An inquiry guidance system based on interactive data collection

Abstract

The present invention discloses an interrogation guidance system based on interactive data collection, which relates to the technical field of interrogation guidance. It includes a data acquisition module, a hypertension feature extraction module, a hypertension risk prediction module, and a guidance adjustment module. By integrating static and dynamic blood pressure data, time series analysis, anomaly detection, and machine learning models, it realizes the accurate identification of white coat hypertension and masked hypertension. It can not only combine the user's physiological data with the characteristics of the measurement environment to improve the accuracy of hypertension risk prediction, but also adjust personalized measurement suggestions through dynamic interrogation guidance and generate targeted blood pressure management plans. This technology effectively reduces the situation of misjudging hypertension due to emotional fluctuations or single measurement, improves the scientific nature of hypertension screening and health management, optimizes the patient's medical decision-making, reduces the risk of unnecessary medical intervention, and thus improves the reliability and clinical application value of the intelligent medical system.

Description

Technical Field

[0001] The present invention relates to the technical field of consultation guidance, and particularly to a consultation guidance system based on interactive data collection. Background Art

[0002] With the development of artificial intelligence (AI), big data analysis, and Internet of Things (IoT) technologies, intelligent medical systems have been widely applied in fields such as remote consultation, health management, and disease prediction. Among them, a consultation guidance system based on interactive data collection can collect patients' symptom information through voice, text, or graphical interaction methods, and combine medical knowledge bases and machine learning models to provide preliminary diagnosis suggestions to assist doctors in making decisions. Such systems are widely used in Internet medical care, intelligent health devices, and hospital pre-triage systems, improving the utilization rate of medical resources and optimizing the patient's medical experience.

[0003] Current consultation guidance systems usually integrate various physiological parameter monitoring functions, such as blood pressure measurement, heart rate monitoring, blood oxygen saturation detection, etc., to provide a more comprehensive health assessment. For example, intelligent wearable devices (such as smart watches, blood pressure monitors) can measure users' blood pressure data in real time and transmit it to the consultation system. Combining with the user's symptom description, the system can further analyze potential disease risks, such as hypertension, cardiovascular diseases, etc. In addition, some systems also support ambulatory blood pressure monitoring, that is, by long-term tracking of blood pressure change trends, to identify morning peak hypertension or masked hypertension, and provide personalized health management plans for users.

[0004] The existing technologies have the following deficiencies:

[0005] When a patient uses an intelligent blood pressure monitoring device at home and uploads blood pressure data through a mobile consultation system. The system uses a traditional decision tree model based on static blood pressure values for hypertension risk assessment and gives a preliminary diagnosis in combination with the patient's main complaint symptoms (such as "occasionally dizzy"). If the patient's blood pressure rises due to emotional tension during a certain measurement (such as 140 / 90 mmHg), the system may misjudge it as grade 1 hypertension and recommend immediate medical treatment. However, this patient actually belongs to white coat hypertension. If subsequent 24-hour ambulatory blood pressure monitoring (ABPM) is not performed, unnecessary antihypertensive treatment may be received, affecting health. On the other hand, a certain patient has masked hypertension, with a daily blood pressure of 150 / 95 mmHg, but during the visit, the measured blood pressure is only 120 / 80 mmHg due to mental relaxation. Since the consultation system relies on the single measurement value during the visit and fails to combine the patient's daily blood pressure trend, it wrongly determines that the patient's blood pressure is normal, resulting in the patient not receiving timely intervention, which may increase the risk of stroke or heart disease in the long run. Summary of the Invention

[0006] The object of the present invention is to provide an interrogation guidance system based on interactive data collection to solve the deficiencies in the background art.

[0007] To achieve the above object, the present invention provides the following technical solution: An interrogation guidance system based on interactive data collection, comprising a data acquisition module, a hypertension feature extraction module, a hypertension risk prediction module, and a guidance adjustment module;

[0008] Data acquisition module: Obtain the real-time health data of the user, including blood pressure measurement data, heart rate, blood oxygen saturation, and symptom information independently input by the user;

[0009] Hypertension feature extraction module: Combine time series analysis methods to calculate the blood pressure change trend, distinguish short-term fluctuations from long-term trends, and use anomaly detection algorithms to identify the characteristics of white coat hypertension and masked hypertension;

[0010] Hypertension risk prediction module: Adopt a machine learning model that integrates static and dynamic blood pressure data, combine the historical hypertension characteristics of the patient and the characteristics of the measurement environment to predict the hypertension risk, and dynamically adjust the weights of the machine learning model in combination with the individual characteristics of different users;

[0011] Guidance adjustment module: After the user uploads blood pressure data, the system automatically adjusts the questioning content based on the measurement data through dynamic interrogation guidance. If suspected white coat hypertension or masked hypertension is detected, the system will remind the user to perform multi-period blood pressure measurements and generate a personalized blood pressure management plan.

[0012] Preferably, in the data acquisition module, blood pressure, heart rate, and blood oxygen saturation are collected through intelligent wearable devices; measurement data is transmitted through intelligent medical devices; the main complaints of the user are collected through the user's independent input and associated with physiological parameters for analysis.

[0013] Preferably, the isolation forest algorithm is used to detect the abnormal deviation between the blood pressure values measured in the hospital and those measured at home to identify white coat hypertension; after analyzing the abnormal deviation between the blood pressure values measured in the hospital and those measured at home in the historical hypertension characteristics of the patient, a blood pressure deviation index is generated. The method for obtaining the blood pressure deviation index is: Calculate the average blood pressure of the hospital and home: ; Where represents the i-th blood pressure measurement value in the hospital, represents the j-th blood pressure measurement value at home; Define the systolic blood pressure deviation index and the diastolic blood pressure deviation index , and the expression is: ; ; is the average systolic blood pressure measured in the hospital, The average systolic blood pressure measured at home; The average diastolic blood pressure measured at the hospital; The average diastolic blood pressure measured at home; The calculated systolic blood pressure deviation index and the diastolic blood pressure deviation index are weighted and averaged to calculate the blood pressure deviation index.

[0014] Preferably, after analyzing the measurement environment characteristics of the patient, a measurement emotional interference index is generated. The method for obtaining the measurement emotional interference index is as follows:

[0015] Collect and encode the measurement environment characteristics, including the time, ambient noise level, and measurement posture data during the past N blood pressure measurements; Calculate the feature vector as the benchmark data set of the measurement environment, where n is the number of features, X is the feature vector, and R is the feature data set; Use the variational autoencoder VAE to train the data distribution: Learn the latent distribution P(X) of the normal blood pressure measurement environment through VAE; If the likelihood probability P(Xcurrent) of the current measurement environment characteristics is less than the training distribution, it is determined that there is emotional interference; Use One-Class SVM for anomaly detection: Calculate the deviation of the current measurement environment characteristics Xcurrent in the latent space generated by VAE; Use One-Class SVM to calculate the anomaly score S: ; where μ and σ are the mean and standard deviation of the normal measurement environment generated by VAE respectively; If is greater than the anomaly threshold, calculate the measurement emotional interference index MEII, and the expression is: ; where the Sigmoid normalization index ranges from 0 to 1.

[0016] Preferably, convert the blood pressure deviation index and the measurement emotional interference index into a comprehensive feature vector, use the comprehensive feature vector as the input of the machine learning model, take predicting the hypertension risk analysis value label for each group of comprehensive feature vectors as the prediction target, and take minimizing the sum of the prediction errors for all hypertension risk analysis value labels as the training target, train the machine learning model until the sum of the prediction errors reaches convergence and then stop the model training, and determine the hypertension risk analysis value according to the model output result. Among them, the machine learning model is a polynomial regression model.

[0017] Preferably, compare the obtained hypertension risk analysis value with the gradient risk threshold. The gradient risk threshold includes a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold. Compare the hypertension risk analysis value with the first risk threshold and the second risk threshold respectively;

[0018] If the hypertension risk analysis value is greater than the second risk threshold, it is marked as a high-risk level; if the hypertension risk analysis value is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, it is marked as a medium-risk level; if the hypertension risk analysis value is less than the first risk threshold, it is marked as a low-risk level.

[0019] Preferably, for hypertensive patients at the medium-risk level, obtain the hypertension risk analysis value L within a fixed time period, and dynamically adjust the weights of the machine learning model in combination with the individual characteristics of different users, including: using the hypertension risk analysis value and the individual characteristics of different users as input items of fuzzy logic, and using the weights of the machine learning model as output items of fuzzy logic; formulating fuzzy rules to describe the influence degree of the hypertension risk analysis value and the individual characteristics of different users on the weights of the machine learning model, and dynamically adjusting the weights of the machine learning model according to the fuzzy logic reasoning results.

[0020] Preferably, if the individual characteristics of the user show a high-risk trend of hypertension, increase the sensitivity of the model to the hypertension risk analysis value; if the individual characteristics of the user show a low-risk trend, reduce the weight of the model to reduce misjudgment of short-term blood pressure fluctuations.

[0021] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0022] The present invention realizes the accurate assessment and personalized management of hypertensive patients through the combination of interactive data collection, time series analysis, machine learning risk prediction, and dynamic consultation guidance. Compared with the traditional hypertension risk assessment method based on static blood pressure values, the present invention can effectively identify white coat hypertension and masked hypertension, and reduce misdiagnosis and missed diagnosis caused by emotional tension or measurement environment. By calculating the blood pressure deviation index and the measurement emotion interference index, the system can integrate static and dynamic blood pressure data to improve the accuracy of hypertension risk prediction. Based on fuzzy logic reasoning, the system can combine individual characteristics (such as age, BMI, family medical history, etc.) to dynamically adjust the weights of the machine learning model and optimize the hypertension risk assessment. Through dynamic consultation guidance, the system can automatically adjust the questioning content according to the user's measurement data, generate a personalized blood pressure management plan, and provide accurate health intervention suggestions. The present invention improves the scientificity and reliability of hypertension detection, reduces unnecessary medical interventions, and at the same time optimizes the patient's health management experience, and has broad clinical application value. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0024] Figure 1 This is the system module diagram of the present invention. Detailed implementation manners

[0025] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0026] For the embodiments, please refer to Figure 1 As shown, a consultation guidance system based on interactive data collection in this embodiment includes a data acquisition module, a hypertension feature extraction module, a hypertension risk prediction module, and a guidance adjustment module;

[0027] Data acquisition module: Obtain the real-time health data of the user, including blood pressure measurement data, heart rate, blood oxygen saturation, and symptom information independently input by the user;

[0028] Hypertension feature extraction module: Combine time series analysis methods to calculate the blood pressure change trend, distinguish short-term fluctuations from long-term trends, and use anomaly detection algorithms to identify the characteristics of white coat hypertension and masked hypertension;

[0029] Hypertension risk prediction module: Use a machine learning model that integrates static and dynamic blood pressure data, combine the historical hypertension characteristics of the patient and the characteristics of the measurement environment to predict the hypertension risk, and dynamically adjust the weights of the machine learning model in combination with the individual characteristics of different users;

[0030] Guidance adjustment module: After the user uploads the blood pressure data, the system automatically adjusts the question content based on the measurement data through dynamic consultation guidance. If suspected white coat hypertension or masked hypertension is detected, the system will remind the user to perform multi-period blood pressure measurements and generate a personalized blood pressure management plan.

[0031] The data acquisition module supports multiple data input methods, including: intelligent wearable devices (such as smart watches, sphygmomanometers): used to measure physiological parameters such as blood pressure, heart rate, and blood oxygen saturation; intelligent medical devices (such as electronic sphygmomanometers, blood glucose meters): transmit measurement data through wireless or wired connections; user independent input (such as text, voice interaction): the user can input their own symptoms through mobile applications, mini-programs or web pages, such as "dizziness", "chest tightness", etc.

[0032] When the user measures blood pressure, heart rate, or blood oxygen saturation, the data acquisition module automatically identifies new data and receives the data through Bluetooth, Wi-Fi, or API interfaces; the user can manually input subjective symptoms through the consultation interface, such as "continuous headache and elevated blood pressure in the recent few days".

[0033] Unify the units of the collected physiological data (such as converting the blood pressure value from mmHg uniformly, normalizing the heart rate in bpm); use a filtering algorithm to remove measurement errors, such as abnormal values caused by incorrect arm position of the sphygmomanometer; combine the measurement time and the user's activity status to label the data background, such as "higher heart rate after exercise" as a normal phenomenon.

[0034] The collected data is stored in the local database and encrypted for transmission to the cloud health record; an authentication mechanism is used to ensure the privacy of the user's health data, such as using biometric login to prevent unauthorized access.

[0035] Case 1: Abnormality detection based on continuous blood pressure monitoring: The user wears a smart watch for 24-hour blood pressure monitoring, and the data acquisition module automatically collects the blood pressure change trend. After the system detects "abnormally elevated blood pressure at night", the data is uploaded to the health record, triggering the intelligent consultation module to ask the user whether there are related symptoms such as insomnia and anxiety.

[0036] Case 2: Intelligent interaction based on user's independent input: The user inputs "continuous dizziness in the recent three days", and the data acquisition module combines the user's past blood pressure measurement data and finds that the recent blood pressure fluctuates greatly. Trigger a personalized consultation process to ask the user about diet and exercise to determine whether the blood pressure abnormality is caused by lifestyle.

[0037] Data sources include: Static blood pressure data: single blood pressure measurement values from devices such as electronic sphygmomanometers and smart watches; Ambulatory blood pressure monitoring (ABPM) data: sequential blood pressure data obtained through 24-hour ambulatory blood pressure devices; User manual input: multiple blood pressure measurement values recorded by the user, such as morning and evening measurement data; Physiological and environmental data: factors affecting blood pressure including heart rate, exercise status, emotional state (such as anxiety), medication status, etc.

[0038] Use moving average filtering to smooth short-term fluctuations and remove measurement noise; calculate the blood pressure change trend and distinguish the following patterns:

[0039] Short-term fluctuations (caused by emotions and measurement environment, such as elevated blood pressure due to nervousness during measurement); Long-term trend (judging whether there is a continuous upward trend based on 7-day or 30-day data). Construct a blood pressure diurnal rhythm model to detect morning peak hypertension. If the morning systolic blood pressure is more than 20 mmHg higher than at night, it indicates the risk of morning peak hypertension.

[0040] White Coat Hypertension (WCH) identification, feature extraction: The blood pressure measured in the medical environment is high (≥140 / 90 mmHg), but the home blood pressure or ambulatory blood pressure monitoring is normal (≤130 / 80 mmHg); combined with the user's heart rate data, detect the state of tension during measurement (such as abnormal increase in heart rate > 100 bpm); statistically analyze the difference in blood pressure measurements in different environments. If the hospital measurement value is consistently higher than the home measurement value by more than 15 mmHg, it is marked as suspected white coat hypertension.

[0041] Use the Isolation Forest algorithm to detect abnormal deviations between hospital blood pressure values and home blood pressure values; use the DBSCAN clustering algorithm to divide the user's blood pressure data into "hospital measurement group" and "home measurement group", calculate the mean difference between the two. If the difference is significant, it is determined as white coat hypertension.

[0042] Masked Hypertension (MH) identification, feature extraction: The blood pressure measured in the hospital is normal (≤130 / 80 mmHg), but the home blood pressure or 24-hour ambulatory blood pressure monitoring is higher than 140 / 90 mmHg; combined with the long-term blood pressure change trend, detect the phenomenon of high blood pressure in the evening (nighttime blood pressure > 125 / 75 mmHg); correlate with the exercise state. If the blood pressure is high after exercise and the recovery is slow, it may indicate masked hypertension.

[0043] Use a time series anomaly detection algorithm (such as the Holt-Winters model) to analyze whether the nighttime blood pressure data shows abnormal increase; calculate the ratio of the user's nighttime blood pressure to daytime blood pressure (Nocturnal BP Ratio). If the ratio > 1, there may be a risk of masked hypertension.

[0044] Through time series analysis and anomaly detection algorithms, the system can effectively identify white coat hypertension and masked hypertension, reduce the misdiagnosis rate; combined with the user's personalized blood pressure change trend, improve the accuracy of the intelligent consultation system, and provide a more accurate health management plan for users.

[0045] Hypertension risk prediction module: Use a machine learning model that integrates static and dynamic blood pressure data, combined with the patient's historical hypertension characteristics and measurement environment characteristics, to predict hypertension risk;

[0046] After analyzing the abnormal deviation between the patient's hospital blood pressure value and home blood pressure value in the patient's historical hypertension characteristics, generate a blood pressure deviation index. The method for obtaining the blood pressure deviation index is: calculate the average blood pressure of the hospital and home: ; Among them, represents the i-th hospital blood pressure measurement value, represents the j-th home blood pressure measurement value; define the systolic blood pressure deviation index and diastolic blood pressure deviation index , the expression is: ; ; is the average systolic blood pressure measured in the hospital, is the average systolic blood pressure measured at home; is the average diastolic blood pressure measured in the hospital; is the average diastolic blood pressure measured at home; The calculated systolic blood pressure deviation index and diastolic blood pressure deviation index are weighted and averaged to calculate the blood pressure deviation index.

[0047] Measurement environment characteristics: including measurement time (morning, night), measurement posture (sitting, lying), emotional state (tense, relaxed), exercise state (resting, after exercise), etc.

[0048] After analyzing the measurement environment characteristics of the patient, a measurement emotional interference index is generated. The acquisition method of the measurement emotional interference index is:

[0049] When the historical blood pressure measurement environment characteristics of the user are relatively stable, a low-dimensional feature representation of the measurement environment can be generated using VAE, and then One-Class SVM is used to detect abnormal emotional states, such as tension or anxiety.

[0050] Collect and encode measurement environment characteristics, including time (morning, night), environmental noise level (dB), and measurement posture (sitting, standing, lying) data during the past N blood pressure measurements;

[0051] Calculate the feature vector as the benchmark data set of the measurement environment, n is the number of features, X is the feature vector, and R is the feature data set; Use the variational autoencoder VAE to train the data distribution: learn the latent distribution P(X) of the normal blood pressure measurement environment through VAE; If the likelihood probability P(Xcurrent) of the current measurement environment characteristics is less than the training distribution, it is judged that there is emotional interference; Use One-Class SVM for anomaly detection: calculate the deviation of the current measurement environment characteristics Xcurrent in the latent space generated by VAE; Use One-Class SVM to calculate the anomaly score S: ; where μ and σ are the mean and standard deviation of the normal measurement environment generated by VAE respectively; If > θ (abnormal threshold), it indicates that the measurement emotional interference is relatively large, and calculate the measurement emotional interference index MEII, the expression is: ; where, the Sigmoid normalization index is between 0-1.

[0052] Convert the blood pressure deviation index and the measured emotional interference index into a comprehensive feature vector, and use the comprehensive feature vector as the input of a machine learning model. The machine learning model takes predicting the label of the hypertension risk analysis value for each group of comprehensive feature vectors as the prediction target, and takes minimizing the sum of the prediction errors for all hypertension risk analysis value labels as the training target. Train the machine learning model until the sum of the prediction errors reaches convergence and then stop the model training. Determine the hypertension risk analysis value according to the model output result, where the machine learning model is a polynomial regression model.

[0053] The method for obtaining the hypertension risk analysis value is as follows: Obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: ; In the formula, is the output function of the model, HK is the blood pressure deviation index, MSII is the measured emotional interference index, is the hypertension risk analysis value.

[0054] Compare the obtained hypertension risk analysis value with the gradient risk thresholds. The gradient risk thresholds include a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold. Compare the hypertension risk analysis value with the first risk threshold and the second risk threshold respectively;

[0055] If the hypertension risk analysis value is greater than the second risk threshold, it indicates that the user has a high hypertension risk. It is recommended to seek medical attention in a timely manner and conduct 24-hour ambulatory blood pressure monitoring (ABPM); drug intervention may be required, and lifestyle adjustments (such as low-salt diet, regular exercise, etc.) should be made; mark it as a high-risk level;

[0056] If the hypertension risk analysis value is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, it indicates that the user has a moderate hypertension risk. It is recommended to continue monitoring and take appropriate health management measures (such as regularly measuring blood pressure, adjusting diet and work and rest, etc.); mark it as a moderate-risk level;

[0057] If the hypertension risk analysis value is less than the first risk threshold, it indicates that the user has a low hypertension risk. It is recommended to maintain good living habits and regularly measure blood pressure for health monitoring; mark it as a low-risk level.

[0058] For hypertension patients at the moderate-risk level, obtain the hypertension risk analysis value L within a fixed time period, and dynamically adjust the weights of the machine learning model in combination with the individual characteristics of different users.

[0059] Among them, individual characteristics include: Age: The older the age, the higher the risk of hypertension; BMI (Body Mass Index): Individuals with a higher BMI are more likely to develop hypertension; Family History: If there is a family history of hypertension, the risk weight increases; Exercise Level: If the user has a low exercise frequency, a higher prediction weight may be required; Dietary Habits: For example, a high-salt diet may lead to higher blood pressure fluctuations.

[0060] Take the hypertension risk analysis value and the individual characteristics of different users as the input items of fuzzy logic, and take the weights of the machine learning model as the output items of fuzzy logic;

[0061] If the individual characteristics of the patient show a high-risk trend (such as older age, higher BMI), increase the sensitivity of the model to the hypertension risk analysis value (L), making it easier to predict the possible future hypertension escalation trend;

[0062] If the individual characteristics of the patient show a low-risk trend (such as regular exercise, long-term stable blood pressure), reduce the weight of the model to make it insensitive to short-term fluctuations to avoid misjudgment.

[0063] The fuzzy logic inference system infers the input information through a fuzzy rule set to determine the final weight adjustment plan of the model.

[0064] First, fuzzify the input items. For example, the hypertension risk analysis value (L) can be divided into "lower", "medium", "higher"; age can be divided into "young", "middle-aged", "elderly"; BMI can be divided into "normal", "higher", "obese"; exercise habits can be divided into "regular", "occasional", "almost none".

[0065] Based on the expert rules of fuzzy logic, the following inference rules can be set:

[0066] Rule 1: If the hypertension risk analysis value is higher and the patient is older, increase the attention weight of the model to the long-term trend data.

[0067] Rule 2: If the hypertension risk analysis value is medium, but the patient has a higher BMI and almost no exercise, increase the sensitivity of the model to short-term blood pressure fluctuations in order to detect potential high risks earlier.

[0068] Rule 3: If the hypertension risk analysis value is medium, but the patient exercises regularly and has no family history of hypertension, reduce the weight of short-term fluctuations to avoid misjudgment.

[0069] Rule 4: If the patient is taking medicine and the blood pressure fluctuates slightly, the impact on the single hypertension analysis value is reduced, and the long-term monitoring of the medicine-taking effect is increased.

[0070] After fuzzy inference, the system outputs an adjusted model weight parameter and applies it to the machine learning model to dynamically optimize the hypertension risk prediction.

[0071] Dynamically adjust the weights of the machine learning model through fuzzy logic so that it can adapt to the personalized needs of different patients and improve the accuracy of hypertension risk prediction.

[0072] Reduce misdiagnosis and missed diagnosis. For example, for the increase in hypertension risk value caused by short-term mood fluctuations, the system will not intervene excessively, while for users with higher long-term risks, the system will be more sensitive.

[0073] Enhance the adaptability of the prediction, and provide more accurate health management suggestions for users with different lifestyles and age groups.

[0074] After the user uploads blood pressure data, the system uses a dynamic consultation guidance mechanism to automatically adjust the consultation content according to the characteristics of the user's measurement data. If suspected white coat hypertension (WCH) or masked hypertension (MH) is detected, the system will provide targeted measurement suggestions and generate a personalized blood pressure management plan to help users more accurately evaluate their blood pressure status and optimize their health management strategies.

[0075] After the user uploads blood pressure measurement data, the system will automatically analyze the blood pressure trend, measurement environment, and historical blood pressure data. If abnormal blood pressure fluctuations are detected (such as high blood pressure measured in the hospital and low blood pressure measured at home, or abnormal increase in blood pressure at night), the system will trigger intelligent consultation guidance and dynamically adjust the question content to further collect relevant information.

[0076] The system will dynamically adjust the question content according to the characteristics of white coat hypertension and masked hypertension. For example, if suspected white coat hypertension (WCH) (high blood pressure in the hospital and normal blood pressure at home) is detected: "Do you feel nervous or anxious when measuring blood pressure in the hospital?" "Have you ever measured your blood pressure at home? If so, please provide your home blood pressure data." "Is your heart rate higher than usual when measuring blood pressure in the hospital?" "Have you experienced stress, exercise, or mood fluctuations before measuring blood pressure?" If suspected masked hypertension (MH) (normal blood pressure in the hospital and high blood pressure at home) is detected: "Has your home blood pressure been higher than the blood pressure measured in the hospital for a long time?" "Has your blood pressure been recorded as high at night?" "Do you have a family history of hypertension? Are there any chronic diseases (such as diabetes, kidney disease)?" "Is your daily diet salty? Do you consume high-sodium foods for a long time?"

[0077] If a user is suspected of having white coat hypertension or masked hypertension, the system will recommend that they measure their blood pressure multiple times at different times to more comprehensively evaluate their blood pressure situation.

[0078] Recommendations for users with white coat hypertension (WCH): Use the home self - measurement blood pressure method to measure blood pressure at fixed times in the morning, afternoon, and evening every day for 1 - 2 consecutive weeks; it is recommended to use 24 - hour ambulatory blood pressure monitoring (ABPM) to exclude the influence of the white coat effect; perform deep breathing relaxation training before measurement to avoid increased blood pressure caused by tension.

[0079] Recommendations for users with masked hypertension (MH): Increase nighttime blood pressure measurement, and it is recommended to measure blood pressure once before going to bed and once after getting up, continuously monitor for 7 days; pay attention to morning peak hypertension, and it is recommended to measure blood pressure within 30 minutes after getting up in the morning; perform blood pressure monitoring after exercise and record the blood pressure changes before and after exercise to determine whether there is abnormal elevation.

[0080] Based on the user's measurement data and response situation, the system will generate personalized lifestyle adjustment recommendations. For example, if a user has too high sodium intake (such as often eating pickled foods and processed foods), it is recommended to reduce salt intake, with the daily salt intake not exceeding 5g; it is recommended to increase potassium intake (such as eating more bananas, spinach, beans, etc.), which helps to lower blood pressure.

[0081] If a user has less physical activity, it is recommended to perform moderate - intensity aerobic exercise (such as brisk walking, swimming, cycling) at least 5 days a week, 30 minutes each time; for users with large blood pressure fluctuations, avoid high - intensity exercise to prevent sudden increase in blood pressure.

[0082] If a user is emotionally tense, the system will recommend relaxation methods such as meditation, deep breathing exercises, and yoga; for users with high stress, the system will recommend using emotion management applications and combining heart rate variability (HRV) analysis to provide stress relief solutions.

[0083] If a user's blood pressure is extremely abnormal or has not improved for a long time, the system will recommend that the user contact a doctor for remote consultation; the system can interface with hospital data and provide the doctor with the user's multi - time blood pressure monitoring report to help the doctor make a more accurate diagnosis; for users at medium risk, the system can set regular follow - up reminders, such as performing a blood pressure assessment once a month and adjusting the management plan according to the trend.

[0084] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by technicians in this field according to the actual situation.

[0085] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0086] As described above, the specific implementation manners of the present application are only provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application.

Claims

1. An interrogation guidance system based on interactive data collection, characterized by: It includes a data acquisition module, a hypertension feature extraction module, a hypertension risk prediction module, and a guidance and adjustment module; Data acquisition module: Obtain the real-time health data of the user, including blood pressure measurement data, heart rate, blood oxygen saturation, and symptom information independently input by the user; Hypertension feature extraction module: Combine time series analysis methods to calculate the blood pressure change trend, distinguish short-term fluctuations from long-term trends, and use anomaly detection algorithms to identify the characteristics of white coat hypertension and masked hypertension; Hypertension risk prediction module: Use a machine learning model that integrates static and dynamic blood pressure data, combine the historical hypertension characteristics of the patient and the characteristics of the measurement environment to predict the hypertension risk, and dynamically adjust the weights of the machine learning model according to the individual characteristics of different users; Guidance and adjustment module: After the user uploads the blood pressure data, the system automatically adjusts the question content based on the measurement data through dynamic inquiry guidance. If suspected white coat hypertension or masked hypertension is detected, the system will remind the user to perform multi-period blood pressure measurements and generate a personalized blood pressure management plan; In the data acquisition module, blood pressure, heart rate, and blood oxygen saturation are collected through intelligent wearable devices; measurement data is transmitted through intelligent medical devices; the user's chief complaint symptoms are collected through the user's independent input and correlated with physiological parameters for analysis; The isolation forest algorithm is used to detect the abnormal deviation between the blood pressure values measured in the hospital and the blood pressure values measured at home to identify white coat hypertension; after analyzing the abnormal deviation between the blood pressure values of the patient in the hospital and at home in the patient's historical hypertension characteristics, a blood pressure deviation index is generated. The method for obtaining the blood pressure deviation index is: Calculate the average blood pressure in the hospital and at home: Among them, BP avgclinic represents the i-th hospital blood pressure measurement value, and BP avghome represents the j-th home blood pressure measurement value; define the systolic blood pressure deviation index BPDI sys and the diastolic blood pressure deviation index BPDI dia , and the expression is: BP augclinic,sys is the average systolic blood pressure measured in the hospital, and BP aughome,sys is the average systolic blood pressure measured at home; BP avgclinic,dia is the average diastolic blood pressure measured in the hospital; BP avghome,dia is the average diastolic blood pressure measured at home; the calculated systolic blood pressure deviation index BPDI sys and the diastolic blood pressure deviation index BPDI dia are weighted and averaged and summed to obtain the blood pressure deviation index; After analyzing the characteristics of the patient's measurement environment, a measurement emotional interference index is generated. The method for obtaining the measurement emotional interference index is: Collect and encode the characteristics of the measurement environment, including the time, ambient noise level, and measurement posture data during the past N blood pressure measurements; calculate the feature vector X ∈ R n As the benchmark data set of the measurement environment, n is the number of features, X is the feature vector, and R is the feature data set; use the variational autoencoder VAE to train the data distribution: learn the latent distribution P(X) of the normal blood pressure measurement environment through VAE; if the likelihood probability P(Xcurrent) of the current measurement environment characteristics is less than the training distribution, it is determined that there is emotional interference; use One-Class SVM for anomaly detection: calculate the deviation of the current measurement environment characteristics Xcurrent in the latent space generated by VAE; use One-Class SVM to calculate the anomaly score S: where μ and σ are the mean and standard deviation of the normal measurement environment generated by VAE respectively; if S is greater than the anomaly threshold, calculate the measurement emotional interference index MEII, and the expression is: MEII = Sigmoid(S); where the Sigmoid normalizes the index to between 0 and 1; Convert the blood pressure deviation index and the measurement emotional interference index into a comprehensive feature vector, use the comprehensive feature vector as the input of the machine learning model, use the machine learning model to predict the hypertension risk analysis value label for each group of comprehensive feature vectors as the prediction target, and use minimizing the sum of the prediction errors for all hypertension risk analysis value labels as the training target to train the machine learning model until the sum of the prediction errors reaches convergence and stop the model training. Determine the hypertension risk analysis value according to the model output result. Among them, the machine learning model is a polynomial regression model.

2. The inquiry guidance system based on interactive data collection according to claim 1, wherein: Compare the obtained hypertension risk analysis value with the gradient risk threshold. The gradient risk threshold includes a first risk threshold and a second risk threshold, and the first risk threshold is less than the second risk threshold. Compare the hypertension risk analysis value with the first risk threshold and the second risk threshold respectively; If the hypertension risk analysis value is greater than the second risk threshold, mark it as a high risk level; If the hypertension risk analysis value is greater than or equal to the first risk threshold and less than or equal to the second risk threshold, mark it as a medium risk level; If the hypertension risk analysis value is less than the first risk threshold, mark it as a low risk level.

3. The inquiry guidance system based on interactive data collection according to claim 2, wherein: For hypertensive patients at a medium risk level, obtain the hypertension risk analysis value L within a fixed time period, and dynamically adjust the weights of the machine learning model in combination with the individual characteristics of different users, including: using the hypertension risk analysis value and the individual characteristics of different users as input items of fuzzy logic, and using the weights of the machine learning model as output items of fuzzy logic; formulating fuzzy rules to describe the influence degree of the hypertension risk analysis value and the individual characteristics of different users on the weights of the machine learning model, and dynamically adjusting the weights of the machine learning model according to the fuzzy logic reasoning results.

4. The inquiry guidance system based on interactive data collection according to claim 3, characterized in that: If the individual characteristics of the user show a high-risk trend of hypertension, increase the sensitivity of the model to the hypertension risk analysis value; if the individual characteristics of the user show a low-risk trend, reduce the weight of the model to reduce the misjudgment of short-term blood pressure fluctuations.

Images