Medical environment analysis system based on deep learning and Internet of Things technology

Abstract

The invention discloses a medical environment analysis system based on deep learning and the Internet of Things technology, and relates to the technical field of medical environment analysis, after behavior mode data of an old user is identified, if the similarity between the behavior mode data and a target behavior mode is lower than expectation, the collection frequency of life state data is constrained according to the similarity; collecting real-time health index data of the elderly user, and if the health state index data is abnormal and the abnormal degree exceeds the expectation, outputting activity state data of the elderly in real time; identifying the health score of the elderly, sending out a multi-level alarm to the outside when the health score is lower than the expectation, and generating a comprehensive health report; and according to the relationship between the health score and the health threshold value, a corresponding processing mode is adopted for the elderly user in the non-health state. Through a multi-level alarm mechanism, when the health state of the elderly user is abnormal, processing can be carried out in time, and omission of the abnormal condition is avoided.

Description

Technical Field

[0001] The present invention relates to the field of medical environment analysis technology, and specifically to a medical environment analysis system based on deep learning and Internet of Things technologies. Background Art

[0002] The analysis of the medical environment for the elderly uses technical means to conduct a comprehensive assessment of the medical system and related factors, with special attention to the special needs and challenges of the elderly. Through data collection and analysis, the barrier-free design of medical facilities, the ease of use of medical equipment, the geriatric expertise of medical staff, the adaptability of nursing processes, and the satisfaction and experience of the elderly with medical services are evaluated. The warning value model used in the technical solution can be used to monitor the health status of elderly patients in real time and timely warn of potential health risks. By reasonably setting the warning weight parameters, especially considering factors such as common chronic disease management, drug use, mental health support and social relationships among the elderly, the warning value can be dynamically adjusted to provide personalized and timely medical intervention. Finally, by combining these technical analyses and situational assessments, a scientific basis is provided for improving the elderly-friendly medical environment and formulating relevant policies, thereby improving the health and quality of life of the elderly.

[0003] The target population of elderly health management is mainly the elderly aged 65 and above, including the elderly who have lived in the community for more than half a year. Regardless of their household registration status, they can enjoy corresponding health management services in the township health center, village clinic or community health service center where they live. Elderly health management is an important public health service, which is of great significance to improving the quality of life of the elderly and reducing the disability rate and mortality rate. By providing comprehensive health management services for the elderly, potential health problems can be discovered early, and preventive measures and treatment measures can be taken in time, thereby ensuring the physical and mental health of the elderly.

[0004] In the Chinese invention patent with the authorization announcement number CN113643808B, a method for realizing health management of the elderly by software is disclosed. The user information collection module of the client collects basic information of the user and issues a health assessment request; the health assessment module performs health assessment according to the user information, returns the health risk assessment results and self-care ability assessment report to the display module of the client for result display, and sends it to the database for storage at the same time; the health assessment module matches the health resource package and sports health video according to the health risk assessment results and self-care ability level of the user and gives reminders. Health assessment is performed according to the basic health status data and self-care ability of the elderly, and sports videos and health resource packages are recommended in a targeted manner according to the assessment results, reminding them to take medicine and exercise on time. Only basic information needs to be entered to obtain health resources and exercise programs suitable for them.

[0005] Combining the above applications and the prior art:

[0006] When the elderly live alone, due to the lack of appropriate care, guardians or doctors may find it difficult to understand the elderly's physical health status in a timely manner. Even when the elderly's health status becomes abnormal, they cannot be diagnosed and treated in time, which can easily lead to further deterioration of the health status of elderly users in unhealthy states.

[0007] In the existing medical environment analysis system, for elderly users living alone, various health indicator data are usually collected in real time. By identifying their health indicator data, an alarm is issued to the outside when health indicator data, such as blood pressure, blood lipids, etc., are abnormal, so as to play a role of reminder or prediction;

[0008] However, in the activity area, the living conditions of elderly users may be abnormal to a certain extent, such as less activity, or even when the external environmental conditions are abnormal or mutated, the health status of the elderly will be greatly affected, but this degree of influence may take some time to manifest. Therefore, although the identification results of health indicator data can reflect the health status of the elderly, there may be a certain lag and it is relatively one-sided. When external intervention is needed, it cannot play an effective guiding role.

[0009] To this end, the present invention provides a medical environment analysis system based on deep learning and Internet of Things technology. Summary of the invention

[0010] 1. Technical issues to be resolved

[0011] In view of the deficiencies of the prior art, the present invention provides a medical environment analysis system based on deep learning and Internet of Things technology. After identifying the behavior pattern data of elderly users, if the similarity between the behavior pattern data and the target behavior pattern is lower than expected, the collection frequency of life status data is constrained according to the similarity; real-time health index data of elderly users are collected, and if the health status index data is abnormal and the abnormality exceeds expectations, the activity status data of the elderly is output in real time; the health score of the elderly is identified, and a multi-level alarm is issued to the outside when the health score is lower than expected, and a comprehensive health report is generated; according to the relationship between the health score and the health threshold, a corresponding treatment method is adopted for elderly users in an unhealthy state. Through the multi-level alarm mechanism, when the health status of elderly users is abnormal, they can be processed in time to avoid omission of abnormal situations, thereby solving the technical problems recorded in the background technology.

[0012] (II) Technical solution

[0013] To achieve the above objectives, the present invention is implemented through the following technical solutions: a medical environment analysis system based on deep learning and Internet of Things technology, including a data analysis unit, which collects the life status data of elderly users, and if there is an abnormality in the life status data, generates a data abnormality degree from the abnormal recognition data, and issues a behavior pattern recognition instruction when the data abnormality degree exceeds expectations;

[0014] A constraint unit, after identifying the behavior pattern data of the elderly user, if the similarity between the behavior pattern data and the target behavior pattern is lower than expected, constrains the collection frequency of the life status data according to the similarity;

[0015] The monitoring unit collects real-time health indicator data of elderly users. If the health status indicator data is abnormal and the degree of abnormality exceeds expectations, the activity status data of the elderly is output in real time;

[0016] The scoring unit identifies the health score of the elderly, issues a multi-level alarm to the outside world when the health score is lower than expected, and generates a comprehensive health report;

[0017] The solution output unit takes corresponding measures for elderly users in unhealthy conditions based on the relationship between the health score and the health threshold.

[0018] Furthermore, data is collected and detected in real time in the activity area of ​​the elderly user to obtain corresponding life status data, which is aggregated and generated into a life status data summary after preprocessing;

[0019] Taking real-time life status data as input, the trained abnormal data recognition model is used to perform abnormality recognition, obtain the abnormal data and the corresponding abnormal proportion, and thus generate the data abnormality degree.

[0020] Further, after receiving the behavior pattern recognition instruction, the user's daily activity data is collected in the living area and aggregated to generate a daily behavior data set;

[0021] Taking the user's daily behavior data as input, the trained behavior pattern recognition model is used to perform behavior pattern recognition to obtain the user's normal behavior pattern, abnormal behavior pattern and the distribution time of different behavior patterns, and then aggregate them to generate the user's behavior pattern data set.

[0022] Furthermore, a similarity analysis is performed between the user's behavior pattern data set and the expected behavior pattern data to obtain corresponding similarity data, and the user's life status data is collected based on the collection frequency that meets the constraint conditions.

[0023] Furthermore, the real-time health indicator data is used as input, and the trained abnormal indicator recognition model is used to identify abnormalities in the user's physical health indicators; after screening out the abnormal health indicators, the corresponding abnormal degree and the abnormal nodes that generate the abnormality are confirmed; and after aggregation, an abnormal indicator data set is generated.

[0024] Furthermore, the health abnormality degree of the user's health indicator is generated from the abnormal indicator data set. If the abnormality degree exceeds the health threshold, a health trend analysis instruction is issued to the outside.

[0025] Further, after receiving the health trend analysis instruction, the health indicator data of the user in several sub-periods is continuously obtained, the health trend analysis is performed on the user, and the corresponding health trend data is obtained;

[0026] The user's health indicator data, as well as the user's posture data, activity frequency and real-time location data are collected and aggregated and sent out as user activity status data.

[0027] Furthermore, the health indicator data is used as input to perform a health score on the user using the trained user health score model; if the user's health score is lower than expected, the user is treated as an unhealthy elderly user and a first-level alarm instruction is issued to the outside.

[0028] Furthermore, if the number of consecutive first-level alarm instructions received within a preset observation period exceeds expectations, a warning value is constructed based on the score gap between the health score and the target score, as well as the distribution state of the time nodes when the first-level alarm instruction is received each time;

[0029] If the acquired warning value exceeds the warning threshold, a secondary alarm command is issued to the outside;

[0030] Summarize the user's health indicator data, activity status data and environmental parameters, and use heat maps to represent the user's high-frequency activity areas, and generate a comprehensive health report after aggregation.

[0031] Furthermore, if the user's health score exceeds the health threshold, no additional processing is performed, and the user's health indicator data is sent in real time;

[0032] If the user's health score is within the health threshold, the elderly user health management knowledge graph will output a personalized health management plan, such as dietary adjustment suggestions and exercise plans;

[0033] If the user's health score is below the health threshold, a video consultation will be conducted on the user, and the comprehensive health report will be shared on the video consultation platform.

[0034] (III) Beneficial effects

[0035] The present invention provides a medical environment analysis system based on deep learning and Internet of Things technology, which has the following beneficial effects:

[0036] 1. Judge the abnormality of various environmental conditions and user activity data in the activity area based on the data abnormality. When the abnormality is high, timely adjustments and processing can be made to ensure the health status of elderly users.

[0037] 2. By identifying the behavior patterns of elderly users, verify whether there are any abnormalities in the user's current behavior patterns. If the proportion of abnormal user behavior patterns is high, the user's health status can be pre-judged and evaluated, reflecting the life and health status of elderly people living alone in real time.

[0038] 4. After constraining the frequency of life status collection by similarity, the amount of data collected is matched with the user's current actual status, avoiding omissions when data is collected and improving reliability when analyzing user status.

[0039] 5. The user's current health status can be judged and evaluated based on the degree of health abnormality, and elderly users can be determined as healthy or unhealthy, and targeted treatment can be carried out; by real-time output of health indicator data, posture data, activity frequency and real-time location data, after receiving these data, elderly users can be treated in a timely manner.

[0040] 6. Use the trained user health scoring model to perform health scoring and achieve a comprehensive assessment of the health status of elderly users. Through the multi-level alarm mechanism constructed by the first-level alarm command and the second-level alarm command, when the health status of elderly users is abnormal, it can be handled in a timely manner to avoid missing abnormal situations.

[0041] 7. Take targeted measures for elderly users in unhealthy conditions and respond quickly to unhealthy conditions in medical scenarios within the current living area to avoid situations where elderly users do not care about their own health status, or know their own status but do not take any action. This ensures the health status of elderly users, especially those living alone, and the efficiency of medical intervention, and can also fully play a guiding role. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of the process of the medical environment analysis method of the present invention;

[0043] Figure 2 It is a schematic diagram of the medical environment analysis process structure of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] See also Figure 1 The present invention provides a medical environment analysis method based on deep learning and Internet of Things technology, including:

[0046] Step 1: Collect the living status data of elderly users in the activity area. If there is an abnormality in the living status data, generate data abnormality from the abnormal recognition data, and issue a behavior pattern recognition instruction when the data abnormality exceeds expectations;

[0047] The step 1 includes the following contents:

[0048] Step 101: When it is necessary to monitor and collect the life and health data of the elderly user, a data collection device, such as a health monitoring device, an environmental sensor, and an activity monitoring device, is placed in the activity area of ​​the elderly user; the data collection device is used to collect and detect data in real time to obtain corresponding life status data, such as the user's activity status data and environmental parameters in the activity area; the collected real-time life status data is pre-processed and aggregated to generate a life status data summary;

[0049] Step 102: train a convolutional neural network with the labeled sample data to obtain a trained abnormal data recognition model; use the real-time data in the life status data set as input, use the trained abnormal data recognition model to perform abnormal recognition, obtain the abnormal data and the corresponding abnormal ratio, and generate the data abnormality degree in the following manner:

[0050]

[0051] Where: t1 and t2 are the start and end times of the monitoring period, m is the total number of abnormal data, P i (t) is the abnormal proportion of the i-th abnormal data at time t, S i is the anomaly score of the ith anomaly data, and its value falls between [0,1]. w(t) is the weight function at time t, which indicates the weight of anomalies in different time periods and can be set according to actual conditions.

[0052] Pre-set abnormal thresholds based on historical data and expectations for abnormal management of life and health data;

[0053] If the data anomaly exceeds the anomaly threshold, it means that the frequency of data anomalies in the living area is also high, and the health status of the elderly user may also be abnormal. It is necessary to identify the current life and health pattern of the elderly user. At this time, a behavior pattern recognition instruction is issued to the outside;

[0054] When using, combine the contents in steps 101 and 102:

[0055] When elderly users are in the activity area, based on the collected data, the abnormal data is identified and obtained, the data abnormality in the activity area is judged, and it is verified whether these abnormal data will have a negative impact on the elderly. In particular, the data abnormality degree is generated on this basis, and the abnormal degree of the various environmental conditions in the activity area and the user's activity data is judged according to the data abnormality degree. When the abnormality degree is high, timely adjustments and processing can be made to ensure the health status of the elderly users.

[0056] Combining the above applications and the prior art:

[0057] When the elderly live alone, due to the lack of appropriate care, guardians or doctors may find it difficult to understand the elderly's physical health status in a timely manner. Even when the elderly's health status becomes abnormal, they cannot be diagnosed and treated in time, which can easily lead to further deterioration of the health status of elderly users in unhealthy states.

[0058] In the existing medical environment analysis system, for elderly users living alone, their various health indicator data are usually collected in real time. By identifying their health indicator data, an alarm is issued to the outside when health indicator data, such as blood pressure, blood lipids, etc., are abnormal, so as to serve as a reminder or prediction; however, in the activity area, the living conditions of elderly users have certain abnormalities, such as less activity, or even when the external environmental conditions are abnormal or mutated, the health status of the elderly will also be greatly affected, but this degree of influence may be manifested after a period of time. Therefore, although the identification results of health indicator data can reflect the health status of the elderly, there may be a certain hysteresis and it is relatively one-sided. When external intervention is needed, it cannot play an effective guiding role.

[0059] Step 2: After identifying the behavior pattern data of elderly users, if the similarity between the behavior pattern data and the target behavior pattern is lower than expected, the collection frequency of life status data is constrained according to the similarity;

[0060] The step 2 includes the following contents:

[0061] Step 201: After receiving the behavior pattern recognition instruction, collect the user's daily activity data in the living area, including the number of steps, exercise frequency, and stationary time, and generate a daily behavior data set after aggregation;

[0062] The deep learning algorithm is trained with the labeled sample data to obtain a trained behavior pattern recognition model;

[0063] Taking the user's daily behavior data as input, the trained behavior pattern recognition model is used to perform behavior pattern recognition, obtain the user's normal behavior pattern, abnormal behavior pattern and the distribution time of different behavior patterns, and generate the user's behavior pattern data set after aggregation;

[0064] When in use, on the basis of completing various data collection, the behavior patterns of elderly users are identified to verify whether there are any abnormalities in the user's current behavior patterns. If the proportion of abnormal user behavior patterns is high, it means that in the current activity area, the health status of the elderly may be abnormal to a certain extent, thereby achieving preliminary judgment and evaluation of the user's health status, reflecting the life and health status of elderly people living alone in real time.

[0065] Step 202: Perform similarity analysis on the user's behavior pattern data set and the expected behavior pattern data to obtain corresponding similarity data, and constrain the data collection frequency Cjp according to the similarity Xs, and collect the user's life status data according to the collection frequency that meets the constraint conditions, in the following manner:

[0066]

[0067] Among them, n is the number of similarities, Xs ij is the difference between the i-th similarity and the j-th similarity, Xs a is the average value of the difference; weight coefficient: 0≤S1≤1, 0≤S2≤1, and S1+S2=1; the weight coefficient can be obtained by referring to the hierarchical analysis method;

[0068] When using, combine the contents in steps 201 and 202:

[0069] After analyzing the user's behavior pattern, based on the similarity generated by the behavior pattern data and after constraining the frequency of life status collection by similarity, the data collection amount is matched with the user's current actual status to avoid omissions when data is collected and improve reliability when analyzing the user status.

[0070] Step 3: Collect real-time health indicator data of elderly users. If the health status indicator data is abnormal and the degree of abnormality exceeds expectations, output the activity status data of the elderly in real time;

[0071] The step three includes the following contents:

[0072] Step 301: training a convolutional neural network with the labeled sample data to obtain a trained abnormal indicator recognition model;

[0073] Using real-time health indicator data as input, the trained abnormal indicator recognition model is used to identify abnormalities in the user's physical health indicators; after screening out abnormal health indicators, the corresponding abnormal degree and abnormal nodes that produce abnormalities are confirmed; after aggregation, an abnormal indicator data set is generated;

[0074] Step 302: Generate the health abnormality degree of the user health indicator from the abnormal indicator data set in the following manner:

[0075]

[0076] In the formula, k is the total number of abnormal health indicators, f j (t) is the time distribution function, which represents the distribution state of the j-th abnormal health indicator in time. j (t) is the number of abnormal health indicators of the jth type at time t, S j is the severity of the j-th abnormal health indicator, and its value falls between [0,1]. w(t) is the weight function at time t, which indicates the weight of the abnormality in different time periods and can be set according to the actual situation.

[0077]

[0078] Where, t i is the ith abnormal node, n is the total number of abnormal nodes, h is the bandwidth, and its value falls within [0.1,1], and K is the Gaussian kernel function;

[0079] Based on historical data and management expectations of the patient's health status, health thresholds are pre-set; if the abnormality exceeds the health threshold, it means that the user's health status has certain abnormalities and needs to be processed in a timely manner. At this time, a health trend analysis instruction is issued to the outside;

[0080] When in use, after obtaining the patient's health indicator data in real time, the abnormal health indicators are identified and marked, and then the health abnormality degree is generated. Based on the health abnormality degree, the user's current health status can be judged and evaluated, and the elderly users can be determined as healthy or unhealthy, so that targeted treatment can be carried out on this basis.

[0081] Step 303: after receiving the health trend analysis instruction, continuously obtain the health indicator data of the user in several sub-periods, perform health trend analysis on the user, and obtain corresponding health trend data;

[0082] Collect and obtain the user's real-time heart rate, blood pressure, body temperature, blood oxygen data and other health indicator data, as well as the user's posture data, activity frequency and real-time location data, and summarize the above as user activity status data to send to the outside;

[0083] When using, combine the contents in steps 301 to 303:

[0084] When the elderly user is in an unhealthy state, the health indicator data, posture data, activity frequency and real-time location data are output in real time. After receiving these data, the elderly user can be processed in time, for example, treatment or corresponding suggestions can be given to ensure his health status.

[0085] Step 4: Identify the health score of the elderly, issue a multi-level alarm to the outside when the health score is lower than expected, and generate a comprehensive health report;

[0086] The step 4 includes the following contents:

[0087] Step 401: training a deep learning algorithm with the labeled sample data to obtain a trained user health scoring model;

[0088] The health index data is used as input, and the trained user health scoring model is used to score the user's health and obtain the corresponding health score; if the user's health score is lower than expected, it is regarded as an unhealthy elderly user and a first-level alarm instruction is issued to the outside;

[0089] Step 402: If the number of consecutive first-level alarm instructions received during the preset observation period exceeds the expected number, a warning value is constructed based on the score difference between the health score and the target score, and the time node distribution state when the first-level alarm instruction is received each time, in the following manner:

[0090]

[0091] g(t,t i )=exp(-λ(tt i ))

[0092] Where: t i is the time point when the first-level alarm command is received, λ is the decay rate parameter, n is the number of first-level alarms, S i is the score difference at the i-th level 1 alarm, g(t,t i ) is the time decay function, α and β are weight coefficients, both of which fall within [0,1], and their sum is 1; w i It is the weight parameter of the early warning, which is determined according to the actual situation;

[0093] Based on historical data and the user's health management expectations, the warning threshold is pre-set; if the obtained warning value exceeds the warning threshold, it means that the user's current health status is abnormal, unhealthy and may further deteriorate. At this time, a secondary alarm instruction is issued to the outside;

[0094] Summarize the user's health indicator data, activity status data, and environmental parameters, and use heat maps to represent the user's high-frequency activity areas, and generate a comprehensive health report after aggregation;

[0095] When using, combine the contents in steps 401 and 402:

[0096] If the screened out elderly users in unhealthy conditions are taken as the target, and health scores are performed using the trained user health scoring model based on the acquisition of various health indicator data, a comprehensive assessment of the health status of elderly users can be achieved. At the same time, through the multi-level alarm mechanism constructed by the first-level alarm instructions and the second-level alarm instructions, when the health status of elderly users is abnormal, it can be handled in a timely manner to avoid missing abnormal situations.

[0097] Step 5: Based on the relationship between the health score and the health threshold, take appropriate measures for elderly users in unhealthy conditions;

[0098] The step five includes the following contents:

[0099] Step 501: After obtaining several health scores Pk continuously, construct a health threshold [Qa, Qb] from the health scores in the following manner:

[0100]

[0101] Among them, Pk i is the health score of the ith person, Pk avg , Pk max and Pk min are the mean, maximum and minimum values ​​of the quality score respectively;

[0102] When in use, by constructing health thresholds from continuous health scores, dynamic evaluation of the health status of elderly users can be achieved, so that the evaluation results are adapted to the activity area;

[0103] Step 502: taking corresponding processing methods according to the relationship between the health score and the health threshold, including:

[0104] If the user's health score exceeds the health threshold, no additional processing is performed and the user's health indicator data is sent in real time;

[0105] If the user's health score is within the health threshold, the corresponding health features are obtained after feature extraction of the patient's comprehensive health report; taking the health management of elderly users as the target word, after deep retrieval and entity relationship building, a knowledge graph of health management for elderly people living alone is pre-built; based on the correspondence between health management plans and health features, the knowledge graph of health management for elderly users outputs personalized health management plans, such as dietary adjustment suggestions and exercise plans, etc.;

[0106] If the user's health score is below the health threshold, the user will be given a video consultation and the comprehensive health report will be shared on the video consultation platform;

[0107] When using, combine the contents in steps 501 and 502:

[0108] Based on the relationship between the health score and the health threshold, the degree of the unhealthy state can be further differentiated after the elderly users enter the unhealthy state, and targeted coping methods can be taken for the elderly users in the unhealthy state. In the medical scenario within the current living area, the unhealthy state can be responded to quickly to avoid the situation where the elderly users do not care about their own health status, or understand their own status but do not take any action. This ensures the health status and efficiency of medical intervention for elderly users, especially the elderly living alone, and can also fully play a guiding role.

[0109] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0110] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0111] The construction method of the knowledge graph of health management for elderly people living alone can refer to the following:

[0112] Clear goal: The main goal of building a knowledge graph for health management of elderly people living alone is to integrate and manage various information related to the health of elderly people living alone, including diseases, symptoms, treatment methods, preventive measures, living habits, etc., in order to provide intelligent support for health management.

[0113] Define the scope: Determine the areas and specific content that the knowledge graph needs to cover, such as common diseases (hypertension, diabetes, heart disease, etc.), health indicators (blood pressure, blood sugar, blood lipids, etc.), lifestyle habits (diet, exercise, sleep, etc.) and related medical knowledge and health management recommendations.

[0114] Data source selection: Obtain health-related knowledge and information from a variety of data sources, including medical literature databases (such as PubMed, Wanfang, etc.), medical treatment records, public health data sets (such as data released by the Centers for Disease Control and Prevention), user health data (such as data collected by wearable devices), etc.

[0115] Data cleaning: Clean the collected data to remove duplicate, erroneous or irrelevant information to ensure the accuracy and consistency of the data.

[0116] Data preprocessing: Formatting data, such as converting text data into structured data for subsequent knowledge extraction and storage.

[0117] Entity Recognition: Use named entity recognition (NER) technology to identify medical entities such as diseases, symptoms, drugs, treatments, etc. from text data.

[0118] Relationship extraction: Extract the relationship between entities through semantic parsing technology, such as "a certain drug treats a certain disease", "a certain disease causes a certain symptom", etc.

[0119] Knowledge representation: The extracted knowledge is represented in the form of triples (entity-relationship-entity) and a suitable storage solution (such as graph database Neo4j or RDF database) is selected for storage.

[0120] Entity and relationship construction: Create entities and relationships in the knowledge graph, such as creating a "hypertension" disease entity and associating it with symptom entities such as "dizziness" and "palpitations" and the "antihypertensive drug" treatment entity.

[0121] Attribute filling: Add attribute information to the entity, such as the name of the disease, description of the symptoms, dosage of the drug, etc.

[0122] Graph optimization: Optimize the constructed knowledge graph, such as removing redundant relationships, merging similar entities, etc., to improve the quality and efficiency of the graph.

[0123] Personalized recommendations: Based on knowledge graphs and user data, combined with recommendation algorithms, personalized health management plans are generated for elderly people living alone, such as dietary advice, exercise plans, and guidance on drug use.

[0124] Disease warning and intervention: Predict possible health risks for elderly people living alone through knowledge graph reasoning, and provide corresponding warnings and intervention recommendations.

[0125] Health education and consultation: Use knowledge graphs to provide health education and consultation services to elderly people living alone, helping them better understand and manage their health conditions.

[0126] See also Figure 2 The present invention provides a medical environment analysis system based on deep learning and Internet of Things technology, including:

[0127] The data analysis unit collects the living status data of the elderly users in the activity area. If there is abnormal data in the living status data, the abnormal recognition data generates the data abnormality degree, and issues a behavior pattern recognition instruction when the data abnormality degree exceeds expectations;

[0128] A constraint unit, after identifying the behavior pattern data of the elderly user, if the similarity between the behavior pattern data and the target behavior pattern is lower than expected, constrains the collection frequency of the life status data according to the similarity;

[0129] The monitoring unit collects real-time health indicator data of elderly users. If the health status indicator data is abnormal and the degree of abnormality exceeds expectations, the activity status data of the elderly is output in real time;

[0130] The scoring unit identifies the health score of the elderly, issues a multi-level alarm to the outside world when the health score is lower than expected, and generates a comprehensive health report;

[0131] The solution output unit takes corresponding measures for elderly users in unhealthy conditions based on the relationship between the health score and the health threshold.

[0132] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0133] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0134] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0135] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0136] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0137] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0138] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A medical environment analysis system based on deep learning and IoT technology, characterized by: include, The data analysis unit collects the life status data of the elderly users. If there is an abnormality in the life status data, the abnormality degree of the data is generated by the abnormality recognition data, and a behavior pattern recognition instruction is issued when the data abnormality degree exceeds the expectation; A constraint unit, after identifying the behavior pattern data of the elderly user, if the similarity between the behavior pattern data and the target behavior pattern is lower than expected, constrains the collection frequency of the life status data according to the similarity; The monitoring unit collects real-time health indicator data of elderly users. If the health status indicator data is abnormal and the degree of abnormality exceeds expectations, the activity status data of the elderly is output in real time; The scoring unit identifies the health score of the elderly, issues a multi-level alarm to the outside world when the health score is lower than expected, and generates a comprehensive health report; The solution output unit takes corresponding measures for elderly users in unhealthy conditions based on the relationship between the health score and the health threshold.

2. The medical environment analysis system according to claim 1, characterized in that: Collect and detect data in real time in the activity area of ​​elderly users, obtain corresponding life status data, and summarize and generate life status data summary after pre-processing; Taking real-time life status data as input, the trained abnormal data recognition model is used to perform abnormality recognition, obtain the abnormal data and the corresponding abnormal proportion, and thus generate the data abnormality degree.

3. The medical environment analysis system according to claim 2, characterized in that: After receiving the behavior pattern recognition instruction, the user's daily activity data is collected in the living area and aggregated to generate a daily behavior data set; Taking the user's daily behavior data as input, the trained behavior pattern recognition model is used to perform behavior pattern recognition to obtain the user's normal behavior pattern, abnormal behavior pattern and the distribution time of different behavior patterns, and then aggregate them to generate the user's behavior pattern data set.

4. The medical environment analysis system according to claim 3, characterized in that: Perform similarity analysis on the user's behavior pattern data set and the expected behavior pattern data to obtain the corresponding similarity data, and collect the user's life status data based on the collection frequency that meets the constraint conditions.

5. The medical environment analysis system according to claim 4, characterized in that: The real-time health indicator data is used as input, and the trained abnormal indicator recognition model is used to identify abnormalities in the user's physical health indicators; after screening out abnormal health indicators, the corresponding abnormal degree and the abnormal nodes that generate the abnormality are confirmed; after aggregation, an abnormal indicator data set is generated.

6. The medical environment analysis system according to claim 5, characterized in that: The health abnormality of the user's health indicator is generated by the abnormal indicator data set. If the health abnormality exceeds the health threshold, a health trend analysis instruction is issued to the outside.

7. The medical environment analysis system according to claim 6, characterized in that: After receiving the health trend analysis instruction, continuously obtain the health indicator data of the user in several sub-periods, perform health trend analysis on the user, and obtain corresponding health trend data; The user's health indicator data, as well as the user's posture data, activity frequency and real-time location data are collected and aggregated and sent out as user activity status data.

8. The medical environment analysis system according to claim 7, characterized in that: The health indicator data is used as input, and the user's health score is scored using the trained user health score model; If the user's health score is lower than expected, he or she will be treated as an unhealthy elderly user and a first-level alarm will be issued to the outside.

9. The medical environment analysis system according to claim 8, characterized in that: If the number of consecutive first-level alarm instructions received during the preset observation period exceeds expectations, the warning value is constructed based on the score gap between the health score and the target score, as well as the distribution of time nodes when the first-level alarm instruction is received each time; If the acquired warning value exceeds the warning threshold, a secondary alarm command is issued to the outside; Summarize the user's health indicator data, activity status data and environmental parameters, and use heat maps to represent the user's high-frequency activity areas, and generate a comprehensive health report after aggregation.

10. The medical environment analysis system according to claim 9, characterized in that: If the user's health score exceeds the health threshold, no additional processing is performed and the user's health indicator data is sent in real time; If the user's health score is within the health threshold, the elderly user health management knowledge graph will output a personalized health management plan, such as dietary adjustment suggestions and exercise plans; If the user's health score is below the health threshold, a video consultation will be conducted on the user, and the comprehensive health report will be shared on the video consultation platform.

Images