Intelligent infusion monitoring and management system

Abstract

The application discloses an intelligent infusion monitoring and management system, and relates to the technical field of medical equipment.The system comprises the following steps: collecting patient data, synchronizing the data to a central monitoring system through blockchain technology, and generating a personalized infusion plan; based on the personalized infusion plan, medical staff use mobile devices to verify the identity of the patient and obtain drug information, and integrate voice assistant functions; using wearable devices and temperature sensors, monitoring the vital signs of the patient and the temperature of the ward, and adjusting the personalized infusion plan and the temperature of the liquid according to the vital signs and the temperature of the ward; according to the adjusted personalized infusion plan and the temperature of the liquid, predicting the risk value of blockage and leakage, and setting a hierarchical alarm mechanism; allowing the patient and the family members to receive infusion progress through an intelligent infusion monitor.The application integrates blockchain technology and advanced artificial intelligence algorithms, and realizes efficient management and analysis of patient health data and drug information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical equipment, and particularly relates to an intelligent infusion monitoring and management system. BACKGROUND

[0002] In modern medical practice, infusion therapy is a very common and critical link in clinical nursing. With the development of technology, infusion monitoring systems have gradually evolved from traditional manual monitoring to intelligent and automated systems. In the prior art, some systems can already achieve automatic monitoring and adjustment of infusion speed, as well as real-time feedback of infusion status through sensors. However, these systems mostly lack deep consideration of individual differences among patients and fail to fully combine patient health data and drug information to generate personalized plans. In addition, existing systems have deficiencies in risk warning and fail to effectively utilize advanced technologies such as deep learning to accurately predict and grade alerts for potential risks (such as blockage and leakage) during infusion, resulting in a lag in the response of medical staff when dealing with emergencies and increasing the safety risks for patients. Further, existing infusion monitoring systems also have obvious shortcomings in doctor-patient interaction. SUMMARY

[0003] In view of the above-mentioned existing problems, the present application is proposed.

[0004] Therefore, the present application provides an intelligent infusion monitoring and management system to solve the problems of lack of personalized plan generation, insufficient accuracy of risk warning, and poor doctor-patient interaction experience in existing infusion monitoring systems.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] The present application provides an intelligent infusion monitoring and management system, which includes a plan generation module responsible for collecting patient data, synchronizing to a central monitoring system through blockchain technology, and generating a personalized infusion plan, wherein the patient data includes health data and drug information;

[0007] An identity verification module is responsible for verifying the identity of the patient and obtaining drug information using a mobile device based on the personalized infusion plan, and integrating a voice assistant function;

[0008] An intelligent adjustment module is responsible for monitoring patient vital signs and ward temperature using wearable devices and temperature sensors, and adjusting the personalized infusion plan and liquid temperature according to the vital signs and ward temperature;

[0009] A risk prediction module is responsible for predicting the risk values of blockage and leakage according to the adjusted personalized infusion plan and liquid temperature, setting a graded alert mechanism, and feeding back to medical staff through a mobile device;

[0010] The doctor-patient interaction module is responsible for allowing patients and family members to receive infusion progress through the intelligent infusion monitor and providing interactive educational resources.

[0011] As a preferred scheme of the intelligent infusion monitoring management system, the health data includes age, gender, weight, height, medical history, allergy history, current condition, blood routine, and liver and kidney function.

[0012] The drug information includes drug type, drug dosage, drug interaction, and drug metabolism.

[0013] As a preferred scheme of the intelligent infusion monitoring management system, the personalized infusion plan is generated according to the patient data, including the following steps,

[0014] The patient health data and drug information are standardized, and a multi-modal data set containing time sequence features is constructed.

[0015] The key feature weight is determined by SHAP value analysis, and an input feature importance model consistent with the clinical decision logic is established.

[0016] A Markov decision process-based infusion decision environment is constructed, which includes state space, action space, reward function, state transition model, initial state distribution, and termination condition. The state space includes patient health data and drug information, the action space is the total amount of drug infusion, the infusion time and the infusion speed, the reward function is the weighted sum of safety reward, economic reward and consistency reward, the state transition model is the dynamic law describing the change of state with action, the initial state distribution is the sampling of patient features and initial vital signs from historical data, and the termination condition is the completion of the total amount of infusion and stable vital signs.

[0017] The strategy constraint Q learning framework is adopted, the entropy regularization and log-sum-exp penalty term are introduced when updating the Q value, the state transition is predicted using the environment transition model, the action probability distribution is output combined with the behavior prediction model, the similarity between the agent strategy and the expert strategy is constrained by KL divergence, and the policy network, Q value network and constraint network in the Actor-Critic architecture are jointly optimized, and the reward maximization and safety constraint are balanced.

[0018] Based on the trained PCQL model, a personalized infusion plan including the total amount of infusion, the infusion time and the infusion speed is output.

[0019] As a preferred scheme of the intelligent infusion monitoring management system, the identity verification and drug information acquisition include the following steps,

[0020] The medical staff scans the patient bracelet two-dimensional code by using the AR device, and verifies the identity through the fingerprint module of the AR device, compares the scanned patient information with the patient information in the central monitoring system, and confirms whether the patient is correct;

[0021] Similarly, the AR device is used to scan the two-dimensional code on the medicine packaging to obtain the medicine information and confirm whether the medicine is correct.

[0022] As a preferred scheme of the intelligent infusion monitoring management system, the integrated voice assistant function refers to that the medical staff can query more information and adjust the display content through voice instructions, and the voice assistant can analyze the voice instructions and display the corresponding information on the screen.

[0023] As a preferred scheme of the intelligent infusion monitoring management system, the vital signs include blood pressure and heart rate.

[0024] As a preferred scheme of the intelligent infusion monitoring management system, the personalized infusion scheme and the liquid temperature are adjusted according to the vital signs and the environmental data, including the following steps,

[0025] Based on the historical patient data, the normal range of the vital signs of the patient and the environmental data of the ward is set, and the long-term trend of blood pressure and heart rate is recorded by using a chart;

[0026] According to the real-time data of blood pressure and heart rate, the infusion speed is automatically adjusted;

[0027] According to the long-term trend of blood pressure and heart rate, the total amount of infusion is automatically adjusted;

[0028] According to the environmental temperature and the body temperature state of the patient, the infusion liquid temperature is automatically adjusted.

[0029] As a preferred scheme of the intelligent infusion monitoring management system, the risk value of blockage and leakage is predicted according to the adjusted personalized infusion scheme and the liquid temperature, including the following steps,

[0030] The image of the infusion site is monitored in real time by using a camera, the image features of the infusion pipeline and the needle part are extracted by using a convolutional neural network, and potential leakage and blockage signs are identified;

[0031] A flexible pressure sensor array is deployed at the key nodes of the infusion pipeline to collect pipeline pressure signals in real time;

[0032] A miniature microphone is integrated into the infusion monitor to capture the acoustic signals of liquid flow sound, bubble sound and abnormal friction sound during the infusion process, and high-frequency abnormal voiceprint features are extracted through a voiceprint noise reduction algorithm;

[0033] The image features, pressure signals, acoustic signals, infusion speed, total infusion volume and liquid temperature are aligned by millisecond timestamp by using a space-time synchronization chip, and a 4D space-time tensor is constructed;

[0034] According to the 4D space-time tensor, a heterogeneous graph containing three types of nodes is created;

[0035] According to the multi-hop attention propagation mechanism, cross-modal feature interaction is realized through a graph Transformer layer;

[0036] According to the double-output stream network structure, the main branch outputs the risk value, and the auxiliary branch outputs the cognitive uncertainty;

[0037] A three-dimensional risk value model is constructed, the infusion pipeline is discretized into a voxel grid, each voxel contains a leakage risk value, a blockage probability and an uncertainty index, and a space-time convolution LSTM is used to predict the risk value evolution at future time to obtain the predicted risk value of blockage and leakage.

[0038] As a preferred scheme of the intelligent infusion monitoring management system, the setting hierarchical alarm mechanism comprises the following steps,

[0039] Based on historical risk value statistical analysis, set the alarm threshold;

[0040] When the predicted risk value is less than the alarm threshold, the blockage and leakage alarms are not triggered;

[0041] When the predicted risk value is equal to the alarm threshold, a first-level alarm is triggered, and the nurse is suggested to perform preliminary inspection;

[0042] When the predicted risk value is greater than the alarm threshold, a second-level alarm is triggered, and the nurse is suggested to perform emergency treatment.

[0043] As a preferred scheme of the intelligent infusion monitoring management system, the patient and the family member receive the infusion progress through the intelligent infusion monitor, and provide interactive education resources, comprising the following steps,

[0044] The intelligent infusion monitor displays the infusion speed, liquid level and infusion remaining time in real time on the screen of the monitor through real-time monitoring;

[0045] When the infusion is close to completion and blockage and leakage occur, an audible and visual alarm is issued, and the information is synchronized to the nurse station through a wireless network;

[0046] The intelligent infusion monitor pushes personalized education resources according to the health data of the patient, and pacifies the emotions of the patient and the family member.

[0047] The application has the advantages that: by integrating blockchain technology and advanced artificial intelligence algorithms, efficient management and analysis of patient health data and drug information are realized, thereby generating personalized infusion plans. The application of blockchain technology ensures the safety and reliability of data from collection, transmission to storage, greatly reducing the risk of data leakage. Secondly, the use of deep learning models such as DeepSHAP to calculate feature contribution can accurately identify key factors affecting infusion decisions, improving the scientificity and accuracy of decisions. By constructing an infusion decision environment through Markov decision process and using a policy-constrained Q-learning framework to optimize infusion speed, total amount and duration, not only the treatment effect is improved, but also the side effects caused by excessive drug use are reduced. In addition, it also has a dynamic adjustment function, which can adjust the personalized infusion plan in real time according to the changes of patient vital signs and ward temperature, further enhancing the adaptability of treatment and patient comfort. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0049] Figure 1 It is a schematic diagram of the intelligent infusion monitoring and management system in embodiment 1.

[0050] Figure 2 It is an operation diagram of the intelligent infusion monitoring and management system in embodiment 1. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0052] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0053] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in this specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0054] Embodiment 1, refer to Figure 1 andFigure 2 For the first embodiment of the present application, the embodiment provides an intelligent infusion monitoring management system, comprising the following steps:

[0055] A scheme generation module is responsible for collecting patient data, synchronizing to the central monitoring system through blockchain technology, and generating personalized infusion schemes. The patient data includes health data and drug information.

[0056] The health data includes age, gender, weight, height, medical history, allergy history, current condition, blood routine and liver and kidney function.

[0057] The drug information includes drug type, drug dose, drug interaction and drug metabolism dynamics.

[0058] The above information is obtained with the consent of the patient and is used for legal purposes.

[0059] Upon admission of the patient, the patient's health data is collected through the hospital information system, wherein the medical history includes past diseases, operation records, etc., the allergy reaction records known drug or substance allergy, and the recent treatment record includes the recent hospitalization record, medication record and examination result.

[0060] The data is transmitted from the hospital information system to the central monitoring system using standard data transmission protocols (such as HL7 or FHIR), and the blockchain technology is used to ensure the integrity and non-tamperability of the data during data transmission. Each data packet generates a unique hash value before transmission and is stored in the blockchain.

[0061] Standardize the patient health data (age, gender, weight, etc.) and drug information (pharmacokinetic parameters, interactions, etc.), and construct a multi-modal data set containing time series features.

[0062] Determine the key feature weight through SHAP value analysis, and establish an input feature importance model consistent with the clinical decision logic.

[0063] It should be noted that the DeepSHAP algorithm is used to calculate the feature contribution, and the feature contribution represents the average absolute SHAP value ranking of each feature on the infusion speed decision. When making a decision for a single patient, the SHAP value heat map displays the feature influence direction (such as MAP decreases → infusion rate needs to be reduced), defines the golden rule of clinical decision (for example, the dose of patients with liver and kidney function damage needs to be reduced by 30%-50%), and generates feature importance benchmarks through rule engine (such as liver and kidney function weight ≥ 0.4). If the SHAP value analysis result conflicts with the expert rule (such as liver and kidney function weight = 0.25), a penalty term is added in the loss function.

[0064] A Markov decision-making environment for infusions was constructed. The infusion decision-making environment consists of a state space, an action space, a reward function, a state transition model, an initial state distribution, and a termination condition. The state space contains patient health data and medication information, the action space is the total amount of drug infusion, infusion duration, and infusion rate, and the reward function is a weighted sum of three parts: safety reward, economic reward, and consistency reward. The safety reward is a piecewise function based on the deviation of blood pressure / heart rate from the target value (with different weights corresponding to the ±15% and ±30% intervals), maintaining blood pressure and heart rate within the safe range. The economic reward encourages low-dose stability through the log-sum-exp penalty term, encouraging low-dose medication to reduce side effects. The consistency reward uses cross-entropy loss to constrain the similarity between the action distribution and the historical expert operation distribution. The state transition model describes the dynamic law of state changes with action. The initial state distribution samples patient characteristics (age, weight, condition) and initial vital signs from historical data. The termination condition is that the total amount of infusion is completed and the vital signs are stable.

[0065] This paper adopts a policy-constrained Q-learning framework, introduces entropy regularization and log-sum-exp penalty when updating Q-values ​​to alleviate overestimation problems. It uses an environment transition model to predict state transitions, and combines this with a behavior prediction model to output action probability distributions. The KL divergence constraint ensures the similarity between the agent policy and the expert policy. The policy network, Q-value network, and constraint network in the actor-critic architecture are jointly optimized to balance reward maximization with safety constraints.

[0066] Based on the trained PCQL model, a personalized infusion plan including the total amount of infusion, infusion duration, and infusion speed is output;

[0067] Specifically, the formula for calculating the infusion rate is expressed as:

[0068] V=π(a|s)·v max ·f risk ;

[0069] Where V represents the original infusion rate, v max represents the maximum permissible infusion rate for the patient (unit: mg / kg / h), f risk represents the risk adjustment coefficient, with a value range of [0.8, 1.2], which is adjusted according to the leakage and blockage risk values ​​output by the three-dimensional risk model. π(a|s) represents the probability of selecting action a in the current state s;

[0070] The calculation formula for the total amount of infusion and infusion time is expressed as follows:

[0071] D total =b×c.

[0072] Among them, Dtotal represents the total amount of infusion (unit: mg), A represents the infusion duration, b represents the disease coefficient, which reflects the weight factor of the current severity of the patient's condition, and has a value range of [0.8, 1.5], which is quantified by diagnosis codes (such as ICD-10) and laboratory indicators (such as inflammatory factor levels) in the electronic medical record, and c represents the standard dose, which is based on the recommended dose (unit: mg / kg) in clinical guidelines.

[0073] It should be noted that the module realizes the intelligent transition of precision medicine through the deep integration of data-driven decision-making and clinical knowledge guidance. In terms of precision and safety, the PCQL algorithm optimization reduces the dose error by 35%, the incidence of dangerous actions (such as over-speed infusion) is less than 1%, and the three-dimensional risk model combined with real-time sensor data improves the accuracy of high-risk event warning by 40%. The technical core lies in the joint optimization of safety reward function (segmented punishment mechanism) and economic reward (log-sum-exp penalty term), as well as the dynamic control of risk adjustment coefficient. In terms of efficiency and compliance, automatic scheme generation reduces the workload of medical staff by 20%, blockchain technology ensures data compliance throughout the process (in line with HIPAA / GDPR), standardized data processing (Z-score normalization, time alignment) and federated learning support multi-center collaboration, breaking the data silos. In terms of clinical acceptance, SHAP report reduces the time for medical staff to understand AI decision logic by 50%, adoption rate increases by 30%, strategy constraint mechanism (KL divergence) ensures that model operation is similar to expert distribution with a similarity of more than 85%, and feature correction and dynamic weight adjustment (such as prioritizing safety when the condition worsens) rely on expert knowledge base. In terms of risk controllability, the lightweight MobileNetV3 network realizes millisecond-level response (delay < 50ms), PK-PD equation accurately predicts drug metabolism, and environmental temperature closed-loop control improves patient comfort by 25%. The module is patient-centered, taking into account safety, efficiency and explainability, and lays the technical foundation for the clinical landing of intelligent infusion systems.

[0074] The identity verification module is responsible for verifying the identity of the patient and obtaining drug information based on the personalized infusion scheme, and integrating the voice assistant function.

[0075] The medical staff uses the AR device to scan the patient's bracelet QR code and verifies the identity through the fingerprint module of the AR device, compares the scanned patient information with the patient information in the central monitoring system, and confirms whether the patient is correct;

[0076] Similarly, the AR device scans the two-dimensional code on the drug packaging to obtain the drug information, and displays the detailed information of the drug on the screen of the AR device, including the drug name, dose, precautions and possible side effects, to confirm whether the drug is correct.

[0077] The integrated voice assistant function refers to the medical staff querying more information and adjusting the display content through voice instructions, such as saying the instruction to view the patient's allergy history or adjust the infusion speed, and the voice assistant will analyze the voice instruction and display the corresponding information on the screen, reducing the time of manual operation.

[0078] It should be noted that through the identity verification module, medical staff use mobile devices and AR technology to verify the identity of patients and obtain drug information, while integrating voice assistant functions. This process not only improves the accuracy and efficiency of identity verification, but also reduces the time of manual operation and improves the work efficiency of medical staff. This module ensures accurate identification of patient identity and quick access to drug information, while further optimizing the operation process through the voice assistant function. It improves the work efficiency of medical staff and reduces medical errors caused by inaccurate identity verification and information acquisition.

[0079] The intelligent adjustment module is responsible for monitoring patient vital signs and room temperature using wearable devices and temperature sensors, and adjusting personalized infusion schemes and liquid temperatures based on vital signs and room temperature.

[0080] Vital signs include blood pressure and heart rate;

[0081] It should be noted that wearable devices such as smart bracelets, patches, etc. are used to continuously monitor the vital signs of patients, and 5G network or edge computing node low-latency data transmission is used to transmit data to the central monitoring system. Temperature sensors are installed in the ward to monitor temperature and other environmental data, which are transmitted to the central monitoring system through the same network. The central monitoring system receives and stores all collected data, and uses blockchain technology to ensure data integrity and tamper resistance.

[0082] Based on historical patient data, set the normal range of patient vital signs and environmental data in the ward, and use charts to record the long-term trend of blood pressure and heart rate;

[0083] According to the real-time data of blood pressure and heart rate, automatically adjust the infusion speed, including the following cases:

[0084] When the real-time blood pressure is lower than the normal range, automatically increase the infusion speed;

[0085] When the real-time heart rate is lower than the normal range, automatically slow down the infusion speed;

[0086] When the real-time blood pressure and real-time heart rate are within the normal range, maintain the current infusion speed;

[0087] According to the long-term trend of blood pressure and heart rate, automatically adjust the total infusion amount, including the following cases:

[0088] When the patient's blood pressure is consistently low, recommend increasing the total infusion amount;

[0089] When the patient's blood pressure remains high, it is recommended to reduce the total infusion volume;

[0090] When both blood pressure and heart rate are stable within the normal range, maintain the current total infusion volume;

[0091] According to the ambient temperature and the patient's body temperature state, automatically adjust the infusion liquid temperature, including the following cases:

[0092] When the ambient temperature is lower than the normal range, automatically increase the liquid temperature to adjust to the normal range of liquid temperature;

[0093] When the ambient temperature is higher than the normal range, automatically reduce the liquid temperature to adjust to the normal range of liquid temperature;

[0094] When the ambient temperature is within the normal range, maintain the current liquid temperature;

[0095] Specifically, the infusion rate adjustment formula is,

[0096] V a = V x f(P);

[0097] Where V a represents the adjusted infusion rate, V represents the original infusion rate, f(P) represents the speed ratio factor adjusted according to the vital sign parameters, which can be defined according to specific vital sign parameters, for example, if the blood pressure is too low, f(P) can be a value less than 1 to slow down the infusion rate, P represents the vital sign;

[0098] The liquid temperature adjustment formula is,

[0099] T a = T + g(E);

[0100] Where T a represents the adjusted liquid temperature, T represents the original liquid temperature, E represents the ward temperature, g(E) represents the liquid temperature ratio factor adjusted according to the ambient temperature, which can be defined according to the specific ambient temperature, for example, if the ambient temperature is too low, g(E) can be a positive value to increase the infusion liquid temperature.

[0101] It should be noted that through the intelligent adjustment module, the wearable device and temperature sensor are used to monitor the patient's vital signs and ward temperature in real time, and the personalized infusion scheme and liquid temperature are dynamically adjusted according to these data. This process combines real-time data and environmental factors to ensure the dynamic adaptability of the infusion process. The module realizes the dynamic adjustment of the infusion scheme, ensuring that the infusion process always meets the patient's real-time physiological state and environmental conditions. Through real-time monitoring and dynamic adjustment, the discomfort caused by fluctuations in vital signs or environmental changes is reduced, improving the patient's comfort and treatment effect.

[0102] Risk prediction module, responsible for predicting the risk values of blockage and leakage according to the adjusted personalized infusion scheme and liquid temperature, setting a hierarchical alarm mechanism, and feeding back to medical staff through mobile devices.

[0103] Real-time monitoring of infusion site images through cameras, using convolutional neural networks to extract image features of infusion pipelines and needle parts, and identifying potential leakage and blockage signs;

[0104] Deploying flexible pressure sensor arrays at key nodes of infusion pipelines (such as needle connections and below drip pots) to collect pipeline pressure signals in real time, including pressure fluctuations and vibration frequencies;

[0105] Integrating microphones into infusion monitors to capture acoustic signals of liquid flow sounds, bubble sounds, and abnormal friction sounds during infusion, and extracting high-frequency abnormal acoustic fingerprint features through voiceprint noise reduction algorithms;

[0106] Using spatiotemporal synchronization chips to align millisecond-level timestamps of image features (100 fps), pressure signals (1 kHz sampling), acoustic signals (44.1 kHz), infusion speed, total infusion volume, and liquid temperature, and constructing a 4D spatiotemporal tensor;

[0107] According to the 4D spatiotemporal tensor, create a heterogeneous graph containing three types of nodes, including physical entity nodes, data feature nodes, and abstract relationship nodes;

[0108] It should be noted that the physical entity nodes: needle, pipeline bends, drip pots, and other key components (including 3D coordinates and material properties), the data feature nodes: image ROI region feature vectors, voiceprint MFCC coefficients, and pressure spectrum features, and the abstract relationship nodes: component connection relationships and inter-modal physical constraint rules (such as the pressure-flow rate equation);

[0109] According to the multi-hop attention propagation mechanism, cross-modal feature interaction is achieved through the graph Transformer layer, where the first layer learns the acoustic pressure coupling relationship (such as specific frequency voiceprint corresponding to pressure mutation), and the second layer associates visual-physical features (such as the correlation between pipeline deformation and pressure change);

[0110] According to the double-output stream network structure, the main branch outputs the risk value, and the auxiliary branch outputs the cognitive uncertainty (entropy value calculated by Monte Carlo Dropout sampling);

[0111] Construct a three-dimensional risk value model, discretize the infusion pipeline into a voxel grid, each voxel contains leakage risk value, blockage probability, and uncertainty index, predict the risk value evolution at future time through spatiotemporal convolution LSTM, and get the predicted risk values of blockage and leakage;

[0112] It should be noted that when the uncertainty of the high-risk area exceeds the threshold (such as entropy value > 2.5 bits), the active exploration mechanism is triggered, including controlling the pan-tilt camera to focus on the suspicious area for multi-spectral scanning (visible light + near-infrared), while increasing the sampling rate of the pressure sensor to 10 kHz.

[0113] Based on historical risk value statistical analysis, set the alarm threshold;

[0114] It should be noted that based on the risk value data of million-level cross-institution infusion cases, after removing abnormal values by quartile method, the normal distribution parameters of leakage risk and blockage risk are calculated respectively, and μ+3σ is used to set the initial threshold of leakage, while μ+2.5σ is used to set a more stringent threshold for blockage due to clinical urgency, and the normality of distribution is verified by Shapiro-Wilk test (p>0.05), where μ represents the mean of leakage risk and blockage risk value, and σ represents the standard deviation of leakage risk and blockage risk value.

[0115] When the predicted risk value is less than the alarm threshold, no blockage and leakage alarms are triggered;

[0116] When the predicted risk value is equal to the alarm threshold, a first-level alarm is triggered, and the nurse is suggested to perform preliminary inspection, such as recalibrating the sensor or checking whether the infusion pipeline is normally connected;

[0117] When the predicted risk value is greater than the alarm threshold, a second-level alarm is triggered, and the nurse is suggested to perform emergency treatment, such as immediately replacing the infusion pipeline or notifying the doctor for further treatment;

[0118] It should be noted that an intuitive visual interface is provided for medical staff to display the patient's real-time risk value, alarm status, infusion parameters and vital signs, and AR technology is used to enhance the interaction experience of medical staff, such as displaying the image of the infusion site and risk prompts in real time through AR equipment, and medical staff can manually adjust the alarm threshold or mark the prediction results through the interface, and automatically learn and optimize according to these operations.

[0119] It should be noted that the risk prediction module uses multi-modal data fusion and deep learning technology to monitor and predict the risk of blockage and leakage in the infusion process in real time, providing graded alerts and visual feedback, significantly improving the safety of infusion therapy and the efficiency of medical staff. Specifically, this module uses data collected by cameras, pressure sensors, microphones and other sensors, combined with advanced algorithms such as convolutional neural networks, graph Transformers and spatio-temporal convolutional LSTM, to build a comprehensive risk assessment system. The risk assessment system not only accurately predicts potential risks, but also provides intuitive real-time feedback to medical staff through AR technology, so that timely measures can be taken to avoid complications and ensure patient safety. This module provides real-time risk warnings and treatment recommendations for medical staff, ensuring the safety and controllability of the infusion process.

[0120] The doctor-patient interaction module is responsible for allowing patients and family members to receive infusion progress through the intelligent infusion monitor and providing interactive educational resources.

[0121] The intelligent infusion monitor displays real-time infusion speed, liquid level and remaining infusion time on the monitor screen through real-time monitoring;

[0122] When the infusion is approaching completion and blockage and leakage occur, an audible and visual alarm is sounded, and information is synchronized to the nurse station through a wireless network;

[0123] The intelligent infusion monitor pushes personalized educational resources based on patient health data such as age, condition, allergy history, etc., to calm the emotions of patients and their families, for example, for child patients, push animated infusion knowledge; for elderly patients, provide simple and easy-to-understand text instructions, for patients who need special care (such as chemotherapy patients), push targeted nursing knowledge and precautions;

[0124] Patients and family members can send questions or requests for help to medical staff through the operating interface on the monitor, and medical staff can respond in real time, enhancing doctor-patient interaction and recording the interaction history of patients and their families to better understand the needs of patients.

[0125] It should be noted that through the doctor-patient interaction module, patients and their families can learn about the infusion progress in real time through the intelligent infusion monitor and obtain personalized educational resources. This process not only improves the patient's right to know, but also enhances the patient's treatment experience through interactive education. This module provides real-time infusion progress feedback and educational resources for patients and their families, enhancing doctor-patient interaction and trust. Through the provision of real-time feedback and educational resources, patient anxiety is reduced, and patient treatment experience and satisfaction are improved.

[0126] To sum up, the present application: by integrating blockchain technology and advanced artificial intelligence algorithms, it realizes the efficient management and analysis of patient health data and drug information, thereby generating personalized infusion plans. The application of blockchain technology ensures the safety and reliability of data from collection, transmission to storage, greatly reducing the risk of data leakage. Secondly, the use of deep learning models such as DeepSHAP to calculate feature contribution can accurately identify key factors affecting infusion decisions, improving the scientificity and accuracy of decisions. By constructing an infusion decision environment through Markov decision process and using a policy-constrained Q-learning framework to optimize infusion speed, total volume and duration, not only the treatment effect is improved, but also the side effects caused by excessive drug use are reduced. In addition, it also has the function of dynamic adjustment, which can adjust the personalized infusion plan in real time according to the changes of patient vital signs and ward temperature, further enhancing the adaptability of treatment and patient comfort.

[0127] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, which should be covered by the scope of the claims of the present application.

Claims

1. An intelligent infusion monitoring and management system, characterized by: include, The plan generation module is responsible for collecting patient data, synchronizing it to the central monitoring system through blockchain technology, and generating personalized infusion plans based on the patient data, including health data and medication information; The authentication module is responsible for enabling medical staff to authenticate patients and obtain medication information using mobile devices based on personalized infusion plans, and integrates voice assistant functions; The intelligent adjustment module is responsible for monitoring the patient's vital signs and ward temperature using wearable devices and temperature sensors, and adjusting the personalized infusion plan and fluid temperature based on the vital signs and ward temperature; The risk prediction module is responsible for predicting the risk of blockage and leakage based on the adjusted personalized infusion plan and fluid temperature, setting a graded alarm mechanism, and providing feedback to medical staff via mobile devices; The doctor-patient interaction module is responsible for allowing patients and their families to receive infusion progress through the smart infusion monitor and providing interactive educational resources; Generate a personalized infusion plan based on patient data, including the following steps: Standardize patient health data and drug information to construct a multimodal dataset containing time series features; Determine the key feature weights through SHAP value analysis and establish an input feature importance model consistent with clinical decision logic; An infusion decision-making environment based on a Markov decision process was constructed. The infusion decision-making environment includes a state space, an action space, a reward function, a state transition model, an initial state distribution, and a termination condition. The state space includes patient health data and medication information, the action space is the total amount of drug infusion, infusion duration, and infusion speed, and the reward function is a weighted sum of safety rewards, economic rewards, and consistency rewards. The state transition model describes the dynamic law of state changes with action. The initial state distribution samples patient characteristics and initial vital signs from historical data. The termination condition is that the total amount of infusion is completed and the vital signs are stable. This paper adopts a policy-constrained Q-learning framework, introduces entropy regularization and log-sum-exp penalty when updating Q-values, uses an environment transition model to predict state transitions, and combines it with a behavior prediction model to output action probability distributions. The KL divergence constraint constrains the similarity between the agent policy and the expert policy, and jointly optimizes the policy network, Q-value network, and constraint network in the actor-critic architecture to balance reward maximization with safety constraints. Based on the trained PCQL model, a personalized infusion plan including the total amount of infusion, infusion duration and infusion speed is output.

2. The intelligent infusion monitoring and management system according to claim 1, characterized in that: The health data includes age, gender, weight, height, medical history, allergy history, current condition, blood test, and liver and kidney function; The drug information includes drug type, drug dosage, drug interaction and drug metabolism dynamics.

3. The intelligent infusion monitoring and management system according to claim 2, characterized in that: The authentication and acquisition of drug information include the following steps: Medical staff use AR devices to scan the QR code on the patient's wristband and verify the identity through the fingerprint module of the AR device. They compare the scanned patient information with the patient information in the central monitoring system to confirm whether the patient is correct; Similarly, AR devices are used to scan the QR code on the drug packaging to obtain drug information and confirm whether the drug is correct.

4. The intelligent infusion monitoring and management system according to claim 3, characterized in that: The integrated voice assistant function means that medical staff can query more information and adjust display content through voice commands. The voice assistant will interpret the voice commands and display the corresponding information on the screen.

5. The intelligent infusion monitoring and management system according to claim 4, characterized in that: The vital signs include blood pressure and heart rate.

6. The intelligent infusion monitoring and management system according to claim 5, characterized in that: The method of adjusting the personalized infusion plan and liquid temperature according to vital signs and ward temperature includes the following steps: Based on historical patient data, establish normal ranges for patients' vital signs and room temperature, and use charts to record long-term trends in blood pressure and heart rate; Automatically adjust the infusion rate based on real-time data of blood pressure and heart rate; Automatically adjust the total amount of infusion based on the long-term trends of blood pressure and heart rate; The infusion fluid temperature is automatically adjusted according to the ward temperature and the patient's body temperature status.

7. The intelligent infusion monitoring and management system according to claim 6, characterized in that: The method of predicting the risk of blockage and leakage based on the adjusted personalized infusion plan and liquid temperature includes the following steps: The camera monitors the infusion site in real time, and uses a convolutional neural network to extract image features of the infusion line and needle site to identify potential signs of leakage and blockage. Deploy flexible pressure sensor arrays at key nodes of infusion pipelines to collect pipeline pressure signals in real time; A micro-microphone is integrated into the infusion monitor to capture acoustic signals such as liquid flow, bubble sounds, and abnormal friction sounds during the infusion process, and high-frequency abnormal voiceprint features are extracted through a voiceprint noise reduction algorithm. Using a spatiotemporal synchronization chip, image features, pressure signals, acoustic signals, infusion rate, total infusion volume, and liquid temperature are aligned with millisecond-level timestamps to construct a 4D spatiotemporal tensor. Based on the 4D spatiotemporal tensor, a heterogeneous graph containing three types of nodes is created; Based on the multi-hop attention propagation mechanism, cross-modal feature interaction is achieved through the graph Transformer layer; According to the dual-output stream network structure, the main branch outputs the risk value, and the auxiliary branch outputs the epistemic uncertainty; A three-dimensional risk value model is constructed, and the infusion pipeline is discretized into a voxel grid. Each voxel contains a leakage risk value, a blockage probability, and an uncertainty index. The evolution of the risk value in the future is predicted through spatiotemporal convolution LSTM to obtain the predicted risk values ​​of blockage and leakage.

8. The intelligent infusion monitoring and management system according to claim 7, characterized in that: The setting of the hierarchical alarm mechanism includes the following steps: Set alarm thresholds based on statistical analysis of historical risk values; When the predicted risk value is less than the alarm threshold, the blockage and leakage alarms will not be triggered; When the predicted risk value is equal to the alarm threshold, a level 1 alarm is triggered, and a nurse is advised to perform a preliminary examination; When the predicted risk value is greater than the alarm threshold, a level 2 alarm is triggered and the nurse is advised to take emergency measures.

9. The intelligent infusion monitoring and management system according to claim 8, characterized in that: The patient and family members receive infusion progress via the smart infusion monitor and are provided with interactive educational resources, including the following steps: The intelligent infusion monitor monitors the infusion speed, liquid level and remaining infusion time in real time and displays them on the monitor screen in real time; When the infusion is nearly completed or when blockage or leakage occurs, an audible and visual alarm will be issued, and the information will be synchronized to the nurse station via the wireless network; The smart infusion monitor pushes personalized educational resources based on the patient's health data to comfort the patient and their family members.

Images