Medical intelligent health propaganda and education method and system

Abstract

The invention provides a medical intelligent health propaganda and education method and system. The method comprises the following steps: constructing a patient portrait according to patient static data and time sequence data; according to the patient portrait, propaganda and education prompt content is determined, a propaganda and education material library is generated through an AI large model, a propaganda and education agent and an aggregation diagnosis and disease knowledge base, and a patient personalized rich media propaganda and education document is generated; health propaganda and education plans and contents are pushed to medical staff, patients and family members through preset configuration; and in combination with propaganda and education result evaluation feedback of the patient and the family members on propaganda and education activities, through training model self-learning, propaganda and education contents are updated. The problems of single propaganda and education material, low pertinence, low efficiency, incapability of realizing closed-loop management and the like in the existing hospital propaganda and education mode are effectively solved, the propaganda and education efficiency and quality are high, the workload of medical personnel is effectively reduced, intelligent optimization and upgrading of propaganda and education closed-loop management are realized, the requirements of patients for health knowledge are met, and the popularization value is high. The hospital propaganda and education effect is obviously improved.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent medical technology, and more particularly to a medical intelligent health education method and system. Background Art

[0002] Health education is a crucial medical activity during a patient's hospitalization. Medical staff need to conduct health education for patients and their families at multiple stages and stages, with the education plan covering the entire cycle before, during, and after hospitalization. Education activities are also a crucial part of a patient's treatment. Existing health education is primarily conducted through manual labor, with manual preparation of educational materials, including writing educational text, creating educational images, editing documents, and filming educational audio and video. Traditional health education methods are primarily manual: 1) In most cases, medical staff first develop an education plan and then conduct face-to-face education with the patient or family according to the plan; 2) Educational videos are broadcast on the ward's television; and 3) General health education is conducted through the hospital website or WeChat official account.

[0003] Existing health education methods suffer from several significant issues: 1) slow updating of materials and content, outdated content, a single format, and a lack of targetedness; 2) low efficiency and quality of education. 3) frequent education sessions during patient readmissions, a heavy workload for medical staff, and the inability to integrate patient information, diagnosis, surgery, and medical orders to generate personalized education plans and achieve targeted, individualized education; 4) inability to rapidly implement closed-loop management of education programs, including intelligent optimization and upgrading of education plans, content, feedback, and evaluation. Therefore, how to effectively improve and enhance the consistency of education quality is a critical issue facing hospital medical staff. With the rapid development of AI big model technology, artificial intelligence technology, and medical information technology, patients' demand for health knowledge is increasing, despite existing health education programs. However, traditional education methods suffer from inaccurate content, single formats, and a lack of interactivity. Therefore, it is crucial to develop an intelligent health education system that can integrate multi-source patient data, construct patient profiles, automatically generate personalized education content based on an education knowledge base, and deliver it in real time. Summary of the Invention

[0004] The purpose of the embodiments of the present disclosure is to provide a medical smart health education method and system to solve the aforementioned problems existing in the prior art.

[0005] In order to achieve the above objectives, the technical solutions adopted in the embodiments of the present disclosure are as follows:

[0006] In one aspect, the present disclosure provides a medical smart health education method, the method comprising:

[0007] Build patient portraits based on patient static data and time series data;

[0008] Determine the content of education prompts based on the patient's profile. Through the AI ​​big model and education agent, aggregate the diagnosis and disease knowledge base to generate an education material library and generate personalized rich media education documents for patients.

[0009] Push health education plans and content to medical staff, patients and their families through preset configurations;

[0010] Combined with the evaluation and feedback from patients and their families on the results of the education activities, the education content is updated through self-learning through the training model.

[0011] Optionally, the patients include: potential patients and hospitalized patients;

[0012] Before building a patient portrait, determine the patient's static data and time series data.

[0013] If the patient is a potential inpatient, the static data of the potential inpatient includes: patient statistics and disease data; the time series data includes: based on the static data of the potential inpatient, the estimated t One or more time-varying physical sign sequence data related to a disease;

[0014] If the patient is an inpatient, the static data of the inpatient includes: patient statistical data, clinical characteristic data, occupational medicine special data and gender-specific medical data; time series data includes: patient data in the preset time series T t One or more vital sign sequence data that changes over time.

[0015] Optionally, the static data of potential hospitalized patients is estimated in a preset time series T t One or more time-varying sequence data of disease-related signs, including:

[0016] Calculate the static feature similarity between each inpatient and potential patient in the historical database;

[0017] The top K inpatients ranked by static feature similarity are selected as candidates, and their true vital sign sequences are extracted;

[0018] Align the candidate patient's vital sign sequence to the unified time axis T by linear interpolation t ;

[0019] The similarity of candidate patients is used as the weight, and the weighted average of each aligned vital sign sequence is performed to generate the potential inpatient sequence in T t Estimated sign sequence within.

[0020] Optionally, when the patient is a potential inpatient, obtain inpatients with the same occupation, age, and disease as the potential inpatient;

[0021] High-frequency common complications were extracted from the concurrent diseases of each hospitalized patient;

[0022] Each high-frequency common complication is included as part of static data to build a patient portrait, which is used to generate more forward-looking and personalized rich media education documents.

[0023] Optionally, the method of constructing a patient portrait based on the patient's static data and time series data includes:

[0024] The static eigenvalues ​​are obtained by weighted calculation of each factor in the patient's static data;

[0025]

[0026] in, represents the static eigenvalue, n represents the patient characteristic dimension, which is adjusted according to the department, and a i represents the feature vector of dimension i, F i represents the weight of dimension i, where i is one of age, gender, occupation or disease;

[0027] According to the temporal feature model TemporalNet (T t ), the patients are placed in the preset time series T t One or more time-varying vital sign sequence data are calculated as time series feature values;

[0028] According to the patient's static eigenvalues ​​and time series eigenvalues, a eigenvector is obtained;

[0029]

[0030] Among them, PatientProfile represents the patient profile, β represents the time series weight coefficient of physical signs;

[0031] Based on patients' static and dynamic data, key tags are generated using one or more methods including rule engines, classifiers, knowledge graph reasoning, and NLP extraction.

[0032] Optionally, when the time series data of the medical system changes, the weight of the time series features is adjusted, and the patient portrait is updated to update the education material library.

[0033] Optionally, when a pre-set expert doctor comes to the medical system for consultation or the patient changes the attending physician, the time series feature weight is increased, the education material library is updated, and the corresponding doctor information is added when the health education plan and content are pushed.

[0034] Optionally, the time series data includes one or more of pulse, heart rate, systolic blood pressure, and blood oxygen.

[0035] Another aspect of the present disclosure provides a medical smart health education system, comprising:

[0036] Patient portrait construction module, used to construct patient portraits based on patient static data and time series data;

[0037] The rich media education document generation module is used to determine the content of education prompts based on the patient profile. Through the AI ​​large model and education agent, it aggregates the diagnosis and disease knowledge base to generate an education material library and generate personalized health rich media education documents for patients.

[0038] The push module is used to push health education plans and content to medical staff, patients and their families through preset configurations;

[0039] The update module is used to update the education content based on the evaluation feedback of patients and their families on the education results of the activities, through self-learning of the training model.

[0040] On the other hand, an embodiment of the present disclosure provides an electronic device, comprising: at least one processor; a memory; and at least one application, wherein the at least one application is stored in the memory and configured to be executed by the at least one processor, and the at least one application is configured to: execute the steps of the method described above.

[0041] Another aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above-described method when executed by a processor.

[0042] The beneficial effects of the embodiments of the present disclosure are:

[0043] The disclosed embodiment constructs a patient portrait based on the patient's static data and time series data, generates personalized rich media education documents in combination with the AI ​​big model, updates the patient portrait in a timely manner, and updates the AI ​​big model based on the patient's and family's evaluation and feedback on the education results of the education activities, thereby realizing intelligent upgrading of the education content. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flowchart of a medical smart health education method proposed in an embodiment of the present disclosure;

[0045] Figure 2 This is a schematic diagram of the flow of various models used in a medical smart health education method proposed in an embodiment of the present disclosure;

[0046] Figure 3 This is a schematic diagram of the structure of a medical smart health education system proposed in an embodiment of the present disclosure;

[0047] Figure 4 This is a source diagram of a health education material library in a medical smart health education method proposed in an embodiment of the present disclosure;

[0048] Figure 5 A schematic diagram of the structure of a rich media health education document is provided by using a medical smart health education method proposed in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present disclosure more clear, the embodiments of the present disclosure are further described in detail below with reference to the accompanying drawings. It should be understood that the specific implementation methods described herein are only used to explain the embodiments of the present disclosure and are not intended to limit the embodiments of the present disclosure.

[0050] like Figure 1 As shown, the embodiment of the present disclosure provides a medical smart health education method, which is mainly used in the field of smart medical technology. The method includes:

[0051] Step S100: Construct a patient portrait based on the patient's static data and time series data.

[0052] The disclosed embodiments can extract basic patient information, namely statistical data such as age, gender, occupation, etc., as well as comprehensive assessment, diagnosis results, treatment plans, medication information, medical history and other data through the hospital's inpatient management system HIS, hospital information system LIS, computerized medical record system EMR, anesthesia and other systems to construct a patient education feature portrait.

[0053] It should be noted that it is also possible to connect with the hospital's existing health education material library, physical examination appointment system, health management system, chronic disease management system and other integrated platforms with multiple sources and heterogeneous systems to extract, integrate, and standardize patient education feature data, such as basic patient information, risk assessment indicators, surgical information, diagnosis information, and treatment information. To construct a high-quality patient portrait, the collected data must be accurate and complete. However, due to the possibility of subjective bias among users and the diversity of data sources, the data may be inaccurate or incomplete. To this end, a variety of data verification methods can be adopted, such as cross-validation of data from different sources and comparison of the consistency of user self-reported data with the diagnosis results of medical institutions, to obtain more accurate and comprehensive data.

[0054] like Figure 4 As shown, the source of health education materials is the service running on the server deployed in the computer room, which is collected through three channels.

[0055] Before step S100, the embodiment of the present disclosure divides the data obtained from each system into static data and time series data. The static data includes: patient statistical data, clinical characteristic data, occupational medicine special data, gender-specific medical data, and other data that remain unchanged for a certain period of time, or data with a low frequency of change; the time series data includes: one or more vital signs of the patient that change over time, such as pulse, heart rate, systolic blood pressure, respiratory rate, blood oxygen, electrocardiogram, pain score, and other information that changes frequently.

[0056] The disclosed embodiment can periodically acquire time series data and static data to update the patient portrait.

[0057] This disclosed embodiment categorizes patients into potential patients and hospitalized patients. Potential patients are those who have undergone physical examinations or outpatient visits at medical institutions but have not been hospitalized. Hospitalized patients are those who have been admitted to medical institutions for treatment. Health education for potential patients includes multiple push notifications related to the user's potential health and preventive information. For hospitalized patients, education content is updated in real time based on changes in the patient profile.

[0058] The patients include: potential patients and hospitalized patients; before constructing a patient portrait, the patient's static data and time series data are determined.

[0059] If the patient is a potential patient, the static data of the potential patient includes: patient statistical data and disease data; the time series data includes: based on the static data of the potential patient, the estimated t One or more time-varying vital sign sequence data related to a disease.

[0060] The static data of potential patients is estimated in the preset time series T t One or more time-varying sequence data of disease-related signs, including:

[0061] Calculate the static feature similarity between each inpatient and potential patient in the historical database;

[0062] The top K inpatients ranked by static feature similarity are selected as candidates, and their true vital sign sequences are extracted;

[0063] Align the candidate patient's vital sign sequence to the unified time axis T by linear interpolation t ; Ensure that the vital sign values ​​at each time point are comparable;

[0064] The similarity of candidate patients is used as the weight (normalized to a total of 1), and the weighted average of each aligned sign sequence is performed to generate the potential patient in T t Multivariate estimated vital sign series (such as blood pressure, pulse, etc.)

[0065] Search the historical database for K real patients that are most similar to the static characteristics (age, occupation, disease, etc.) of the potential patient, and take the weighted average of their time series data as the estimated result.

[0066] If the patient is an inpatient, the static data of the inpatient includes: patient statistical data, clinical characteristic data, occupational medicine special data and gender-specific medical data; time series data includes: one or more vital signs data of the patient that change over time within a preset time period, such as heart rate, blood pressure, etc.; chronic disease progression tracking data, such as chemotherapy cycles, blood sugar trends (daily granularity); cross-scale correlation data, such as connecting short-term anomalies with long-term prognosis (such as postoperative fever and infection risk assessment values).

[0067] Quantitative and qualitative classification is performed based on patient education feature data to construct a patient education portrait, and the qualitative description feature information is converted into quantitative numerical form to improve the accuracy of the patient portrait. Specifically, deep reasoning and analysis of the feature information is performed through algorithms to dynamically obtain the quantitative value of the patient feature portrait, which consists of static feature values ​​and time series feature values.

[0068] Exemplarily, constructing a patient profile based on the patient's static data and time series data includes:

[0069] Step S110: weighted calculation is performed on each factor in the patient's static data to obtain a static characteristic value;

[0070]

[0071] in, represents the static eigenvalue, n represents the patient characteristic dimension, which is adjusted according to the department, and a i represents the feature vector of dimension i, F i represents the weight of dimension i, where i is age, gender, occupation or disease, and a age is a piecewise function, [0.1, <18 years old]\[0.3, 18 to 65 years old]\[0.5, >65 years old], F age Normalized value (0-1), F age =age / 100, other eigenvector values ​​and dimension weights of patients a i Pre-set according to actual situation or experience.

[0072] Static eigenvalues ​​are based on patient information such as age, gender, occupation, and disease. They are selected based on clinical standards such as medical guidelines, gender-specific diseases, occupational medicine, and clinical medicine, and are weighted and adjusted according to rules. The dimension range is set to 64.

[0073] Expanding by clinical dimensions: Dimensions of static features, such as age, gender, occupation, and disease, are grouped according to medical guidelines. Each group is assigned a fixed number of dimensions, determined based on practical experience. For example, age is divided into three segments (<18 / 18-65 / >65) and encoded using a 3-dimensional one-hot encoding. Gender is 2-dimensional (male / female), occupation is 20-dimensional (classified according to occupational medicine standards), and disease is 39-dimensional (selecting high-frequency diseases based on ICD-10 clinical subdivisions). After concatenation, a 64-dimensional sparse vector (3 + 2 + 20 + 39 = 64) is generated, with each dimension corresponding to a clinically interpretable feature. The corresponding contribution value of each group is padded to certain positions (e.g., dimension 0) of the corresponding dimension group, with the remaining positions padded with zeros. After concatenation and merging, the resulting scalar static feature values ​​are assigned to specific positions in the 64-dimensional vector, and the remaining positions are padded with zeros or other clinically relevant features.

[0074] Step S120: According to the temporal feature model TemporalNet (T t ), the patients are placed in the preset time series T t One or more time-varying vital sign sequence data are calculated as time series feature values.

[0075] The temporal feature value is based on the temporal feature model TemporalNet, which focuses on the factors that change in the patient at a certain time, such as changes in pulse, heart rate, systolic blood pressure, etc. The β1 weight coefficient is generally in the range of 0.4 to 0.7 according to clinical standards. The patient's time series T t Measured vital sign data, such as body temperature, heart rate, blood oxygen, blood pressure, etc., minute-level fluctuations of vital signs, and time series features also include chronic disease progression tracking: chemotherapy cycles, blood sugar trends (daily granularity), β2 weight coefficient, 0.2-0.5; cross-scale association: connecting short-term abnormalities with long-term prognosis, such as postoperative fever and infection risk assessment value, β3 weight coefficient, 0.1-0.3.

[0076] The change of one vital sign over time is sequence data, and the change of multiple vital signs over time is sequence matrix data.

[0077] Step S130: Obtain a feature vector based on the patient's static feature value and time series feature value.

[0078]

[0079] Where PatientProfile represents the patient profile and β represents the weight coefficient. The output is a 64-dimensional vector, where each dimension is a linear combination of static and temporal features.

[0080] Step S140: Based on the patient's static data and dynamic data, one or more methods including rule engine, classifier, knowledge graph reasoning and NLP extraction are used to generate key tags.

[0081] When using multiple methods, multiple key labels are obtained. The label fusion and deduplication mechanism are used to generate the final key label. The patient portrait includes: quantitative feature vectors and qualitative key labels.

[0082] Step S200: Determine the content of the education prompts based on the patient portrait, aggregate the diagnosis and disease knowledge base through the AI ​​big model and the education intelligent agent Agent, generate an education material library, and generate patient-personalized rich media education documents.

[0083] Based on the patient education feature portrait, the education prompt description is defined. Through the AI ​​big model and education agent, the diagnosis and disease knowledge base are aggregated to complete the patient education logic deduction and generate the education material library and generate the patient's personalized health rich media education documents, such as Figure 5 As shown, rich media mission documents include: mission plans, mission documents and mission evaluations, etc.

[0084] Based on the qualitative and quantitative data of the patient education feature portrait and combined with the basic rule library, a concise and structured prompt statement framework can be generated, that is, the primary education prompt description, which is a coarse-grained prompt description for subsequent enhancement; the education intelligent agent agent product generates patient rich media education materials and education plans: the education intelligent agent agent completes patient image recognition, realizes full network search, logical reasoning, and deep thinking by calling one of DeepSeek, Huawei Pangu Big Model, Baidu Big Data Model, and Alibaba Tongyi Big Model, and optimizes the existing education knowledge base in the hospital, including information such as patient diseases, treatment status, medication, preferences and habits, to generate rich media health education materials. The education agent uses a dynamic rule library and medical knowledge graph to analyze the patient profile and preliminary education prompt descriptions, outputting a structured context and an enhanced prompt description framework. The structured context includes the profile, rules, strategies, and an enhanced prompt description framework. The enhanced prompt description framework includes more comprehensive behavioral recommendations, push frequency scheduling, rich media resource IDs such as WBC_Diet_301, and multi-dimensional intervention directions such as medication, diet, exercise, and psychology. This can be used as input for LLM / AI natural language generation. The agent outputs the structured context, which includes the complete health guidance text expressed in natural language by the LLM / AI, a rich media organization structure (images, text, video, and reminders), and the final personalized document (JSON or HTML).

[0085] For example, input data: patient information:

[0086] "Basic Information": {

[0087] "Name": "Zhang Wei",

[0088] Age: 72

[0089] "Occupation": "Teacher",

[0090] "Education level": "University",

[0091] "Device": "Tablet"

[0092] "Medical Data": {

[0093] "Diagnosis": ["knee osteoarthritis (ICD-10: M17)", "type 2 diabetes mellitus (E11.9)"],

[0094] "Surgery": "Total knee replacement (2025-05-20)",

[0095] "Doctor's orders":

[0096] "Rivaroxaban 10mg qd*30 days",

[0097] "Rehabilitation training: ankle pump exercise 10 times / hour",

[0098] "Stitches removed 7 days after surgery"

[0099] "Behavior data": {

[0100] "Video viewing completion rate": 78%,

[0101] "Knowledge test score": 42 / 100

[0102] Generated intelligent portrait:

[0103] "Eigenvector": [0.8, 0.6, 1.0, 0.7, 0.3, ...], / / normalized values

[0104] "Key tags": [

[0105] {"tag":"orthopedics postoperative","weight":0.95},

[0106] {"tag":"diabetes management","weight":0.85},

[0107] {"tag":"low health literacy","weight":0.75},

[0108] {"tag":"visual aid needs","weight":0.65};

[0109] Missionary tips:

[0110] "Module": "Medication Reminder";

[0111] "Content": "Take rivaroxaban (white tablet) this morning to avoid missing a dose."

[0112] "Module": "Wound Care";

[0113] "Content": "Keep the dressing dry. If you experience redness, swelling, heat, pain, or a temperature exceeding 38°C, please inform your doctor immediately."

[0114] "Module": "Rehabilitation Training",

[0115] "Content": "Follow the video and do ankle pump exercises (10 times each time, 1 set every hour) to promote blood circulation.

[0116] "

[0117] "Module": "Diabetes Management",

[0118] Content: "Monitor blood sugar in the morning and keep it between 6-10mmol / L. Pay attention to your diet."

[0119] "Module": "Health literacy improvement",

[0120] "Content": "Complete the quiz after watching the video to help you better understand the key points of postoperative care.

[0121] "Module": "Visual Assisted Recommendation",

[0122] "Content": "Animated videos with audio explanations are prepared for you to understand and remember easily."

[0123] "Rivaroxaban is a new anticoagulant drug used to prevent deep vein thrombosis."

[0124] "Ankle pump exercises can help improve blood circulation in the lower limbs and prevent postoperative blood clots."

[0125] Mission Strategy

[0126] "Content complexity": "Elementary (FKGL≤6)";

[0127] "Media Preference": ["animated video", "voice guidance"];

[0128] "Push frequency": "Twice a day (09:00, 19:00)";

[0129] Example of generated educational content:

[0130] Title: Your Joint Replacement Surgery Day 3 Care;

[0131] Media format: 3-minute animated video (with audio)

[0132] Key points:

[0133] Medication reminder: Take rivaroxaban (white tablet) this morning;

[0134] Wound care: Keep the dressing dry and inform the doctor immediately if the body temperature is greater than 38°C;

[0135] Rehabilitation training: follow the video to do ankle pump exercises (10 times each time, 1 set every hour);

[0136] Note for diabetics: Monitor blood sugar in the morning (target value 6-10mmol / L).

[0137] For example, when the patient is a potential inpatient, inpatients with the same or similar occupation, age, and disease as the potential inpatient are obtained.

[0138] High-frequency common complications were extracted from the concurrent diseases of each inpatient; complications occurring in more than 50% of the inpatients with the same or similar occupation, age, and disease as the potential inpatient were considered high-frequency complications.

[0139] Each high-frequency common complication is included as part of static data to build a patient portrait, which is used to generate more forward-looking and personalized rich media education documents.

[0140] In the process of generating education for potential hospitalized patients, it is necessary to introduce the step of "extracting high-frequency concurrent diseases from hospitalized cases" and incorporate it into the patient portrait as a new static data dimension. The AI ​​big model will then generate more forward-looking and targeted personalized rich media education documents.

[0141] Use machine learning algorithms to conduct in-depth analysis of the integrated data and generate personalized education plans.

[0142] Improve the classification accuracy of education plans and materials through education models and intelligent agents. Education models can be AI or large language models such as LLM.

[0143] When doctors in the medical system change, the weight of the time series features is adjusted and the patient portrait is updated to update the educational material library.

[0144] Physician changes, such as a rotation of attending physicians, may alter the diagnosis and treatment plan, necessitating an update to the patient profile by adjusting the weights of temporal features (e.g., increasing the importance of vital signs monitored by the new physician). Alternatively, if a leading expert is available, prioritize educational content in their area of ​​expertise.

[0145] When a pre-set expert doctor comes to the medical system for consultation or the patient changes his attending physician, the time series feature weight is increased, the education material library is updated, and the corresponding expert information is added when pushing health education plans and content.

[0146] Different doctors have different expertise or may focus on different aspects for the same patient. For example, Doctor A is more concerned about blood pressure fluctuations, while Doctor B is more concerned about blood sugar trends. Adjust the weight of time series features based on the doctor, update the education materials, and carry the corresponding information when pushing health education plans and content.

[0147] The time series data includes: one or more of pulse, heart rate, systolic pressure, and blood oxygen.

[0148] Step S300: Push health education plans and content to medical staff, patients and their families through preset configurations.

[0149] In other words, through configuration settings, health education plans and content can be pushed to medical staff, patients, and their families in a targeted manner through multiple channels and methods, meeting the personalized needs of patients for health education. Based on the patient's disease, treatment status, medication, preferences, and habits, appropriate channels and devices are selected to push education content to the patient. Specifically, medical staff, patients, and their families can log in through audio and video terminals such as mobile phones, PDAs, and televisions to view and browse personalized education plans and complete education activities according to the plan.

[0150] Step S400: Based on the evaluation feedback of the education results of the education activities by patients and their families, the education content is updated through self-learning of the training model.

[0151] The method of the disclosed embodiment can effectively solve the problems existing in the existing hospital education methods, such as single education materials, lack of pertinence, low efficiency, and inability to achieve closed-loop management. It has high education efficiency and quality, can effectively reduce the workload of medical staff, realize intelligent optimization and upgrading of education closed-loop management, meet patients' needs for health knowledge, and significantly improve the hospital education effect.

[0152] like Figure 2 As shown in the figure, based on the collection of patient education feedback, the content of the education is evaluated and optimized, and the system interacts with patients, answers questions, and provides personalized suggestions. The accuracy of the education content is also improved through learning models.

[0153] Specifically, the Q-Learning algorithm is used to enhance model self-learning and training, thereby improving the accuracy of educational content. This allows AI models to increase the breadth of educational content, while training models can improve the accuracy of educational content.

[0154] like Figure 3 As shown, another aspect of the present disclosure provides a medical smart health education system, the system comprising:

[0155] The patient portrait construction module 100 is used to construct a patient portrait based on the patient's static data and time series data.

[0156] The rich media education document generation module 200 is used to determine the content of education prompts based on the patient portrait, aggregate the diagnosis and disease knowledge base through the AI ​​big model and the education intelligent agent Agent, generate an education material library, and generate patient-personalized health rich media education documents.

[0157] Push module 300 is used to push health education plans and content to medical staff, patients and their families through preset configurations.

[0158] The updating module 400 is used to update the content of the education activities by combining the evaluation feedback of the patients and their families on the education results through self-learning of the training model.

[0159] The system of the embodiment of the present disclosure further includes a data acquisition module for acquiring static data and time series data of hospitalized patients and potential hospitalized patients respectively; an estimation module for estimating the potential hospitalized patients' static data within a preset time series T t One or more time-varying vital sign sequence data related to the disease; determining the patient data module, if the patient is a potential inpatient, the static data of the potential inpatient includes: patient statistical data and disease data; time series data includes: based on the static data of the potential inpatient, the estimated t One or more time-varying physical sign sequence data related to a disease;

[0160] If the patient is an inpatient, the static data of the inpatient includes: patient statistical data, clinical characteristic data, occupational medicine special data and gender-specific medical data; time series data includes: patient data in the preset time series T t One or more vital sign sequence data that changes over time.

[0161] The system of the disclosed embodiment will operate an artificial intelligence big model to automatically generate education content, using multi-source heterogeneous data to integrate the education documents, medical records, and disease knowledge accumulated in historical periods with AI loading as an education training material library; automatically update the education material library through the AI ​​big data model; models include: DeepSeek, Huawei Pangu big model, Baidu big data model, Ali Tongyi big model, etc., to complete the automatic generation of health education materials; it can also generate health education materials through the construction of open source private big models, through medical records, medical standards, medical documents and other data information; at the same time, it supports traditional manual editing, recording and custom education materials.

[0162] Another aspect of the disclosed embodiments provides a medical smart health education system. The system comprises a server, a multimedia terminal, a medical staff computer, and multimedia terminals such as televisions, mobile phones, large screens, and ward screens. The system is provided on the server and further includes a rich media interaction subsystem for education, a health education management software subsystem, and a rich media education document generation module 200, i.e., an intelligent education material generation subsystem. The system implements the following steps:

[0163] Step 1: The patient logs in, and the system's multimedia terminal obtains patient information from the server, such as personal information, patient assessment, diagnosis information, and course of disease information, to build a patient education portrait.

[0164] Step 2: Automatically generate patient rich media health education documents, transmit them to the server, and push the content to the corresponding terminal according to the system management configuration requirements.

[0165] Automatically generate patient education rich media documents based on the education rich media configuration template.

[0166] The system server pushes educational rich media documents to the multimedia terminal according to the patient's binding time.

[0167] Step 3: After receiving the education information prompt, the medical staff, patient, or family member clicks to automatically open the education rich media interactive subsystem, completes the education process, and provides evaluation feedback.

[0168] Step 4: The patient answers the education questions and is educated again if necessary.

[0169] Step 5: Answer the questions, record the teaching time, sign electronically or take a photo, and upload them to the server.

[0170] Step 6: During the patient's treatment, repeat the above activities according to the education plan and record the education results.

[0171] Step 7: After the education plan is completed, the patient or family members will score the education effect or fill out feedback, and the results will be uploaded to the server.

[0172] Step 8. Medical staff review the education results through the health education management system, count and analyze the education time, answer scores, education scores, etc., and determine the improvement plan based on the statistical results.

[0173] Step 9: If you are not satisfied with the education materials, start the automatic update program for the education materials.

[0174] Step 10: The system self-learns and trains based on the education feedback to improve the education content.

[0175] Step 11: The educational materials are updated, reviewed, confirmed, and distributed to relevant departments and treatment phases.

[0176] Step 12: The education process is completed.

[0177] The system of the disclosed embodiment consists of a server deployed in the computer room, a call terminal deployed in the ward, computers for doctors and nurses, and a mobile terminal at the nurse workstation. The server and the terminals are connected via a network, and the system can deploy multiple fixed terminals (1, 2, ... n) and multiple mobile terminals (1, 2, ... m).

[0178] The system of the embodiment of the present disclosure uses a server to complete the storage of health education materials, run the education management program, and automatically generate multi-source heterogeneous health education materials, patient portraits, education plans and other functions; the server also completes the data calculation, service scheduling, information processing and other tasks of the education system, and also connects with the hospital's existing information systems such as HIS, EMR, medical record system, and nursing system to obtain patient personal and treatment information and upload education results.

[0179] Multimedia terminals include electronic products commonly found in hospitals, such as televisions, ward screens, mobile or handheld terminals, and mobile phones; they can play, watch, browse, and read educational content. After learning, patients can choose to answer questions in the document, provide scoring and feedback on the effectiveness of the education, evaluate the effectiveness of the education through the terminal, and assess the education services provided by medical staff.

[0180] The health education management software subsystem runs on the computer of the attending physician or responsible nurse, and completes the tasks of viewing patient information, defining health education plans and education rich media document templates, managing education progress, monitoring education effects and education scores, and scheduling personalized education. It also associates and binds patient education terminals (including education methods) with patients, and finally performs statistical analysis on the duration, number of times, execution time, number of people, and evaluation scores of education. Based on the analysis results, it determines whether to upgrade the education materials and whether to adjust the education plan, forming a PDCA closed loop for continuous improvement of medical education quality management. By integrating the above modules, models, and algorithms, this system will effectively realize the intelligentization of the entire process from patient disease characteristic analysis → education content generation → education plan and personalized definition → effect evaluation, reducing the workload of medical staff, while supporting the intelligent iteration of the material library, ultimately improving the efficiency of medical education and patient compliance, and greatly enhancing the quality and effectiveness of education.

[0181] On the other hand, an embodiment of the present disclosure provides an electronic device, characterized in that it includes: at least one processor; a memory; and at least one application, wherein the at least one application is stored in the memory and configured to be executed by the at least one processor, and the at least one application is configured to: execute the steps of the method described above.

[0182] Another aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above-described method when executed by a processor.

[0183] Example 1: Using the patient health education method of the embodiment of the present disclosure, the improvement effect is achieved:

[0184] 1) Using the case of lung cancer chemotherapy patients, the quantitative feature generation implementation method is as follows:

[0185]

[0186] Output patient portrait:

[0187] "Eigenvector": [0.7, 0.5, 0.9, 0.8, 0.4, ...];

[0188] Key tags: [

[0189] In chemotherapy, "weight": 0.95;

[0190] Leukopenia, "weight": 0.90;

[0191] Tumor in the advanced stage (stage IIIB), "weight": 0.85;

[0192] Advanced age risk, "weight": 0.75 age > 65 years;

[0193] Visual aid requirements, "weight": 0.60;]

[0194] Agent generates structured context:

[0195] Health Education Strategy

[0196] Content complexity: elementary (FKGL≤6);

[0197] Media preference: [animated video, audio guide, illustrated manual];

[0198] Push frequency: twice a day (09:00, 18:00);

[0199] Key intervention areas: dietary guidance for white blood cell production, infection prevention, chemotherapy side effect management, and fall prevention;}

[0200] Description of the content of the missionary prompt:

[0201] "Module": "Medication Reminder",

[0202] Content: Please take the medication prescribed by your doctor on time and pay attention to any adverse reactions.

[0203] "Module": "Precautions for low white blood cell count"

[0204] "Content": "The current white blood cell count is low (2.8×10 9 / L), please try to avoid going to crowded places, wash your hands frequently, and pay attention to personal hygiene;

[0205] Finally, the AI ​​model generates personalized rich media education documents:

[0206] title:

[0207] "You are receiving the third cycle of TP chemotherapy - please pay attention to changes in white blood cells and nutritional intake";

[0208] Medical Tips Summary:

[0209] Current treatment status:

[0210] The third cycle of TP chemotherapy is currently underway;

[0211] The pain score was low (2 points) and well controlled;

[0212] White blood cell count was low (2.8×10 9 / L), be alert to the risk of infection;

[0213] Risk Reminder:

[0214] White blood cells <3.0×10 9 / L→The risk of infection is significantly increased;

[0215] Age > 65 years → increased risk of falls (α_fall = 0.5);

[0216] It is recommended to monitor body temperature daily, pay attention to oral hygiene, and avoid crowds;

[0217] Rich media content recommendation:

[0218] Video resources:

[0219] [Whitening Diet Video]

[0220] Recommended viewing time: After breakfast every day;

[0221] Key points: Recommended foods rich in protein, vitamin B6 / B12, and iron, such as beef, eggs, and spinach;

[0222] Illustrated manual:

[0223] "Dietary Guidelines During Chemotherapy" PDF version (available for download and reading);

[0224] Home Care Tips: How to Prevent Infections e-booklet.

[0225] After the implementation of this program, according to feedback statistics, important indicators such as the matching degree of education content and patient compliance have been improved to varying degrees:

[0226] index Traditional systems This program Optimization principle Matching degree of missionary content 68％ 93％ Patient multi-dimensional feature fusion + dynamic weight adjustment Frequency of repetition 100％ 60％ Real-time push linkage + reminder mechanism Improved patient compliance 21％ 58％ Multi-channel, professional characteristics-driven push time optimization

[0227] The above is only a preferred implementation of the embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the embodiment of the present disclosure. These improvements and modifications should also be considered within the scope of protection of the embodiment of the present disclosure.

Claims

1. A medical wisdom health education method, characterized in that: The method comprises: Build patient portraits based on patient static data and time series data; Determine the content of education prompts based on the patient's profile. Through the AI ​​big model and education agent, aggregate the diagnosis and disease knowledge base to generate an education material library and generate personalized rich media education documents for patients. Push health education plans and content to medical staff, patients and their families through preset configurations; Combined with the evaluation and feedback from patients and their families on the results of the education activities, the education content is updated through self-learning through the training model.

2. The method according to claim 1, characterized in that Said patients include: potential patients and hospitalized patients; Before building a patient portrait, determine the patient's static data and time series data. If the patient is a potential inpatient, the static data of the potential inpatient includes: patient statistics and disease data; the time series data includes: based on the static data of the potential inpatient, the estimated t One or more time-varying physical sign sequence data related to a disease; If the patient is an inpatient, the static data of the inpatient includes: patient statistical data, clinical characteristic data, occupational medicine special data and gender-specific medical data; time series data includes: patient data in the preset time series T t One or more vital sign sequence data that changes over time.

3. The method according to claim 2, characterized in that According to the static data of potential hospitalized patients, it is estimated that t One or more time-varying sequence data of disease-related signs, including: Calculate the static feature similarity between each inpatient and potential patient in the historical database; The top K inpatients ranked by static feature similarity are selected as candidates, and their true vital sign sequences are extracted; Align the candidate patient's vital sign sequence to the unified time axis T by linear interpolation t ; The similarity of candidate patients is used as the weight, and the weighted average of each aligned vital sign sequence is performed to generate the potential inpatient sequence in T t Estimated sequence of vital signs within.

4. The method according to claim 2, characterized in that When the patient is a potential inpatient, obtain inpatients with the same occupation, age, and disease as the potential inpatient; High-frequency common complications were extracted from the concurrent diseases of each hospitalized patient; Each high-frequency common complication is included as part of static data to build a patient portrait, which is used to generate more forward-looking and personalized rich media education documents.

5. The method according to claim 1 or 2, characterized in that The method for constructing a patient portrait based on the patient's static data and time series data includes: The static eigenvalues ​​are obtained by weighted calculation of each factor in the patient's static data; in, represents the static eigenvalue, n represents the patient characteristic dimension, which is adjusted according to the department, and a i represents the feature vector of dimension i, F i represents the weight of dimension i, where i is one of age, gender, occupation or disease; According to the temporal feature model TemporalNet (T t ), the patients are placed in the preset time series T t One or more time-varying vital sign sequence data are calculated as time series feature values; According to the patient's static eigenvalues ​​and time series eigenvalues, a eigenvector is obtained; Among them, PatientProfile represents the patient profile, β represents the time series weight coefficient of physical signs; Based on patients' static and dynamic data, key tags are generated using one or more methods including rule engines, classifiers, knowledge graph reasoning, and NLP extraction.

6. The method according to claim 5, characterized in that When the patient stage time series data in the medical system changes, the weight of the time series features is adjusted, the patient portrait is updated, and the education material library is updated.

7. A medical wisdom health education system, characterized by: The system comprises: Patient portrait construction module, used to construct patient portraits based on patient static data and time series data; The rich media education document generation module is used to determine the content of education prompts based on the patient profile. Through the AI ​​large model and education agent, it aggregates the diagnosis and disease knowledge base to generate an education material library and generate personalized health rich media education documents for patients. The push module is used to push health education plans and content to medical staff, patients and their families through preset configurations; The update module is used to update the education content based on the evaluation feedback of patients and their families on the education results of the activities, through self-learning of the training model.

8. An electronic device, characterized in that: include: at least one processor; Memory; At least one application, wherein the at least one application is stored in a memory and configured to be executed by at least one processor, and the at least one application is configured to: execute the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images