Patient nursing intelligent follow-up visit system and method based on Internet of Things

Abstract

The invention relates to the technical field of health data processing, and particularly discloses a patient nursing intelligent follow-up visit system and method based on the Internet of Things, and the system comprises a sensing layer, an edge computing device, a network layer, a platform layer, and an application layer. The sensing layer covers various Internet of Things devices, the devices transmit data to the edge computing device or the network layer through various communication protocols, a rich and real-time data basis is provided for subsequent analysis, the edge computing device performs format conversion, abnormal data filtering and duplicate removal processing on the data, the data format is unified, and the data processing efficiency is improved. Invalid data transmission is reduced, the operation pressure of a core server is reduced, the data processing efficiency and the network bandwidth utilization rate are improved, meanwhile, a network layer adopts multiple communication technology collaboration, an encryption technology and message queue middleware, the continuity, safety and reliability of data transmission are guaranteed, and the service life of the network layer is prolonged. And the patient data can be accurately transmitted to the platform layer.

Description

Technical Field

[0001] The present invention relates to the technical field of health data processing, and in particular to an intelligent patient care follow-up system and method based on the Internet of Things. Background Art

[0002] In the modern medical system, patient care follow-up is an important part of ensuring the patient's rehabilitation effect and health management. Traditional patient care follow-up mainly relies on regular telephone inquiries by medical staff or patients' follow-up visits to the hospital, which has many limitations. On the one hand, information acquisition is not comprehensive and timely, and it is impossible to grasp the changes in patients' physiological indicators and living environment conditions in real time. For example, it is difficult to accurately understand the patient's daily exercise, sleep quality, and the temperature, humidity, and air quality of the living environment that affect health through telephone inquiries alone. On the other hand, there is a lack of basis for the formulation of personalized nursing plans, and they can often only be guided by general nursing standards. It is impossible to fully consider factors such as the patient's individual disease type, severity of the disease, rehabilitation stage, and individual differences, resulting in unsatisfactory nursing results.

[0003] With the development of Internet of Things technology, although some medical devices have achieved remote data transmission, there are still deficiencies in data processing and application. The data formats of different devices are different, which makes it difficult to integrate and analyze, and the data security and stability during the transmission process are difficult to guarantee. At the same time, the existing patient care follow-up system lacks in-depth mining and intelligent analysis of data, and cannot generate accurate care plans and risk warnings in a timely manner according to the specific conditions of patients, and cannot meet the growing demand for intelligent and personalized medical care. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent follow-up system and method for patient care based on the Internet of Things, in order to solve the above-mentioned technical defects.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent follow-up system for patient care based on the Internet of Things, including a perception layer, an edge computing device, a network layer, a platform layer and an application layer; The perception layer includes wearable devices, home medical detection equipment and environmental monitoring sensors; The edge computing device is based on the intelligent gateway, with a built-in high-performance processor and a customized Linux operating system, and is equipped with edge data processing software; The network layer switches to 4G / 5G backup transmission when the network is poor, and uses SSL / TLS encryption protocol to ensure data security; The platform layer includes a personalized nursing plan formulation unit and a patient health status risk warning unit; The application layer provides medical staff with complete patient health data, nursing plan execution and risk warning information, and supports video remote guidance and plan adjustment.

[0006] Furthermore, the wearable device collects the patient's physiological and motion data at a specific frequency, the home medical detection equipment generates medical data in real time, the environmental monitoring sensor collects environmental information, and all devices transmit raw data via Bluetooth, Wi-Fi or 4G / 5G communication protocols.

[0007] Furthermore, the edge computing device is used to unify data formats, identify and filter abnormal data, deduplicate data with minimal changes, and reduce invalid data transmission.

[0008] Furthermore, a message queue middleware RabbitMQ is deployed between the network layer and the platform layer to ensure the reliability and asynchrony of data transmission. When stable, data is uploaded at high speed via wired Ethernet according to the TCP / IP protocol.

[0009] Furthermore, the personalized care plan formulation unit at the platform layer extracts care cases and patient health data, processes them with Apache Hadoop after cleaning, writes a decision tree algorithm using TensorFlow, and constructs a decision tree model through information gain IG to generate a personalized care plan.

[0010] Furthermore, the patient health status risk warning unit at the platform layer sets the risk indicator time series data, selects the moving average period Q according to the condition and indicator characteristics, calculates the moving average and the difference, sets the fluctuation threshold, and determines whether to trigger a risk warning according to the rules.

[0011] Furthermore, the application layer also provides health status display, nursing advice reception and mutual assistance community functions for patients and their families, and is provided with a feedback collection module for system optimization.

[0012] Furthermore, an intelligent follow-up method for patient care based on the Internet of Things includes: perception layer data collection, using various devices to collect data and transmit it to edge computing devices; edge computing device preprocessing, unifying the format, filtering out exceptions, and deduplicating data; network layer data transmission and security assurance, transmitting via a wired network when stable, switching and encrypting, and caching data when unstable; formulation of personalized care plans, extracting and cleaning data, and building a decision tree model to generate plans; patient health status risk warning, setting parameters and judging according to rules; application layer data display and interaction, providing functions to all parties and collecting feedback to optimize the system.

[0013] Furthermore, in the perception layer data collection step, wearable devices collect heart rate, exercise, and sleep data, home medical testing equipment generates blood pressure and blood sugar data, and environmental monitoring sensors collect air quality, temperature and humidity data.

[0014] Furthermore, in the personalized nursing plan formulation step, the decision tree model input includes a feature set including disease type, disease severity, rehabilitation stage and individual differences, and the output is a nursing plan content set; in the patient health status risk warning step, the moving average period Q is selected based on the stability of the disease and the fluctuation characteristics of the indicators.

[0015] Beneficial effects of the present invention: 1. The perception layer of the system of the present invention covers a variety of Internet of Things devices, wearable devices, home medical detection equipment and environmental monitoring sensors, which can comprehensively and frequently collect patients' physiological data, motion data, medical detection data and living environment data. These devices transmit data to edge computing devices or network layers through multiple communication protocols, providing a rich and real-time data basis for subsequent analysis. The edge computing devices perform format conversion, abnormal data filtering and deduplication processing on the data, which not only unifies the data format and reduces invalid data transmission, but also reduces the computing pressure of the core server, improves data processing efficiency and network bandwidth utilization. At the same time, the network layer adopts multiple communication technologies to coordinate as well as encryption technology and message queue middleware to ensure the continuity, security and reliability of data transmission, and ensure that patient data can be accurately transmitted to the platform layer.

[0016] 2. The personalized nursing plan formulation unit at the platform level, through in-depth mining and analysis of nursing case data and patient health data, uses big data analysis tools and artificial intelligence algorithms to build a decision tree model. This model can fully consider factors such as the patient's disease type, disease severity, rehabilitation stage, and individual differences, and accurately learn the mapping relationship between different feature combinations and corresponding nursing plans, thereby generating an exclusive personalized nursing plan for each patient. This personalized nursing plan can better meet the individual needs of patients and improve nursing effects and rehabilitation quality.

[0017] 3. The patient health status risk warning unit uses time series analysis to select the appropriate moving average period according to the patient's condition and indicator characteristics, and dynamically analyzes the patient's health data. By calculating the moving average, difference, and setting the fluctuation threshold, it can timely and accurately determine whether the health indicators have abnormal fluctuations, thereby triggering a risk warning. This risk warning mechanism enables medical staff to understand the patient's health risk status in a timely manner and take appropriate intervention measures to prevent the deterioration of the disease and the occurrence of complications. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below in conjunction with the accompanying drawings.

[0019] Figure 1 This is a principle block diagram of an intelligent patient care follow-up system based on the Internet of Things according to an embodiment of the present invention; Figure 2 This is a logical flow chart of the in-depth analysis of the patient health data set by the personalized nursing plan formulation unit in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0021] Example 1: Please refer to Figure 1 As shown, an intelligent patient care follow-up system based on the Internet of Things includes: a perception layer, a network layer, a platform layer and an application layer. The perception layer sends the collected raw data to the edge computing device through communication protocols such as Bluetooth, Wi-Fi or 4G / 5G. The edge computing device uses a preset algorithm to convert the format of the collected raw data, unifies the different data formats of different devices, and identifies and filters out abnormal data, retaining only valid data. The data pre-processed by the edge is transmitted to the platform layer through the communication technology of the network layer. In the data center of the platform layer, the computer starts the data cleaning algorithm to obtain a complete and accurate patient health data set, and uses big data analysis tools and artificial intelligence algorithms to deeply analyze the integrated patient health data set to generate exclusive care for the patient. Management plan, and at the same time, the system analyzes and judges the patient's health data, obtains the risk level of the patient's physiological indicators and issues early warning information. The patient's health data after analysis and processing is presented in an intuitive form at the application layer. Medical staff use follow-up management applications to view the patient's complete health data, the implementation of personalized nursing plans, and risk warning information, and provide remote guidance and adjust nursing plans for patients based on this. Patients and their families can understand their health status, receive nursing advice, and participate in mutual assistance communities on their respective application ends. At the same time, the system collects feedback data from medical staff, patients, and their families during use, such as satisfaction with the nursing plan and suggestions for improvement of system functions. These feedback data enter the data processing process again to optimize algorithms, improve system functions, and achieve continuous iteration and improvement of the system.

[0022] It should be noted that the perception layer of the system includes various types of IoT devices, including wearable devices, home medical testing equipment, and environmental monitoring sensors. Wearable devices, such as smart bracelets, continuously monitor patients' heart rate, number of steps, sleep duration and other physiological and motion data, and collect and transmit them at a frequency of seconds. Home medical monitoring equipment, such as electronic blood pressure monitors and blood glucose meters, instantly generate blood pressure and blood sugar data when patients are measured. Environmental monitoring sensors are responsible for collecting environmental information such as indoor air quality, temperature and humidity. These devices send the collected raw data to edge computing devices or directly upload them to the network layer through communication protocols such as Bluetooth, Wi-Fi or 4G / 5G, providing basic materials for subsequent processing.

[0023] It should also be noted that edge computing devices can use smart gateways to collect data from wearable devices, home medical monitoring equipment, and environmental monitoring sensors, and deduplicate data that changes very little or is repeated over a continuous time period, greatly reducing the amount of invalid data transmission, alleviating the computing pressure on the core server, improving overall data processing efficiency, and optimizing network bandwidth utilization.

[0024] It is also necessary to further explain that when the network signal is poor or the data volume is large, the 4G / 5G network is used as a backup transmission path to ensure the continuity of data transmission; during the transmission process, encryption technology is used to encrypt the data to ensure the security of sensitive patient information during the transmission link and prevent data leakage. At the same time, a message queue middleware is deployed between the network layer and the platform layer to ensure the reliability and asynchrony of data transmission between the perception layer equipment and the platform layer. When the network fluctuates, the data is temporarily cached. After the network returns to normal, the data is pushed to the platform layer in order and accurately to maintain the reliability and stability of data transmission.

[0025] Specifically, the wearable device uses a smart bracelet with high-precision sensors. Its built-in heart rate sensor can accurately monitor the patient's heart rate at a frequency of once per second, and transmit the data in real time to the surrounding edge computing device or directly upload it to the network layer through low-power Bluetooth 5.0 technology; for motion data, the smart bracelet uses accelerometers and gyroscopes to collect information such as the patient's steps, motion trajectory, and motion posture 10 times per second to ensure that the patient's daily activity status is fully captured; in terms of sleep monitoring, the smart bracelet uses photoelectric sensors to record the patient's sleep status data every 5 seconds, including the time to fall asleep, light sleep, deep sleep and wake-up time, etc., to analyze sleep quality.

[0026] Among home medical monitoring devices, electronic blood pressure monitors use the oscillometric method to measure blood pressure. After the patient presses the measurement button, the device instantly starts the measurement program. After the measurement is completed, the systolic blood pressure, diastolic blood pressure, heart rate and other data are transmitted to the network layer or edge computing device in a standard data format within 1 second through the Wi-Fi module; the blood glucose meter uses the glucose oxidase method. After the patient draws blood, the device obtains the blood glucose value within 3 seconds and sends the data with the help of Bluetooth 4.0 technology. Environmental monitoring sensors are deployed in the patient's living space. For example, the indoor air quality sensor uses electrochemical principles to detect the concentration of harmful gases in the air and collects data every 30 seconds. The temperature and humidity sensor uses a capacitive humidity sensor and a thermistor to measure the indoor temperature and humidity once a second. These environmental data are aggregated to the smart gateway via ZigBee wireless communication technology, and then uploaded to the system platform by the smart gateway via wired Ethernet or 4G / 5G network.

[0027] Specifically, the edge computing device is centered on the intelligent gateway. The intelligent gateway has a built-in high-performance processor, runs a customized Linux operating system, and is equipped with specially developed edge data processing software. After receiving the raw data from the perception layer device, the software first unifies the data format of different devices according to the preset data format conversion rules; for example, the heart rate data of the smart bracelet is converted from a custom binary format to a JSON format that complies with international standards for subsequent transmission and processing within the system.

[0028] In terms of data anomaly judgment, the smart gateway uses a simple threshold algorithm. Taking heart rate data as an example, if the received heart rate value instantly exceeds 220 beats / minute (the maximum heart rate of an average adult), the smart gateway determines that the data is abnormal data and discards it, retaining only data within the normal range. For data with very small changes in a continuous time period, such as the changes in 10 consecutive temperature and humidity data measurements are all within 0.1%, the smart gateway uses a sliding window algorithm to deduplicate data and only retains one of the data, which greatly reduces the amount of invalid data transmission, effectively reduces the computing pressure of the core server, optimizes network bandwidth utilization, and ensures the efficiency and stability of data transmission.

[0029] Specifically, the network layer uses a variety of communication technologies to work together for data transmission. In indoor scenarios with stable network environments, edge pre-processed data is uploaded to the cloud server through the wired Ethernet interface in accordance with the TCP / IP protocol at a rate of 100Mbps or above. When the network signal is poor or in an outdoor mobile scenario, the system automatically switches to the 4G / 5G network. Taking the 4G network as an example, the data transmission rate can reach 10Mbps-100Mbps, ensuring continuous data transmission.

[0030] To ensure the security of sensitive patient information during transmission, the system uses the SSL / TLS encryption protocol. Before data transmission, the sending device and the receiving server negotiate the encryption key through a handshake protocol, and then the data is encrypted into ciphertext during transmission to prevent data leakage. At the same time, the message queue middleware RabbitMQ is deployed between the network layer and the platform layer. When the network fluctuates briefly, RabbitMQ temporarily caches the data in the queue. After the network returns to normal, it pushes the data to the platform layer accurately in a first-in-first-out order, maintaining the reliability of data transmission and avoiding data loss or confusion.

[0031] In a specific embodiment, the perception layer of the system of the present invention covers a variety of Internet of Things devices, wearable devices, home medical detection equipment and environmental monitoring sensors, which can comprehensively and frequently collect patients' physiological data, motion data, medical detection data and living environment data. These devices transmit data to edge computing devices or network layers through multiple communication protocols, providing a rich and real-time data basis for subsequent analysis. The edge computing devices perform format conversion, abnormal data filtering and deduplication processing on the data, which not only unifies the data format and reduces invalid data transmission, but also reduces the computing pressure of the core server, improves data processing efficiency and network bandwidth utilization. At the same time, the network layer uses multiple communication technologies to collaborate as well as encryption technology and message queue middleware to ensure the continuity, security and reliability of data transmission, ensuring that patient data can be accurately transmitted to the platform layer.

[0032] Embodiment 2: It needs to be further explained that the platform layer includes a personalized nursing plan formulation unit and a patient health status risk warning unit.

[0033] The personalized care plan formulation unit obtains complete and accurate patient health data sets, uses big data analysis tools and artificial intelligence algorithms to conduct in-depth analysis of the integrated patient health data sets, and generates exclusive care plans for patients, such as Figure 2 The specific analysis process is as follows: Extract nursing case data and patient health data from the system's data center. Suppose the nursing case data set is ,in represents the i-th nursing case, each case contains nursing measures and effect information; the patient health data set is ,in Represents the health data of the jth patient, including the disease type , severity of illness , Rehabilitation stage and individual difference information Among them, the disease type is represented by a classification code, the severity of the disease is 1, 2, and 3, corresponding to mild, moderate, and severe respectively, the recovery stage is 1, 2, and 3, corresponding to early, middle, and late recovery respectively, and individual difference information includes age A, gender G, and allergy history code H, etc.

[0034] Clean the above extracted data. For the nursing case dataset C and the patient health dataset P, traverse all data records. If there is key information in the record (such as If the age is missing or the data is incorrect (such as age A is a negative number), the record will be deleted from the corresponding data set.

[0035] Use the big data analysis tool Apache Hadoop to distribute and process nursing case data sets and patient health data sets, and use the artificial intelligence algorithm library TensorFlow to write decision tree algorithm code.

[0036] Defining feature sets , which contains the key attributes of the patient's health data and is used as the input of the decision tree. The nursing plan content set is set to , k represents the number of different nursing plan contents, Represents specific care plan content such as dietary recommendations and exercise guidance as output labels of the decision tree.

[0037] Information gain IG is used to select the optimal partitioning feature to determine the branch structure of the decision tree. , the information gain calculation formula is: .

[0038] Among them, S in the calculation formula represents the sample set currently involved in the division, is the information entropy of the sample set S, which is used to measure the uncertainty of the sample set. is the number of care regimen categories, is the proportion of samples belonging to the i-th type of care plan in the sample set S, is the value set of feature f, for example, the value set of disease severity S is , The feature f in the sample set S has the value A subset of samples, and Respectively represent subsets and the number of samples in set S.

[0039] It should be noted that the number of nursing plan categories is determined based on the degree of difference between different nursing plans in actual nursing cases. For example, according to different nursing methods, it can be divided into several categories such as dietary care, exercise care, drug care, and psychological care. Under each major category, it is further subdivided according to specific nursing measures and goals. For example, dietary care can be subdivided into low-sugar diet, low-salt diet, etc.; exercise care can be divided into aerobic exercise, strength training, etc. Through in-depth analysis of the nursing case data set and combined with the professional experience of medical staff, the final number of nursing plan categories is determined.

[0040] The principle of selecting the sample set S is to cover representative patient data. In actual operation, patient data of different disease types, different disease severity, different rehabilitation stages, and different individual differences within a certain period of time (such as the past year) are extracted from the system's data center as the sample set S. To ensure the randomness and representativeness of the samples, a stratified sampling method is used to stratify according to factors such as disease type and disease severity, and then a certain proportion of samples are randomly selected from each layer. For example, in the layer of patients with diabetes, further subdivided according to the severity of the disease (mild, moderate, severe), 10% of the samples are selected from each layer to form a sample set S, to ensure that the model training can fully learn the relationship between the characteristics and care plans in different situations.

[0041] In the process of building a decision tree, for each internal node, the information gain of each feature in the feature set F is calculated, and the feature with the largest information gain is selected as the partition feature of the node. By repeating this process, the sample set is gradually divided until the stopping condition that all samples belong to the same category is met, and a complete decision tree model is constructed.

[0042] Moreover, in this process, the model continuously adjusts the structure of the decision tree by learning and training historical data (nursing case dataset C and patient health dataset P), so that the decision tree can accurately learn the mapping relationship between different feature combinations and corresponding nursing plans.

[0043] Get new patient data And input into the decision tree model, first of all, the disease type, severity of the disease, recovery stage and individual difference information of the new patient are encoded and organized, and then starting from the root node of the decision tree, according to the value of each feature in the new patient data, traverse downward according to the determined branching rules of the decision tree until the leaf node of the decision tree is reached. The nursing plan content corresponding to the leaf node is the personalized nursing plan generated for the new patient, and the generated personalized nursing plan is transmitted to the application layer for display notification.

[0044] Specifically, for downward traversal according to the branching rules determined by the decision tree, if the root node uses disease type D as the dividing feature, and the disease type of the new patient is For a specific branch, continue down along the branch; if the leaf node you finally reach corresponds to the dietary advice of "daily carbohydrate intake should be controlled within x grams, and it is recommended to eat oats, broccoli, etc.", the exercise guidance is "take y times a week, z minutes of walking each time", and the medication reminder is "take c tablets of medication b at a time every morning", etc., these contents will be provided to new patients through the application layer to guide their subsequent care and rehabilitation.

[0045] In a specific embodiment, the personalized care plan formulation unit at the platform layer, through in-depth mining and analysis of nursing case data and patient health data, uses big data analysis tools and artificial intelligence algorithms to build a decision tree model. This model can fully consider factors such as the patient's disease type, disease severity, rehabilitation stage, and individual differences, and accurately learn the mapping relationship between different feature combinations and corresponding care plans, thereby generating an exclusive personalized care plan for each patient. This personalized care plan can better meet the individual needs of patients and improve care effects and rehabilitation quality.

[0046] Embodiment 3: The patient health status risk warning unit uses the time series analysis method to make a warning judgment on the patient's health risk. The specific judgment process is as follows: Assume that the time series data of risk indicators related to patient health is , g = 1, 2, …, t, It is represented by the measured value of the z-th health data at the g-th time point, g is represented by the number of each time point, and t is represented by the total number of each time point. These data are arranged in chronological order to form the basis for analysis.

[0047] The moving average period Q is selected according to the patient's condition and the characteristics of the indicators. For unstable conditions and large fluctuations in indicators, Q takes a smaller value; for stable conditions and small fluctuations in indicators, Q takes a larger value. A specific expression is: for patients with arrhythmia and large heart rate fluctuations, Q may take a smaller value such as Q1; for hypertensive patients with relatively stable blood pressure, Q may take a larger value such as Q2 (Q1<Q2).

[0048] According to the selected moving average period Q, use the formula Calculate the moving average of period g ; A specific expression is: when monitoring the patient's blood sugar data, if the blood sugar data number z=3, Q=7, g=15, then , which reflects the average blood sugar level of the patient in the past 7 days. As time goes by, the g value increases and is continuously updated. calculations to reflect the latest trends.

[0049] It should be noted that a general moving average period Q selection quantitative method is determined by combining the standard deviation of the disease fluctuation and the frequency of the index fluctuation. First, calculate the standard deviation of the patient's health index data over a period of time (such as the past 30 days). The larger the standard deviation, the more severe the disease fluctuation. Suppose the standard deviation of a patient's heart rate data in the past 30 days is st, and the index fluctuation frequency is fr (such as the number of measurements per day). Set a comprehensive evaluation formula: , where k1 and k2 are adjustment coefficients, which can be adjusted according to actual conditions, and the general range is between 0.5-2. For example, for patients with arrhythmia with large heart rate fluctuations, the standard deviation of their heart rate data is large and the fluctuation frequency is high. If k1=1, k2=1.2, after calculation, Q may be 3-5; for hypertensive patients with relatively stable blood pressure, the standard deviation of blood pressure data is small, the fluctuation frequency is relatively low, and Q may be 7-10. At the same time, according to different diseases and indicators, adjustments are made in combination with clinical experience. For example, for patients with chronic obstructive pulmonary disease (COPD), the fluctuation of their lung function indicators (such as blood oxygen saturation) is affected by many factors, and the disease changes relatively slowly. The Q value can be appropriately increased, and the range is between 10-14; for patients with hyperthyroidism, their thyroid hormone levels fluctuate more frequently, and the Q value can be between 5-8, so as to capture the trend of indicator changes more timely.

[0050] Calculate the current value of the risk indicator time series data of various health data of patients and moving average The difference ,like ,show Higher than the average level of the past Q time points, if ,show Lower than the average level of the past Q time points.

[0051] Set corresponding fluctuation thresholds for the risk indicator time series data of various health data of patients , z represents the number of each patient's health data, It is expressed as the proportional coefficient of each patient's health data determined according to the actual situation. It is expressed as the average value of various health data of patients.

[0052] According to the risk warning rules, if , then it is determined that the health indicator has an abnormal fluctuation and a risk warning is triggered. Otherwise, no risk warning is triggered.

[0053] Furthermore, the application layer is used to present the analyzed and processed data in an intuitive visual interface. Medical staff can log in to the follow-up management application to view the patient's complete health data, including historical physiological indicator change curves, nursing plan execution progress charts, and risk warning information pop-ups. Based on this information, medical staff use the video call function in the application to provide remote guidance to patients, such as guiding rehabilitation training movements; according to the patient's actual recovery situation, they can adjust the nursing plan in the application and modify the diet, exercise, medication and other recommended content.

[0054] Patients and their families can view the health status dashboard through their respective applications, which visually shows whether the current physiological indicators are normal; receive push notifications of nursing suggestions and click to view detailed nursing content; participate in the mutual aid community, post text, pictures, etc. to share rehabilitation experiences, and view other patients' experience sharing and expert online lecture videos.

[0055] At the same time, the system sets up a feedback collection module at the application layer. Medical staff can provide feedback on problems encountered during the implementation of the nursing plan and the convenience of system functions; patients and their families can provide feedback on their satisfaction with the nursing suggestions and whether the system operation is easy. These feedback data are transmitted to the platform layer through the network and enter the data processing process to optimize the algorithm, such as adjusting the frequency of nursing plans according to the feedback of medical staff and optimizing the parameters of the decision tree model; improving system functions, such as improving the application interface design based on the complex operation problems reported by patients' families, realizing continuous iteration and improvement of the system, and better serving the intelligent follow-up needs of patients' nursing care.

[0056] It should be noted that in terms of algorithm optimization, taking the decision tree model as an example, when medical staff feedback that the frequency of nursing plan adjustment is high, it indicates that the current decision tree model is not accurate enough in the division of certain features, and the importance of the features needs to be re-evaluated. After the system collects these feedback data, the decision tree model is retrained. First, according to the patient cases involved in the feedback data, the corresponding feature data and the actual nursing plan adjustment are extracted. Then, the information gain of these features is recalculated. If it is found that a certain feature (such as disease type) has a large change in information gain in the new calculation, it means that the division of the feature may need to be adjusted. For example, if it is found that the nursing plans for patients with different disease severity under a certain disease type are quite different, but the previous decision tree model did not fully consider this factor in the division of the disease type, then adjust the branch nodes of the disease type in the decision tree model and re-divide the sample set to improve the model's prediction accuracy of the nursing plan under the disease type.

[0057] In terms of improving system functions, when patients' families reported that the application interface was complicated to operate, the system development team first conducted a user experience survey to collect more users' operating habits and feedback. Based on the survey results, the application interface was redesigned. For example, the operation process was simplified, and common functions (such as viewing health data and receiving nursing advice) were set in a more prominent position to reduce the number of operation steps; the interface layout was optimized, and concise and clear icons and text prompts were used to improve the readability and operability of the interface. At the same time, user testing was conducted, inviting users of different ages and technical levels to try out the new interface, collect feedback and further optimize it to ensure that the system functions can better meet user needs.

[0058] In a specific embodiment, the patient health status risk warning unit uses time series analysis to select a suitable moving average period according to the patient's condition and indicator characteristics, and dynamically analyzes the patient's health data. By calculating the moving average, difference, and setting the fluctuation threshold, it can timely and accurately determine whether the health indicators have abnormal fluctuations, thereby triggering a risk warning. This risk warning mechanism enables medical staff to understand the patient's health risk status in a timely manner, take corresponding intervention measures, and prevent the deterioration of the disease and the occurrence of complications.

[0059] Embodiment 4: Specifically, this embodiment also discloses a patient care intelligent follow-up method based on the Internet of Things, including the following steps: Step 1: Data collection at the perception layer: Use various IoT devices at the perception layer to collect data. Wearable devices such as smart bracelets collect patients' physiological and motion data at a specific frequency, such as heart rate once per second, step count and other motion data 10 times per second, and sleep status once every 5 seconds; home medical monitoring equipment, electronic blood pressure monitors generate blood pressure and heart rate data instantly after measurement, and blood glucose meters obtain blood glucose values ​​within 3 seconds of blood sampling; environmental monitoring sensors, indoor air quality sensors collect harmful gas concentration data every 30 seconds, and temperature and humidity sensors measure temperature and humidity data every second. These devices send raw data to edge computing devices through communication protocols such as Bluetooth, Wi-Fi or 4G / 5G.

[0060] Step 2: Preprocessing of edge computing devices: The edge computing device with the intelligent gateway as the core receives the raw data, unifies the data formats of different devices according to preset rules, uses the threshold algorithm to identify and filter abnormal data (such as heart rate exceeding 220 times / minute), and uses the sliding window algorithm to deduplicate data with very small continuous changes (such as 10 consecutive temperature and humidity data changes within 0.1%), reducing invalid data transmission and improving data processing efficiency.

[0061] Step 3: Data transmission and security at the network layer: In indoor environments with a stable network, edge-preprocessed data is uploaded to the cloud server via wired Ethernet using the TCP / IP protocol at a rate of 100Mbps or above. When the network is poor or outdoors, it automatically switches to the 4G / 5G network (4G speed can reach 10Mbps-100Mbps). The SSL / TLS encryption protocol is used during transmission to ensure data security. The message queue middleware RabbitMQ is deployed to cache data when the network fluctuates and push it to the platform layer in sequence after recovery to ensure the continuity, security and reliability of data transmission.

[0062] Step 4: Personalized care plan formulation: Extract the nursing case dataset C and the patient health dataset P from the system data center, clean the data, delete the records with missing or incorrect key information, use Apache Hadoop for distributed storage and processing, use TensorFlow to write the decision tree algorithm code, define the feature set F as the decision tree input, and the nursing plan content set N as the output label. Use information gain IG to select the optimal partitioning features to build a decision tree model. After learning and training historical data, adjust the structure so that it can accurately map different feature combinations and nursing plans, obtain new patient data, encode and organize them, and input them into the model. According to the branching rules, traverse to the leaf nodes, generate personalized care plans and transmit them to the application layer.

[0063] Step 5: Patient health status risk warning: Set the time series data of risk indicators related to patient health, select the moving average period Q according to the patient's condition and indicator characteristics, and calculate the moving average using the formula , calculate the difference between the current value and the moving average , set fluctuation thresholds for various health data ,like , determine if the health indicators fluctuate abnormally and trigger a risk warning; otherwise, it will not be triggered.

[0064] Step 6: Application layer data display and interaction: Medical staff can view the patient's complete health data, nursing plan implementation status and risk warning information through the follow-up management application, use video calls to remotely guide patients, and adjust the nursing plan according to the recovery situation. Patients and their families can view their health status, receive nursing advice, participate in mutual aid communities to share and obtain rehabilitation experience and lecture videos through the application.

[0065] Step 7: Application layer data display and interaction: Medical staff can view the patient's complete health data, nursing plan implementation status and risk warning information through the follow-up management application, use video calls to remotely guide patients, and adjust the nursing plan according to the recovery situation. Patients and their families can view their health status, receive nursing advice, participate in mutual aid communities to share and obtain rehabilitation experience and lecture videos through the application.

[0066] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The size of the coefficient is to quantify each parameter to obtain a specific value. Regarding the size of the coefficient, it is acceptable as long as it does not affect the proportional relationship between the parameter and the quantified value.

[0067] In addition, those skilled in the art will appreciate that various aspects of the present invention may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present invention may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present invention may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0068] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0069] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. An intelligent patient care follow-up system based on the Internet of Things, characterized in that: Includes perception layer, edge computing devices, network layer, platform layer and application layer; The perception layer includes wearable devices, home medical detection equipment and environmental monitoring sensors; The edge computing device is based on the intelligent gateway, with a built-in high-performance processor and a customized Linux operating system, and is equipped with edge data processing software; The network layer switches to 4G / 5G backup transmission when the network is poor, and uses SSL / TLS encryption protocol to ensure data security; The platform layer includes a personalized nursing plan formulation unit and a patient health status risk warning unit; The application layer provides medical staff with complete patient health data, nursing plan execution and risk warning information, and supports video remote guidance and plan adjustment.

2. According to the Internet of Things-based patient care intelligent follow-up system according to claim 1, it is characterized in that: The wearable device collects the patient's physiological and motion data at a specific frequency, the home medical detection equipment generates medical data in real time, the environmental monitoring sensor collects environmental information, and all devices transmit raw data via Bluetooth, Wi-Fi or 4G / 5G communication protocols.

3. According to the Internet of Things-based patient care intelligent follow-up system of claim 1, it is characterized in that: The edge computing device is used to unify data formats, identify and filter abnormal data, deduplicate data with minimal changes, and reduce invalid data transmission.

4. According to the Internet of Things-based patient care intelligent follow-up system of claim 1, it is characterized in that: The message queue middleware RabbitMQ is also deployed between the network layer and the platform layer to ensure the reliability and asynchrony of data transmission. When stable, data is uploaded at high speed via wired Ethernet according to the TCP / IP protocol.

5. According to the Internet of Things-based patient care intelligent follow-up system of claim 1, it is characterized in that: The personalized care plan formulation unit at the platform layer extracts care cases and patient health data, processes them with Apache Hadoop after cleaning, writes a decision tree algorithm using TensorFlow, and constructs a decision tree model through information gain IG to generate a personalized care plan.

6. The patient care intelligent follow-up system based on the Internet of Things according to claim 1 is characterized in that: The patient health status risk warning unit at the platform layer sets the risk indicator time series data, selects the moving average period Q according to the condition and indicator characteristics, calculates the moving average and the difference, sets the fluctuation threshold, and determines whether to trigger the risk warning according to the rules.

7. The patient care intelligent follow-up system based on the Internet of Things according to claim 1 is characterized in that: The application layer also provides health status display, nursing advice reception and mutual assistance community functions for patients and their families, and is equipped with a feedback collection module for system optimization.

8. An intelligent follow-up method for patient care based on the Internet of Things, characterized in that: The follow-up method is applied to the intelligent follow-up system for patient care based on the Internet of Things as described in any one of claims 1 to 7 above, including: perception layer data acquisition, using various devices to collect data and transmit it to edge computing devices; edge computing device preprocessing, unified format, filtering exceptions, and deduplication of data; network layer data transmission and security assurance, when stable, it is transmitted via a wired network, and when unstable, it is switched and encrypted, and the data is cached; personalized nursing plan formulation, extraction and cleaning of data, construction of a decision tree model to generate a plan; patient health status risk warning, setting parameters and judging according to rules; application layer data display and interaction, providing functions to all parties and collecting feedback to optimize the system.

9. The method for intelligent follow-up of patient care based on the Internet of Things according to claim 8, characterized in that: In the perception layer data collection step, wearable devices collect heart rate, exercise, and sleep data, home medical testing equipment generates blood pressure and blood sugar data, and environmental monitoring sensors collect air quality, temperature and humidity data.

10. The method for intelligent follow-up of patient care based on the Internet of Things according to claim 8, characterized in that: In the personalized care plan formulation step, the decision tree model input includes a feature set including disease type, severity of the disease, rehabilitation stage and individual differences, and the output is a care plan content set; in the patient health status risk warning step, the moving average period Q is selected based on the stability of the disease and the fluctuation characteristics of the indicators.

Images