data features, the meteorological impact correlation information indicates that the temperature has dropped suddenly recently, the average temperature has dropped from 10°C to about 0°C, and the air humidity has also decreased, with the average humidity dropping from 50% to about 30%. The system extracts and screens these information. According to the characteristics that the influenza virus is more likely to spread in a low-temperature and dry environment, information such as the sudden drop in temperature, the decrease in humidity, and the influenza transmission situation in surrounding cities is integrated to generate potential risk factors for infectious disease transmission. At the same time, meteorological data such as temperature and humidity are recorded as meteorological factor correlation data. Combining the increase in the proportion of influenza patients visiting the hospital outpatient department in this region and the upward trend of the number of confirmed cases, basic information on the regional epidemic transmission trend is generated, providing a basis for subsequent evaluation of the influenza transmission risk.

[0031] Based on the infectious disease transmission model library, the potential risk factors of infectious disease transmission, the associated data of meteorological factors, and the basic information of the regional epidemic transmission trend are processed to generate the infectious disease transmission risk level information and the transmission trend change information. The system calls the infectious disease transmission model library, which contains the transmission models of various infectious diseases under different environmental conditions. Taking the SEIR model as an example, the SEIR model belongs to the deterministic compartmental model. It divides the population into different categories according to their health status and describes the dynamic changes between these categories through mathematical equations to simulate the transmission process of infectious diseases in the population. This model mainly includes four compartments, namely Susceptible (S), Exposed (E), Infectious (I), and Recovered (R). Susceptible (S): Refers to the population that has not been infected with the virus but is likely to be infected. For this 65-year-old elderly person with underlying diseases, due to their relatively weak immunity, they belong to the high-risk individuals in the susceptible population. Exposed (E): Those who have been infected with the virus but are in the incubation period and do not have or have relatively weak infectivity. Infectious (I): The population in the infected state who can transmit the virus to the susceptible. Recovered (R): Those who have been infected with the virus, have now recovered and acquired a certain degree of immunity, and generally will not be infected again. Infection rate (β): Represents the probability of a susceptible being infected after contacting an infectious person. It is affected by various factors, such as contact frequency, protective measures, etc. In this city, since people mainly stay indoors in winter, the contact between people is closer, and the recent implementation of protective measures is not in place, so the infection rate is relatively high. The rate of conversion of exposed to infectious (σ): That is, the proportion of exposed individuals converting to infectious individuals per unit time. Different infectious diseases have different incubation periods, and this parameter will also vary. For influenza, the incubation period is usually short, and the σ value is relatively large. Recovery rate (γ): Refers to the proportion of infectious individuals recovering per unit time. For healthy people and those with underlying diseases, the recovery rate may be different. For example, for this 65-year-old elderly person with hypertension and diabetes, if infected with influenza, their recovery rate may be lower than that of healthy people because the underlying diseases will affect the body's recovery ability.

[0032] The system calls the SEIR model in the infectious disease transmission model library to analyze the influenza transmission situation in this city. Previously, the system has collected and generated basic information on potential risk factors for infectious disease transmission, associated data on meteorological factors, and the regional epidemic transmission trend. Potential risk factors for infectious disease transmission: The population in this city is dense, there is a large flow of people in public places, and some residents have weak awareness of protection and do not wear masks. At the same time, there are a large number of crowded places such as nursing homes and hospitals in the city. This 65-year-old elderly person lives in a nursing home, increasing the risk of infection. Associated data on meteorological factors: Recently, the temperature in this city has dropped sharply, with the average temperature dropping from 10°C to 2°C, and the air humidity has also decreased significantly, with the average humidity dropping from 60% to 30%. The low-temperature and dry environment is conducive to the survival and transmission of influenza viruses. The survival time of the virus on the surface of objects is extended, and the transmission efficiency is improved. Basic information on the regional epidemic transmission trend: The influenza epidemic in the surrounding cities shows a rapid upward trend, with the number of confirmed cases increasing at a rate of 20%-30% per week. Moreover, there are frequent exchanges of people between this city and the surrounding cities, with a large number of business trips, tourism and other activities, increasing the risk of virus importation. Currently, 200 influenza cases have been confirmed in this city, and the number of cases is on the rise.

[0033] The SEIR model integrates and analyzes the above input data. Considering the impact of meteorological conditions on the survival and transmission of the virus, the model adjusts the infection rate β. In a low-temperature and dry environment, the infection rate β is increased from the original 0.1 to 0.15. At the same time, according to the population structure and medical resource situation of this city, the rate σ at which latent individuals turn into infected individuals is determined to be 0.2, and the recovery rate γ is 0.08 (considering that the recovery rate is relatively low for people with underlying diseases such as 65-year-old elderly people). The model also takes into account the current regional epidemic basis, that is, the 200 confirmed cases, and uses it as the initial number of infected individuals I(0). At the same time, according to the epidemic transmission law and population contact situation, the number of latent individuals E(0) is estimated to be 300, the number of susceptible individuals S(0) is 499,500, and the number of recovered individuals R(0) is 0 (assuming no previous recovered cases). For the transmission situation in the surrounding cities, the model reflects its impact by introducing a population flow parameter. It is assumed that 1% of the population flows between this city and the surrounding cities every week, and the infection rate among the flowing population is the same as that in the surrounding cities. By comprehensively analyzing these factors, the model performs calculations.

[0034] After the operation and evaluation of the model, it is determined that the influenza transmission risk level in this area is "high". This is because under the current meteorological conditions, population mobility and epidemic situation, the number of susceptible people is large, the infection rate is relatively high, the virus spreads relatively fast, and there are a large number of potential transmission sources (latent carriers). Therefore, the overall risk is at a high level. For this 65-year-old elderly person with underlying diseases, he / she faces a relatively high infection risk. The model predicts that if the meteorological conditions continue in the next week, the number of influenza confirmed cases may increase at a rate of 10%-15%. Calculated based on the current 200 confirmed cases, the number of confirmed cases may increase to 220-230 after one week. At the same time, due to the characteristics of personnel mobility and virus transmission, the transmission range may further expand to the surrounding areas where large-scale infections have not yet occurred. For example, new cases may emerge successively in some surrounding communities and schools, and the epidemic may gradually spread in these areas. The urban health department can take prevention and control measures in a timely manner according to these prediction results, such as strengthening the protection of key places such as nursing homes and schools, raising the protection awareness of residents, and allocating medical resources, etc., to reduce the influenza transmission risk and protect the health of susceptible people including this 65-year-old elderly person.

[0035] Based on the infectious disease transmission risk level information and the information on the change of transmission trend, classify and identify the relevant data in the environmental characteristics and climate data characteristics to generate the screening result of risk-associated data. Classify the recent meteorological data such as temperature, humidity, air quality, etc. according to the degree of influence on influenza transmission. For example, mark the low temperature and low humidity data as high-risk influencing factors; identify the infectious disease-related data such as the proportion of influenza patients visiting outpatient departments of hospitals and the number of confirmed cases according to the growth rate and severity. Mark the time periods with a fast growth rate and a large number of confirmed cases as high-risk time periods. After classification and identification, the system screens out the data related to high risks to generate the screening result of risk-associated data. For example, screen out the meteorological data with an average temperature below 5°C and an average humidity below 40% in the past week, and the infectious disease data with a daily increase in the number of influenza confirmed cases exceeding 50. The combination of these data constitutes the screening result of risk-associated data, which is convenient for subsequent centralized analysis.

[0036] Summarize and analyze the screening results of risk-associated data to generate infectious disease surveillance characteristics, where the infectious disease surveillance characteristics are used to characterize the transmission risks and trends of infectious diseases under different environmental and climatic conditions. The system summarizes and analyzes the screening results of risk-associated data. Statistically analyze information such as the frequency of high-risk meteorological data occurrences, the duration and growth rate of high-risk infectious disease data, etc. Through analysis, it is found that within the past two weeks, the frequency of high-risk meteorological conditions (low temperature, low humidity) has reached 70%, the number of confirmed influenza cases has been continuously increasing, and the daily growth rate is between 30 - 80 people. Combining this information, the system generates infectious disease surveillance characteristics. These characteristics indicate that under the current environmental and climatic conditions, the influenza transmission risk in this region is at a relatively high level, and the transmission trend shows a rapid upward trend. For this 65-year-old patient with hypertension and diabetes, due to their relatively low immunity, the risk of developing severe complications after being infected with influenza is relatively high. The infectious disease surveillance characteristics generated by the system will be incorporated as an important basis into the patient's health warning system to promptly remind the patient to take protective measures, such as minimizing going out, wearing masks, keeping warm, etc., and at the same time also provide support for subsequent provision of targeted health management advice for the patient.

[0037] S104, Process the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data.

[0038] In one implementation, extract and classify the literature information in the knowledge base data based on the source and type to generate scientific research achievement literature data, authoritative textbook literature data, clinical review literature data, and drug instruction literature data. The system obtains a large amount of knowledge base data, which contains rich medical knowledge. A large number of literatures are obtained from channels such as professional medical databases and academic journal websites. When processing this literature information, the system extracts and classifies it according to the source and type. Extract the latest research results on hypertension and diabetes from the latest medical research journals, such as clinical trial reports of new antihypertensive drugs, research on the pathogenesis of diabetic complications, etc., which constitute the scientific research achievement literature data. Extract relevant content from professional textbooks such as "Internal Medicine" and "Practical Diabetes" published by authoritative medical publishers to form authoritative textbook literature data, which contain knowledge such as disease diagnosis criteria and treatment principles that have been verified over a long period. Search for clinical review articles on hypertension and diabetes on major medical websites and professional databases, such as articles comprehensively analyzing the treatment plans for patients with hypertension and diabetes, to form the clinical review literature data. Collect the instructions of various drugs for treating hypertension and diabetes, and extract information such as the drug ingredients, indications, usage and dosage, and adverse reactions as the drug instruction literature data.

[0039] Process the literature data of scientific research achievements, authoritative textbook literature data, clinical review literature data, and drug instruction literature data based on the literature evidence strength grading system to generate the evidence strength evaluation values of scientific research achievements, authoritative textbooks, clinical reviews, and drug instructions. The system uses the literature evidence strength grading system to process the classified literature data above. For example, for the literature data of scientific research achievements, if the research is a multi-center, large-sample randomized controlled trial, according to the grading system, its evidence strength is relatively high; if it is a small-scale observational study, the evidence strength is relatively low. After evaluation, an evidence strength evaluation value is assigned to each scientific research achievement literature. For the authoritative textbook literature data, because it is a summary of long-term practice and research in the medical field, it usually has a relatively high evidence strength, but the content of different versions and chapters will also vary, and a detailed evaluation is also carried out to give the corresponding authoritative textbook evidence strength evaluation value. For the clinical review literature data, it will be evaluated according to factors such as the quality of the cited literature, the comprehensiveness and scientificity of the review, and the clinical review evidence strength evaluation value is generated. For the drug instruction literature data, the drug instruction evidence strength evaluation value is determined based on the standardization degree of drug research and development, the sufficiency of clinical trials, etc. For example, for drugs that have been verified through a large number of clinical trials and are widely recognized, the evidence strength evaluation value of their instructions is relatively high.

[0040] Based on the evidence strength evaluation values of scientific research achievements, authoritative textbooks, clinical reviews, and drug instructions, mark and screen various types of literature data to generate high-evidence-strength literature screening results, medium-evidence-strength literature screening results, and low-evidence-strength literature screening results. According to the various types of literature evidence strength evaluation values obtained above, the system marks and screens various types of literature data. Mark the scientific research achievement literature with evidence strength evaluation values in the top 30% as high-evidence-strength. These literatures are often the latest and most reliable research results; those in the middle 40% are marked as medium-evidence-strength; and the last 30% are marked as low-evidence-strength. The same method is applied to the authoritative textbook literature data, clinical review literature data, and drug instruction literature data. After screening, high-evidence-strength literature screening results are obtained, such as high-quality scientific research achievement literatures on the treatment of diabetes with novel insulin analogs, the core chapters on the treatment of hypertensive emergencies in authoritative textbooks, etc.; medium-evidence-strength literature screening results, such as some medium-scale hypertensive epidemiology research literatures, analysis articles on the current status of diabetes treatment in specific regions in clinical reviews, etc.; and low-evidence-strength literature screening results, such as some early small-scale diabetes drug efficacy observation literatures, auxiliary information not fully verified in drug instructions, etc.

[0041] Based on the multi-source knowledge conflict resolution information, knowledge conflict detection and resolution processes are respectively carried out on the screening results of high, medium, and low evidence strength literatures to generate high-confidence literature information. An initial evidence weight is set for each scientific research achievement literature, and the determination of the weight comprehensively considers multiple factors, such as the authority of the research institution, the size of the research sample, the scientificity of the research method, etc. For large-scale randomized controlled trials conducted in top medical research institutions, a relatively high initial weight is given; while for research with a small sample size and relatively simple research methods, the initial weight is relatively low. Taking the research on a new antihypertensive drug as an example, for a multi-center, large-sample (sample size exceeding 1000 cases) randomized controlled trial conducted at a globally renowned cardiovascular research center, the initial weight can be set at 0.8; while for an observational study with a sample size of only 100 cases conducted by a small regional research institution, the initial weight may be set at 0.3. The system analyzes the methodologies of different studies in detail to determine whether they meet the standard specifications of medical research. For studies using the standard randomized controlled trial method, a higher credibility score is given; for the evaluation of research methodologies, a credibility scoring system is set. Assuming a full score of 10 points, if a study strictly follows the internationally recognized randomized controlled trial process, conducts random grouping, double-blind treatment, and sets up a reasonable control group, the credibility score is recorded as (C = 8); if there is a defect in a key link such as random grouping, control setting, and blinding implementation in the study, according to the severity of the defect, points are deducted in a certain proportion. For example, 2 points are deducted for each defect, then the credibility score C = 8 - 2×n (n is the number of defects).

[0042] Compare the sample characteristics of different studies, including the age, gender, severity of diabetes and cardiovascular diseases, and comorbidity status of patients, etc. For studies with a high degree of match with the target patient population (such as patients aged 65 with hypertension and diabetes), a higher weight is assigned. Suppose the sample of Study A is mainly patients aged 60 - 70 with type 2 diabetes and mild cardiovascular diseases, which highly matches the characteristics of the target patient population, and its weight can be increased by 20% on the original basis; the sample of Study B is young patients with type 1 diabetes, which has a large difference from the target population, and the weight can be reduced by 30%. Consider the differences in the experimental environments where the studies are carried out, such as factors like region, medical conditions, and living habits that affect the research results. If a study is conducted in an environment similar to the climate and lifestyle of the region where the target patients are located, the reliability of its conclusion is higher. For example, the target patients live in a cold northern region with a slow pace of life and a greasy diet. If there is a study carried out in a similar environment and the living habits of the research subjects are similar, then the conclusion of this study will be given more consideration when dealing with conflicts; conversely, for studies carried out in regions with a hot climate and large lifestyle differences, the weight of their conclusions is appropriately reduced.

[0043] Introduce the knowledge and experience of medical field experts, and invite authoritative experts in related fields such as cardiology and endocrinology to evaluate the conflicting research conclusions. Based on their clinical experience and in-depth understanding of the field, the experts score the reliability of different research conclusions. The scoring results of the experts are comprehensively calculated with the results obtained above. For example, if an expert's reliability score for a certain research conclusion is 8 points (out of 10), combined with the weight of this research calculated previously, the comprehensive score of this research in conflict resolution is finally determined, which is used as an important basis for retaining or comprehensively analyzing the research conclusion.

[0044] Comprehensively analyze the results obtained above. If a certain research conclusion performs excellently in terms of evidence weight, research method consistency, sample characteristic matching, experimental environment similarity, etc., and is highly recognized by experts, then this conclusion will be preferentially retained. If different research conclusions have their own advantages and disadvantages, then these conclusions will be comprehensively analyzed. For example, for different conclusions about the cardiovascular protection effect of a new antihypertensive drug on diabetic patients, after comprehensively considering various factors, it may be concluded that "in patients of a specific age group (60 - 70 years old), with a specific type of diabetes (type 2) and accompanied by mild cardiovascular diseases, in areas with a cold climate and a greasy lifestyle, the new antihypertensive drug has a certain cardiovascular protection effect, but the effect may be affected by individual differences", so as to generate high-confidence literature information and provide a reliable knowledge basis for subsequent health warnings and diagnosis and treatment suggestions.

[0045] Obtain climate data and time information of different regions. The system obtains the climate data of the patient's location through the data interface with the meteorological department. For example, in winter, the average temperature in this area may be between 0°C and 5°C, the air humidity is about 40% - 50%, and the wind force is small. At the same time, obtain time information, such as the current is the high-incidence season of influenza and it is approaching the end of the year, so the indoor activities of residents increase and social gatherings are frequent. Based on the climate data and time information of different regions, weighted processing is performed on the high-confidence literature information to generate updated knowledge base data. Considering the impact of climate and time factors in different regions on diseases, the system performs weighted processing on the high-confidence literature information based on the obtained climate data and time information. For patients with hypertension and diabetes, a cold climate may lead to blood pressure fluctuations and increased difficulty in blood sugar control, and during the high-incidence season of influenza, the risk of infection increases, which will further affect the condition.

[0046] Therefore, the weight of high-confidence literature information related to the management of hypertension and diabetes, as well as the prevention and treatment of influenza in cold climates, will be correspondingly increased. For example, literature on strategies for coping with blood pressure fluctuations in winter hypertensive patients, and literature on how diabetic patients can prevent infections and control blood sugar levels during infectious disease epidemics, will receive higher weights in the knowledge base. Through this weighting process, updated knowledge base data is generated, enabling the knowledge base to more accurately provide health warnings and diagnosis and treatment recommendations tailored to the actual situation of patients. When the system conducts a health assessment and warning for this 65-year-old patient, the updated knowledge base data can more effectively support the system in giving targeted suggestions, such as reminding the patient to keep warm in winter, strengthen blood sugar and blood pressure monitoring, and minimize going out and take protective measures during the high-incidence period of influenza.

[0047] S105, process the target physiological data sequence, medical text features, and infectious disease monitoring features based on the target health warning model to generate target health risk features.

[0048] In one implementation, the medical and health data platform obtains various data and relevant information of other patients through data interaction with various cooperative medical institutions, smart wearable device manufacturers, meteorological departments, and professional medical databases. A large amount of bracelet monitoring data of other patients has been collected. For example, there is bracelet data of 1000 patients, which includes information such as their heart rate, number of steps, and sleep duration. These data detailedly record the daily physiological activities of patients in the form of time series. For example, for a certain patient in the past week, the average heart rate per day was between 70 and 90 beats per minute, the average number of steps was between 3000 and 5000 steps, and the sleep duration was about 6 to 8 hours. Medical text data of these patients has been obtained from the electronic medical record system of the cooperative hospital, including physical examination reports, hospitalization records, and discharge summaries, etc. These texts record information such as the patient's medical history, diagnosis results, and treatment process. For example, the physical examination reports of some patients show that they have hypertension, diabetes, and related complications to varying degrees; the hospitalization records record the changes in the patient's condition and treatment measures during hospitalization.

[0049] Climate data of the patient's location was obtained from the meteorological department, such as daily temperature, humidity, air quality index in the past year, etc.; at the same time, infectious disease surveillance information was obtained from the Center for Disease Control and Prevention, including the incidence and epidemic trends of infectious diseases such as influenza and pneumonia. For example, in winter, the temperature in this area is relatively low, with an average temperature of about 5°C, a humidity of 40%, and the incidence of influenza has increased. The platform has obtained updated knowledge base data according to the previous data processing process for the knowledge base. These data include the latest medical research results, key content of authoritative textbooks, drug instruction information, etc. For example, information such as the latest treatment guidelines for hypertension and diabetes, the mechanism of action and side effects of new drugs are included. The platform obtains medical knowledge update mechanism information by subscribing to professional medical journals, participating in medical academic conferences, and maintaining communication with medical expert teams. For example, recently, new research has shown that a certain new antihypertensive drug has better cardiovascular protection for patients with combined diabetes, and this information has been timely incorporated into the medical knowledge update mechanism information.

[0050] Based on a preset division ratio, other patients' bracelet monitoring data, medical text data, and environmental data are processed to generate training data and validation data. Among them, other patients' bracelet monitoring data, medical text data, and environmental data contain several pieces of information related to health conditions and the external environment. The platform processes the obtained other patients' bracelet monitoring data, medical text data, and environmental data based on a preset division ratio (assumed to be 70% training data and 30% validation data). From the bracelet monitoring data of 1000 patients, 700 patients' data are randomly selected as training data for training the initial health warning model; the remaining 300 patients' data are used as validation data to evaluate the performance of the model. The medical text data and environmental data are also divided according to the same ratio. For example, 70% of the medical text data such as physical examination reports, hospitalization records, and discharge summaries, as well as 70% of the climate data and infectious disease surveillance information, are used as training data; the other 30% are used as validation data. This division ensures that both the training data and the validation data contain information related to different health conditions and the external environment and are representative to a certain extent.

[0051] The initial health warning model is processed based on training data, updated knowledge base data, and medical knowledge update mechanism information to generate a warning model prediction parameter vector. Among them, the warning model prediction parameter vector includes prediction strategies and related parameter information for health risk assessment. The initial health warning model is a machine learning-based model, such as a neural network model. The model learns the relationships between patients' health conditions and bracelet monitoring data, medical text data, and environmental data from the training data. For example, by analyzing the bracelet data of patients with hypertension and diabetes in the training data, it is found that an abnormal increase in heart rate and a sudden decrease in the number of steps may be related to the deterioration of the condition; learn the diagnostic basis and treatment methods of different complications from medical text data; and combine environmental data to understand the impact of cold weather and the prevalence of infectious diseases on patients' health.

[0052] At the same time, the model refers to the updated knowledge base data and medical knowledge update mechanism information to obtain the latest medical knowledge and research results to optimize its prediction strategy. For example, according to the research results of new antihypertensive drugs, adjust the assessment method of the health risks of patients with hypertension. Finally, a warning model prediction parameter vector is generated, which contains prediction strategies for health risk assessment, such as rules for judging patients' health risks based on different data characteristics, and related parameter information, such as weights corresponding to different data characteristics. Based on the warning model prediction parameter vector, the initial health warning model is optimized and trained to generate a trained health warning model. During the training process, the model continuously adjusts its parameters to improve the fitting degree to the training data and the prediction accuracy. For example, through the backpropagation algorithm, continuously adjust the connection weights between neurons in each layer of the neural network model, so that the model can more accurately predict health risks based on the input patient data. After multiple iterative trainings, the model gradually adapts to the characteristics of the training data, and the health risk assessment of patients in different health conditions and external environments is more accurate, thus generating a trained health warning model.

[0053] The trained health warning model is subjected to simulated health warning processing based on the validation data to generate a validation result. Based on the validation result, the trained health warning model is evaluated and adjusted to generate a target health warning model. The validation data of 300 patients is input into the trained model, and the model evaluates and warns the health risks of these patients according to the learned knowledge and prediction strategies. For example, for a patient with hypertension and diabetes in the validation data, the model predicts that the patient has a high risk of infection and deterioration of the condition based on the abnormal heart rate monitored by the bracelet, poor blood sugar control in the medical text data, and the prevalence of influenza in the environmental data, and generates corresponding warning information. These warning information are the validation results, which are used to evaluate the performance of the model. The trained health warning model is evaluated and adjusted based on the validation result. The evaluation metrics can include accuracy, recall rate, F1 value, etc.

[0054] If the model has a low accuracy in predicting the deterioration of the conditions of patients with hypertension and diabetes during the validation process, it may be that the weight settings of some key data features in the model are unreasonable, or the model has not fully learned relevant medical knowledge. At this time, the model needs to be adjusted, such as re-adjusting the weights in the parameter vector, or further introducing more medical knowledge and data for training. After multiple evaluations and adjustments, the model reaches better performance indicators on the validation data, and finally generates the target health warning model. This target health warning model will be used to accurately warn patients (such as 65-year-old patients with hypertension and diabetes), and provide timely health management suggestions for patients, such as reminding patients to strengthen protection during the influenza epidemic period, adjusting drug doses according to physical activity conditions, etc.

[0055] In another implementation, target data extraction and classification processing are performed on the target physiological data sequence, medical text features, and infectious disease monitoring features to generate physiological abnormality index data, medical symptom target feature data, and infectious disease risk level data. The system obtains the target physiological data sequence, medical text features, and infectious disease monitoring features of the patient. From the target physiological data sequence, physiological abnormality index data such as abnormally elevated heart rate (e.g., resting heart rate continuously higher than 90 beats per minute) and sudden decrease in steps (daily steps less than 1000 steps) are extracted; from the medical text features, medical symptom target feature data such as "poor blood pressure control" and "large blood sugar fluctuations" are identified; according to the infectious disease monitoring features, if influenza is in a high-incidence period in the area where the patient is located, and the influenza transmission risk level in this area is evaluated as "high", this is used as the infectious disease risk level data.

[0056] Based on the target health warning model, the physiological abnormality index data, medical symptom target feature data, and infectious disease risk level data are analyzed and processed to generate health risk factor importance assessment information and health risk degree assessment information. The target health warning model receives the above-mentioned extracted and classified data and analyzes the association between these data and health risks. For example, it is found that the patient has an abnormally elevated heart rate and poor blood pressure control. Combining the knowledge about hypertension and cardiovascular diseases in the updated knowledge base, it is judged that these two factors have a greater impact on the patient's health risk; at the same time, considering that influenza is in a high-incidence period and the patient's own immunity is weak due to underlying diseases, the risk of influenza infection also becomes an important health risk factor. Thus, health risk factor importance assessment information is generated, and it is determined that abnormally elevated heart rate, poor blood pressure control, and influenza infection risk are important risk factors. Through the comprehensive analysis of these factors and the model's learning of the disease development law, it is evaluated that the patient's current health risk degree is relatively high, and health risk degree assessment information is generated.

[0057] Based on the importance assessment information of health risk factors and the assessment information of the degree of health risk, the target data in the target physiological data sequence, medical text features, and infectious disease monitoring features are screened and associated to generate the screening result of health risk target data. Data closely related to important risk factors are screened out. For example, continuously abnormal heart rate data and blood pressure fluctuation data are associated with the risk of hypertension exacerbation; data on poor blood glucose control are associated with the risk of diabetic complications; infectious disease monitoring data in high-incidence areas of influenza are associated with the patient's infection risk. These associated data together constitute the screening result of health risk target data.

[0058] Integrate and calculate the screening result of health risk target data, and process it in combination with the weight information in the target health warning model and the updated knowledge base to generate target health risk features. Among them, the target health risk features are used to characterize the degree of risk and the main risk factors faced by the patient's current health status. Combine the weight settings of different risk factors in the target health warning model (for example, the weight of risk factors related to hypertension is relatively high because hypertension poses a greater threat to the health of this patient), and the weight information on disease severity and risk association in the updated knowledge base for comprehensive calculation. Finally, generate target health risk features, indicating that the patient currently faces a relatively high health risk, and the main risk factors are poor control of hypertension, blood glucose fluctuations, and the risk of influenza infection. These factors may lead to consequences such as the onset of cardiovascular diseases, the aggravation of diabetic complications, and the development of severe illnesses caused by influenza infection, providing a basis for generating accurate health warning information and providing targeted health management suggestions in the follow-up.

[0059] S106, process the target health risk features based on the target health warning model and the data in the updated knowledge base to generate the patient's health warning information.

[0060] In one implementation, process the target health risk features based on the target health warning model and the data in the updated knowledge base to generate a health risk association assessment value. The system obtains the target health risk features of this patient, such as poor control of hypertension, blood glucose fluctuations, and a relatively high risk of influenza infection. The target health warning model combines the data in the updated knowledge base to comprehensively analyze these risk features. The knowledge base contains the association information between various disease risks and health status, as well as the quantitative data on the impact of different risk factors on overall health. Through calculation, the model determines the interaction relationship between these risk factors. For example, hypertension and blood glucose fluctuations interact with each other, increasing the risk of cardiovascular disease onset; the risk of influenza infection will further aggravate the underlying diseases in the case of relatively low immunity of the patient. Based on these analyses, the model generates a comprehensive health risk association assessment value. Assuming this assessment value is 80 (out of 100, the higher the value, the higher the degree of risk association), it indicates that the current health risk association degree of this patient is relatively high.

[0061] Optimize and adjust the risk assessment decision vector inside the target health warning model based on the health risk association assessment value to generate a health warning deviation vector. A risk assessment decision vector is preset inside the target health warning model, which is used to judge the health risk status of the patient. After the health risk association assessment value is generated, the model compares it with the preset risk assessment criteria. If the low-risk threshold for the comprehensive assessment of poor control of hypertension, blood sugar fluctuations, and influenza infection risk is 50, and the assessment value of the current patient, 80, exceeds this threshold, it indicates that the actual risk is higher than the expected low-risk level. The model optimizes and adjusts the internal risk assessment decision vector based on this difference. By adjusting each parameter in the vector, such as the weights of different risk factors, the boundary values for risk judgment, etc., a health warning deviation vector is generated. This vector records the deviation between the actual risk and the preset risk status. For example, the elements in the vector may represent information such as the deviation degree of hypertension risk, the deviation degree of blood sugar risk, and the weight adjustment of the impact of influenza infection risk deviation on overall health.

[0062] Analyze and transform the health warning deviation vector to generate patient health warning information, where the patient health warning information is used to characterize the health risk status faced by the patient and the quantitative indicators of the risk severity. The system analyzes and transforms the health warning deviation vector, converting the information in the vector into intuitive and easy-to-understand patient health warning information. For example, if the health warning deviation vector shows a large deviation in hypertension risk, and there are also certain degrees of deviation in blood sugar fluctuations and influenza infection risk, the patient health warning information generated after system analysis may be: "Your current health risk is relatively high. The control of hypertension is not good, the blood sugar fluctuates greatly, and at the same time, due to the high incidence of influenza in the recent period, your risk of influenza infection has increased. These factors may lead to serious consequences such as the onset of cardiovascular diseases and the aggravation of diabetic complications. It is recommended that you immediately strengthen the monitoring of blood pressure and blood sugar, adjust your medication according to the doctor's advice, try to reduce going out, and wear a mask if you need to go out to take protective measures." This health warning information clearly shows the health risk status faced by the patient or their family members and doctors, as well as the severity of the risk, and enables relevant personnel to have a clear understanding of the patient's health status in a quantitative manner (such as reflecting the overall risk level through the risk association assessment value), facilitating the timely adoption of corresponding health management measures.

[0063] In this application, the server collects the patient bracelet monitoring data, medical text data, environmental data, and knowledge base data, and performs preprocessing and feature extraction. For example, format conversion and key information extraction are performed on the medical text data, and meteorological and infectious disease information is screened from the environmental data. The knowledge base data is processed through the literature evidence strength grading system to ensure the accuracy and timeliness of the knowledge. In terms of model construction and application, models such as SEIR are used to process infectious disease-related data and evaluate the transmission risk. At the same time, the initial health warning model is trained by obtaining other patients' data, and the target model is obtained through optimization for analyzing the health risks of patients. For example, by combining the patient's physiological abnormality indicators, medical symptoms, and infectious disease risk levels, the health risk factors and degrees are evaluated.

[0064] Finally, based on the target health warning model and the updated knowledge base, the patient's health risk characteristics are processed to generate a health risk association evaluation value, the internal decision vector is adjusted to obtain a deviation vector, which is then converted into intuitive health warning information to provide disease risk prompts and prevention and treatment suggestions for patients, realizing efficient home health management, improving the patient's quality of life and saving medical resources.

[0065] In one implementation, as Figure 2 shown, this application also provides an artificial intelligence-based home health warning device, including:

[0066] An acquisition module 201, configured to acquire patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information;

[0067] A processing module 202, configured to preprocess the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features; process the environmental features and climate data features to generate infectious disease monitoring features; process the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data; process the target physiological data sequence, medical text features, and infectious disease monitoring features based on the target health warning model to generate target health risk features; process the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information.

[0068] Each embodiment in this application is described in a related manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the method, electronic device, electronic equipment, and readable storage medium for evaluating home health warning based on artificial intelligence, since they are basically similar to the embodiments of the method for home health warning based on artificial intelligence described above, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the embodiments of the method for home health warning based on artificial intelligence described above.

[0069] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application should be subject to the scope defined by the claims.

Claims

1. An artificial intelligence-based home health warning method, characterized in that, Including: Obtaining patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information; Preprocessing the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features; Processing the environmental features and climate data features to generate infectious disease monitoring features; Processing the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data; Processing the target physiological data sequence, medical text features, and infectious disease monitoring features based on the target health warning model to generate target health risk features; Processing the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information.

2. The method according to claim 1, characterized in that, Preprocessing the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features, including: Performing data cleaning on the patient bracelet monitoring data to remove outliers and missing values, and generating preliminarily processed bracelet physiological data; Performing feature extraction on the preliminarily processed bracelet physiological data to generate a target physiological data sequence, where the target physiological data sequence includes a heart rate feature sequence, a step count feature sequence, and a sleep duration feature sequence; Performing format conversion on the medical text data to generate medical text in a target format; Processing the medical text in the target format to generate medical text features, where the medical text features include symptom judgment features and symptom intervention features; Screening and performing feature extraction on the meteorological data in the environmental data to generate climate data features, where the climate data features include temperature features, humidity features, and air quality features; Screening and performing feature extraction on the non-meteorological data in the environmental data to generate environmental features, where the environmental features include infectious disease information in different regions and times.

3. The method according to claim 2, wherein Processing the environmental features and climate data features to generate infectious disease monitoring features, including: Extracting and screening the basic infectious disease information in the environmental features and the meteorological impact correlation information in the climate data features to generate potential risk factors for infectious disease transmission, meteorological factor correlation data, and basic information on the regional epidemic transmission trend; Processing the potential risk factors for infectious disease transmission, meteorological factor correlation data, and basic information on the regional epidemic transmission trend based on the infectious disease transmission model library to generate infectious disease transmission risk level information and transmission trend change information; Classifying and labeling the relevant data in the environmental features and climate data features based on the infectious disease transmission risk level information and transmission trend change information to generate a screening result of risk-associated data; Summarizing and analyzing the screening result of risk-associated data to generate infectious disease monitoring features, where the infectious disease monitoring features are used to characterize the transmission risk and trend of infectious diseases under different environmental and climate conditions.

4. The method according to claim 1, characterized in that, Processing the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data, including: Extract and classify the literature information in the knowledge base data based on the source and type, and generate scientific research achievement literature data, authoritative textbook literature data, clinical review literature data, and drug instruction literature data; Process the scientific research achievement literature data, authoritative textbook literature data, clinical review literature data, and drug instruction literature data based on the literature evidence strength grading system, and generate scientific research achievement evidence strength evaluation values, authoritative textbook evidence strength evaluation values, clinical review evidence strength evaluation values, and drug instruction evidence strength evaluation values; Based on the scientific research achievement evidence strength evaluation values, authoritative textbook evidence strength evaluation values, clinical review evidence strength evaluation values, and drug instruction evidence strength evaluation values, perform marking and screening processing on various types of literature data, and generate high evidence strength literature screening results, medium evidence strength literature screening results, and low evidence strength literature screening results; Based on the multi-source knowledge conflict resolution information, perform knowledge conflict detection and resolution processing on the high, medium, and low evidence strength literature screening results respectively, and generate high-confidence literature information; Obtain climate data and time information of different regions; Based on the climate data and time information of different regions, perform weighted processing on the high-confidence literature information to generate updated knowledge base data.

5. The method according to claim 1, wherein Obtain a target health warning model, including: Obtain other patients' bracelet monitoring data, medical text data, environmental data, updated knowledge base data, and medical knowledge update mechanism information; Based on a preset division ratio, process other patients' bracelet monitoring data, medical text data, and environmental data to generate training data and validation data, where the other patients' bracelet monitoring data, medical text data, and environmental data contain several health conditions and external environment-related information; Based on the training data, updated knowledge base data, and medical knowledge update mechanism information, process the initial health warning model to generate a warning model prediction parameter vector, where the warning model prediction parameter vector includes prediction strategies and related parameter information for health risk assessment; Based on the warning model prediction parameter vector, perform optimized training processing on the initial health warning model to generate a trained health warning model; Based on the validation data, perform simulated health warning processing on the trained health warning model to generate a validation result; Based on the validation result, perform evaluation and adjustment processing on the trained health warning model to generate a target health warning model.

6. The method according to claim 5, characterized in that Based on the target health warning model, process the target physiological data sequence, medical text features, and infectious disease monitoring features to generate target health risk features, including: Perform target data extraction and classification processing on the target physiological data sequence, medical text features, and infectious disease monitoring features to generate physiological abnormality index data, medical symptom target feature data, and infectious disease risk level data; Based on the target health warning model, analyze and process the physiological abnormality index data, medical symptom target feature data, and infectious disease risk level data to generate health risk factor importance evaluation information and health risk degree evaluation information; Based on the importance assessment information of health risk factors and the assessment information of the degree of health risk, screen and correlate the target data in the target physiological data sequence, medical text features, and infectious disease monitoring features to generate the screening result of health risk target data; Integrate and calculate the screening result of health risk target data, and process it in combination with the weight information in the target health warning model and the updated knowledge base to generate target health risk features, where the target health risk features are used to characterize the degree of risk and the main risk factors faced by the patient's current health status.

7. The method according to claim 6, characterized in that, Process the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information, including: Process the target health risk features based on the target health warning model and the updated knowledge base data to generate a health risk correlation assessment value; Optimize and adjust the risk assessment decision vector inside the target health warning model based on the health risk correlation assessment value to generate a health warning deviation vector; Analyze and transform the health warning deviation vector to generate patient health warning information, where the patient health warning information is used to characterize the health risk status faced by the patient currently and the quantitative indicators of the severity of the risk.

8. An artificial intelligence-based home health warning device, characterized in that, The device includes: An acquisition module, configured to acquire patient bracelet monitoring data, medical text data, environmental data, knowledge base data, and medical knowledge update mechanism information; A processing module, configured to preprocess the patient bracelet monitoring data, medical text data, and environmental data to generate a target physiological data sequence, medical text features, environmental features, and climate data features; process the environmental features and climate data features to generate infectious disease monitoring features; process the literature in the knowledge base data based on the literature evidence strength grading system to generate updated knowledge base data; process the target physiological data sequence, medical text features, and infectious disease monitoring features based on the target health warning model to generate target health risk features; process the target health risk features based on the target health warning model and the updated knowledge base data to generate patient health warning information.

9. An electronic device, characterized in that, Including: A first processor; And a memory, configured to store the executable instructions of the first processor; Wherein, the first processor is configured to execute the artificial intelligence-based home health warning method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a second processor, it implements the artificial intelligence-based home health warning method according to any one of claims 1 to 7.

Images