Mobile infusion safety monitoring and remote early warning system and method based on edge calculation

Abstract

The invention belongs to the technical field of biomedical engineering, and discloses a mobile infusion safety monitoring and remote early warning system and method based on edge computation.The mobile infusion safety monitoring and remote early warning system comprises a data stream monitoring unit, a data processing unit and a remote early warning unit, the data stream monitoring unit is used for collecting infusion dripping speed data of a patient in an ICU inpatient area and converting the infusion dripping speed data into dripping speed digital signals through an ADC module arranged in an infusion support edge node; distributing along the vertical direction of the infusion container, dynamically monitoring the residual liquid amount, and eliminating shaking interference in combination with a compensation algorithm to obtain the accurate residual liquid amount; the ICU central monitoring terminal is synchronously connected to obtain physical sign data of the patient; the edge intelligent analysis unit is used for predicting the infusion remaining time according to the accurate remaining liquid amount; acquiring infusion medicine types, and dynamically sequencing sickbed lists on a nurse bracelet interface in combination with infusion remaining time; starting a zero call mode for a common inpatient area, cancelling all local sound-light alarms, and only retaining vibration alarm information sent by a nurse bracelet; the infusion monitoring reliability is improved, and the burden of medical staff is reduced.

Description

Technical Field

[0001] The present invention relates to the field of biomedical engineering technology, and more specifically, to a mobile infusion safety monitoring and remote early warning system and method based on edge computing. Background Art

[0002] The patent with patent announcement number CN112750517A discloses a method and system for tracking intravenous treatment informationization, which initializes the intravenous treatment informationization system; the nurse role submits an intravenous treatment application and sends the intravenous treatment application to the intravenous treatment approval module; the attending physician role approves the intravenous treatment application through the intravenous treatment approval module; the hydraulic sensor set on the infusion tube starts to collect hydraulic values, and when the hydraulic value is less than the set hydraulic threshold, it is judged that the infusion is about to end, and the infusion monitoring module sends an early warning message to the mobile device of the nurse role; the end of the intravenous infusion is registered; access to the database on the Internet is realized, and the catheterization status and patient information of each hospital are uniformly managed and monitored, and the information of catheterized patients is shared, reducing the problems of low automation and low efficiency in traditional management.

[0003] The existing mobile infusion safety monitoring and remote early warning systems and methods have the following main problems:

[0004] Existing capacitive liquid level sensors usually use fixed judgment factors to determine the liquid level and cannot be dynamically adjusted for liquids with different viscosities. Since tiny bubbles may be generated during the infusion process and adhere to the surface of the capacitive sensor electrode, the capacitance value fluctuates for a short time, causing the system to misjudge the bubbles as liquid, affecting the accuracy of the calculation of the remaining liquid volume. Traditional liquid level detection methods are either too sensitive or have insufficient anti-interference capabilities, and it is impossible to achieve a dynamic balance between sensitivity and anti-interference capabilities. The fixed judgment factor method cannot be adaptively adjusted for different infusion environments. Traditional liquid level monitoring is difficult to meet the high-precision monitoring requirements in ICU wards, especially for infusion of critically ill patients. Misjudgment may lead to abnormal liquid supply and affect the treatment effect. For high-viscosity liquids such as fat emulsions, the liquid adheres to the infusion container or sensor surface to form a liquid film residue, causing the capacitance value to drop slowly, causing the system to misjudge that the liquid level has not dropped, affecting the calculation of the actual remaining infusion volume.

[0005] In view of this, the present invention proposes a mobile infusion safety monitoring and remote early warning system and method based on edge computing to solve the above problems. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned objectives, the present invention provides the following technical solution: a mobile infusion safety monitoring and remote early warning system based on edge computing, comprising:

[0007] The data stream monitoring unit collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the ADC module built into the edge node of the infusion stand. It is distributed vertically along the infusion container to dynamically monitor the remaining liquid volume, using a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It also simultaneously connects to the ICU central monitoring terminal to obtain patient vital sign data.

[0008] The edge intelligent analysis unit predicts the remaining infusion time based on accurate remaining fluid volume data; obtains the infusion drug type and, combined with the remaining infusion time, dynamically sorts the bed list on the nurse wristband interface; and enables zero-call mode for general wards, canceling all local sound and light alarms and retaining only the vibration alarm information from the nurse wristband;

[0009] The low-latency communication unit uses slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and defines a three-level QoS policy for dynamic transmission adjustment;

[0010] The alarm interaction unit displays the bed location through a circular LED array and distinguishes the urgency of the vibration alarm information through the color of the LED light strip. If the nurse in the ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

[0011] Preferably, the method for collecting infusion drip rate data of ICU patients includes:

[0012] Using electromagnetic interference-resistant infrared array sensors, adapted to the ICE electromagnetic environment, to collect infusion drip rate data for patients in the ICU ward, the electromagnetic interference-resistant infrared array sensors are installed around the drip bucket of the infusion tube. The infrared array sensors use an n×n infrared transmitter-receiver array layout.

[0013] The array layout rules are defined, including the transmitter, receiver, array arrangement, and vertical spacing. The transmitter and receiver are located on either side of the drip bucket of the infusion tube, forming a penetrating optical path. The transmitter uses a narrow-band infrared LED for transmission, which is focused by a collimating lens to form a parallel beam.

[0014] The receiving end uses a PIN photodiode with a built-in preamplifier and spectral response filter. The array is arranged as p×p detection points, forming a grid-like three-dimensional detection network covering the entire droplet passing area. The center of the grid is aligned with the droplet outlet. The vertical spacing is defined as the distance between adjacent transmitting and receiving pairs, and is designed to be m millimeters.

[0015] Infusion drip rate data includes basic drip rate data and infusion environment data; basic drip rate data includes instantaneous drip rate, average drip rate, drip rate fluctuation and drip rate change rate; infusion environment data includes ward temperature, drip time and infusion tube pressure.

[0016] Preferably, the method of converting into a dripping rate digital signal includes:

[0017] The receiving end of the deployed electromagnetic interference-resistant infrared array sensor detects the light intensity signal reaching the receiving end after being blocked or scattered by droplets when passing through the infusion tube funnel. The light intensity signals detected at different detection points are weighted averaged to extract the optimal dripping rate signal.

[0018] An anti-interference filtering algorithm is added to the front end of the built-in ADC module in the edge node of the infusion stand to optimize the signal's anti-interference capability. The ADC module converts the optimal drip rate signal after anti-interference at a preset sampling rate greater than vHz, and collects timestamp information to convert the optimal drip rate signal (analog voltage signal) into a digital drip rate signal.

[0019] Preferably, the method for dynamically monitoring the remaining liquid volume includes:

[0020] Arrange K capacitive liquid level sensors vertically along the infusion container to form a liquid level detection array, which is attached to the outer wall of the infusion container. The liquid level detection array covers the entire liquid level range from the bottom to the top of the container;

[0021] The initial capacitance value of each capacitive liquid level sensor measured in the absence of liquid is recorded as , the saturated capacitance value of each capacitive liquid level sensor measured when the infusion container is full is recorded as , construct the capacitance-liquid level relationship model of each capacitive liquid level sensor: ;in, Indicates the capacitance value at the current liquid level; Indicates the current liquid level; Indicates the total height of the infusion container;

[0022] Each capacitive sensor measures the capacitance value of the liquid in real time, and the real-time capacitance value measured by each capacitive liquid level sensor is recorded as ; Determine whether there is liquid at any liquid level height. If , it is determined that there is liquid at this liquid level height; Indicates the judgment factor; calculates the current liquid level height through the highest triggered capacitance sensor position;

[0023] The judgment factor is dynamically adjusted through the judgment factor adjustment formula. The judgment factor adjustment formula is: ;in, Represents the judgment factor after dynamic adjustment; Indicates the current liquid level drop rate; Indicates the maximum rate of liquid level drop; Indicates the standard deviation of the current capacitance value; Indicates the maximum standard deviation of capacitance value; represents the rate weight coefficient; represents the capacitance fluctuation weight coefficient;

[0024] The current liquid level drop rate Multiply by the dripping time t to get the total height change of the liquid level drop , the cross-sectional area of ​​the preset infusion container is , the maximum liquid level height at the start of infusion Total height change minus the drop in liquid level The difference is multiplied by the cross-sectional area of ​​the infusion container to obtain the remaining liquid volume.

[0025] Preferably, the method for obtaining the precise remaining liquid volume includes:

[0026] During the preset time period of g, the remaining liquid volume data is collected and smoothed by sliding window filtering to eliminate shaking interference; the total height change threshold of the liquid level drop is preset, and when the total height change of the liquid level drop When the total height change of the liquid level drop is greater than or equal to the preset threshold, it is determined that the total height change of the liquid level drop is disturbed by shaking, and the remaining liquid volume data is abnormal. The abnormal remaining liquid volume data will be eliminated, and the remaining liquid volume will be continuously updated to obtain accurate remaining liquid volume data.

[0027] Preferably, the method for obtaining patient vital sign data includes:

[0028] A protocol conversion adapter is deployed at the edge node of the infusion stand, which is connected to the ICU central monitoring terminal through the ICU intranet. The real-time data stream of the ICU central monitoring terminal is parsed into structured data, and the open API of the ICU central monitoring system is called to authenticate the patient's vital signs data; patient vital signs data include heart rate, blood pressure, blood oxygen saturation, respiratory rate, body temperature, hemodynamic data, brain function and metabolic indicators.

[0029] Preferably, the method for obtaining the deviation index includes:

[0030] Build and train a deviation index prediction model using the deep learning library TensorFlow. The deviation index prediction model consists of an input layer, an LSTM layer, and an output layer. The input layer is used to input historical accurate remaining fluid volume data, and the output layer is used to output the remaining infusion time. The deviation index prediction model is an LSTM model.

[0031] Define the model's loss function and use the L2 regularized mean square error loss function to measure the difference between the predicted remaining infusion time and the actual remaining infusion time; use the trained precise remaining liquid volume data to predict the current precise remaining liquid volume data to obtain the remaining infusion time.

[0032] Preferably, the method for dynamically sorting the bed list includes:

[0033] The types of infused drugs include antibiotics, nutritional support drugs, chemotherapy drugs, analgesics, hormone drugs, antivirals, anticoagulants, immunoglobulins, antiallergic drugs, and detoxification drugs;

[0034] Obtain the infusion drug type and dynamically sort the bed list on the nurse wristband interface based on the remaining infusion time; sort the bed list on the nurse wristband interface based on the patient's remaining infusion time and infusion drug type, and define sorting rules. The sorting rules include priority for the shortest remaining time, drug priority, bed distance, and patient health status.

[0035] Preferably, the method for performing dynamic transmission adjustment includes:

[0036] Real-time perception of the current network status, obtaining bandwidth utilization, link round-trip delay, and packet loss rate, and classifying network status into levels of normal, congested, and faulty. Based on the network status level, different adjustment strategies are triggered to dynamically adjust the transmission of vibration alarm information and the sorted bed list.

[0037] If the network status level is normal, the data is transmitted through the 5G slice primary path, the LoRaWAN backup path is dormant, and the vibration alarm information is transmitted to the ICU central monitoring terminal. The sorted bed list is updated every 5 seconds through the ZigBee mesh network;

[0038] If the network status level is congested, the sorted bed list is first compressed using Delta coding and then updated through the ZigBee mesh network. The preemptive mechanism of the time-sensitive network is triggered, interrupting the transmission of low-priority data, and automatically activating the LoRaWAN backup path to transmit the vibration alarm information to the ICU central monitoring terminal.

[0039] If the network status level is failure, the 5G slice will be shut down, and all data will be transmitted through a mixed ZigBee and LoRaWAN network. The vibration alarm will be split into N LoRa data packets, and Fountain code will be used for rate-free encoding to ensure that the information can be reassembled when any half of the packet arrives. If the local gateway is paralyzed, the data will be relayed through the ZigBee mesh network of the adjacent ward.

[0040] The mobile infusion safety monitoring and remote early warning method based on edge computing includes:

[0041] S1. Collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the built-in ADC module of the infusion stand edge node. It is distributed vertically along the infusion container, dynamically monitors the remaining liquid volume, and uses a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It is also synchronously connected to the ICU central monitoring terminal to obtain patient vital sign data.

[0042] S2. Predict the remaining time of infusion based on accurate remaining fluid volume data; obtain the infusion drug type and, combined with the remaining infusion time, dynamically sort the bed list on the nurse wristband interface; enable zero call mode for general wards, cancel all local sound and light alarms, and retain only the vibration alarm information from the nurse wristband;

[0043] S3: Use slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and define a three-level QoS strategy for dynamic transmission adjustment;

[0044] S4. The bed position is displayed through a circular LED array, and the urgency of the vibration alarm information is distinguished by the color of the LED light strip. If the nurse in this ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention dynamically reduces the judgment factor. When the system detects a rapid drop in the liquid level, the judgment factor is automatically reduced, thereby improving sensitivity and making it easier for the system to detect a drop in the actual liquid level. It solves the measurement lag problem caused by the residual effect of the liquid film and improves the detection accuracy of high-viscosity liquid medicines; ensures that the system can respond in a timely manner when the liquid level drops, avoiding misjudgment that the infusion is still ongoing; is particularly suitable for viscous liquid medicines such as fat emulsions and albumin, and improves the versatility of liquid level detection. The judgment factor is dynamically increased. When a large capacitance fluctuation is detected, the judgment factor is automatically increased, making the system insensitive to capacitance fluctuations in a short period of time, thereby reducing misjudgments caused by bubbles; using standard deviation analysis, it can distinguish between normal liquid level changes and short-term fluctuations caused by bubbles, so that the system can remain stable in complex infusion environments. Enhances the ability to resist bubble interference and avoids false liquid level detection caused by bubble attachment. Improves the accuracy of residual liquid volume calculation, reduces false alarms, and improves the reliability of infusion monitoring.

[0047] By comprehensively calculating the rate of liquid level drop and capacitance fluctuation, an intelligent adjustment judgment factor is implemented. This improves sensitivity in low-noise environments and enhances anti-interference capabilities in noisy environments, allowing the system to adapt to different infusion environments. This ensures a controlled and stable adjustment range, preventing measurement errors caused by over-adjustment. Combined with intelligent adjustment of the judgment factor, this method achieves more accurate liquid level detection, making ICU infusion monitoring more reliable. By optimizing liquid level calculation accuracy, false alarms are reduced, and the accuracy of nurse wristband alarms is improved, reducing the burden on medical staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a schematic diagram of the structure of the mobile infusion safety monitoring and remote early warning system based on edge computing of the present invention;

[0049] Figure 2 This is a flow chart of the mobile infusion safety monitoring and remote early warning method based on edge computing of the present invention. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] Example 1

[0052] See also Figure 1 As shown, this embodiment 1 further illustrates the mobile infusion safety monitoring and remote warning system based on edge computing proposed by the present invention, including:

[0053] Infusion therapy is a widely used treatment method in clinical medicine, especially in intensive care units (ICUs). Accurate infusion monitoring is crucial to patient safety and treatment outcomes. With the continuous advancement of medical technology, infusion therapy is gradually shifting from traditional manual management to intelligent, automated management, requiring systems with high precision, high reliability, and real-time performance. However, existing infusion monitoring systems still face numerous challenges in practical application, particularly in balancing high-precision monitoring with environmental interference.

[0054] Key issues include:

[0055] Disadvantages of capacitive liquid level sensors: Traditional capacitive liquid level sensors use a fixed factor to determine liquid level, a method that is incapable of handling liquids of varying viscosities. In practical applications, the flow characteristics of high-viscosity liquids (such as fat emulsions and albumin) differ significantly from those of low-viscosity liquids, leading to inaccurate measurements. Furthermore, the residual effect of the liquid film slows down the capacitance value, hindering the timely feedback of liquid level changes and, in turn, affecting the accuracy of remaining liquid volume calculations.

[0056] Bubble interference: Tiny bubbles can form during infusion and adhere to the capacitive sensor electrodes, causing short-term fluctuations in capacitance. Because the short-term fluctuations in bubbles resemble those in liquid, the system struggles to distinguish between bubbles and liquid, leading to misjudgments and inaccurate calculations of the remaining liquid volume.

[0057] The contradiction between sensitivity and anti-interference ability: Traditional liquid level detection methods are often too sensitive or have insufficient anti-interference ability, and cannot achieve dynamic balance in different infusion environments, especially in high-interference ICU wards, which may lead to incorrect liquid level judgment and affect the treatment effect.

[0058] Existing liquid level monitoring systems typically use a fixed judgment factor approach. However, due to the variability of different infusion environments, this approach lacks adaptive adjustment, resulting in poor system adaptability to varying liquids and environmental conditions. Fixed judgment factors cannot dynamically adjust to factors such as the rate of liquid level drop, liquid viscosity, or bubble interference, leading to false alarms or missed alerts.

[0059] In order to effectively solve the above problems, the present invention proposes a mobile infusion safety monitoring and remote early warning system based on edge computing, including:

[0060] The data stream monitoring unit collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the ADC module built into the edge node of the infusion stand. It is distributed vertically along the infusion container to dynamically monitor the remaining liquid volume, using a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It also simultaneously connects to the ICU central monitoring terminal to obtain patient vital sign data.

[0061] The edge intelligent analysis unit predicts the remaining infusion time based on accurate remaining fluid volume data; obtains the infusion drug type and, combined with the remaining infusion time, dynamically sorts the bed list on the nurse wristband interface; and enables zero-call mode for general wards, canceling all local sound and light alarms and retaining only the vibration alarm information from the nurse wristband;

[0062] The low-latency communication unit uses slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and defines a three-level QoS policy for dynamic transmission adjustment;

[0063] The alarm interaction unit displays the bed location through a circular LED array and distinguishes the urgency of the vibration alarm information through the color of the LED light strip. If the nurse in the ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

[0064] The method for collecting the infusion drip rate data of patients in the ICU ward includes:

[0065] Using electromagnetic interference-resistant infrared array sensors, adapted to the ICE electromagnetic environment, to collect infusion drip rate data for patients in the ICU ward, the electromagnetic interference-resistant infrared array sensors are installed around the drip bucket of the infusion tube. The infrared array sensors use an n×n infrared transmitter-receiver array layout.

[0066] The array layout rules are defined, including the transmitter, receiver, array arrangement, and vertical spacing. The transmitter and receiver are located on either side of the drip bucket of the infusion tube, forming a penetrating optical path. The transmitter uses a narrow-band infrared LED for transmission, which is focused by a collimating lens to form a parallel beam.

[0067] The receiving end uses a PIN photodiode with a built-in preamplifier and spectral response filter. The array is arranged as p×p detection points, forming a grid-like three-dimensional detection network covering the entire droplet passing area. The center of the grid is aligned with the droplet outlet. The vertical spacing is defined as the distance between adjacent transmitting and receiving pairs, and is designed to be m millimeters.

[0068] Infusion drip rate data includes basic drip rate data and infusion environment data; basic drip rate data includes instantaneous drip rate, average drip rate, drip rate fluctuation and drip rate change rate; drip rate fluctuation is the standard deviation of the drip rate; infusion environment data includes ward temperature, drip time and infusion tube pressure.

[0069] The methods for converting to drip rate digital signal include:

[0070] The receiving end of the deployed electromagnetic interference-resistant infrared array sensor detects the light intensity signal reaching the receiving end after being blocked or scattered by droplets when passing through the infusion tube funnel. The light intensity signals detected at different detection points are weighted averaged to extract the optimal dripping rate signal.

[0071] An anti-interference filtering algorithm is added to the front-end of the built-in ADC module at the edge node of the infusion stand to optimize the signal's anti-interference capability. The front-end of the ADC module is the early stage in the signal processing flow, that is, the part close to the signal source. The ADC module converts the optimal drip rate signal after anti-interference at a preset sampling rate greater than vHz, and collects timestamp information to convert the optimal drip rate signal (analog voltage signal) into a digital drip rate signal.

[0072] Methods for dynamically monitoring the remaining liquid volume include:

[0073] Arrange K capacitive liquid level sensors vertically along the infusion container to form a liquid level detection array, which is attached to the outer wall of the infusion container. The liquid level detection array covers the entire liquid level range from the bottom to the top of the container;

[0074] The initial capacitance value of each capacitive liquid level sensor measured in the absence of liquid is recorded as , the saturated capacitance value of each capacitive liquid level sensor measured when the infusion container is full is recorded as , construct the capacitance-liquid level relationship model of each capacitive liquid level sensor: ;in, Indicates the capacitance value at the current liquid level; Indicates the current liquid level; Indicates the total height of the infusion container;

[0075] Each capacitive sensor measures the capacitance value of the liquid in real time, and the real-time capacitance value measured by each capacitive liquid level sensor is recorded as ; Determine whether there is liquid at any liquid level height. If , it is determined that there is liquid at this liquid level height; Represents the judgment factor, which is used to determine whether the capacitive sensor is covered by liquid; The smaller it is, the more sensitive the detection is (easy to detect low liquid levels), but the risk of false alarms increases; The larger the value, the stronger the anti-interference, but it may miss the detection of viscous liquid; the current liquid level is calculated by the position of the highest triggered capacitance sensor;

[0076] Now considering that some high viscosity liquids (such as fat emulsions) may have residual liquid film, which causes the capacitance value to decrease slowly, if the judgment factor Is fixed, the actual liquid level has dropped, but the capacitance is still higher than the fixed judgment factor, the system will judge that there is liquid; bubbles attached to the electrode will cause the capacitance value to fluctuate. If the judgment factor The system will mistake bubbles for liquid, causing errors in the calculation of the remaining liquid volume.

[0077] Existing capacitive liquid level sensors typically use a fixed judgment factor to determine the liquid level, and are unable to dynamically adjust for liquids of different viscosities. During the infusion process, tiny bubbles may form and adhere to the surface of the capacitive sensor electrode, causing the capacitance value to fluctuate briefly. This can lead the system to misjudge the bubbles as liquid, affecting the accuracy of calculating the remaining liquid volume. Traditional liquid level detection methods are either overly sensitive (leading to false alarms) or lack anti-interference capabilities (leading to missed detections), and are unable to achieve a dynamic balance between sensitivity and anti-interference capabilities. The fixed judgment factor method cannot adaptively adjust to different infusion environments (such as patient movement, different liquid properties, and changes in infusion speed). Traditional liquid level monitoring is difficult to meet the high-precision monitoring requirements in ICU wards, especially for critically ill patients receiving infusions. Misjudgments may lead to abnormal liquid supply and affect the treatment effect.

[0078] The judgment factor is dynamically adjusted through the judgment factor adjustment formula to increase the accuracy of the detection. The judgment factor adjustment formula is: ;in, Represents the judgment factor after dynamic adjustment; Indicates the current liquid level drop rate; Indicates the maximum rate of liquid level drop; Indicates the standard deviation of the current capacitance value (indicates the degree of fluctuation of the capacitance value); Indicates the maximum standard deviation of capacitance value; represents the rate weight coefficient; represents the capacitance fluctuation weight coefficient;

[0079] It should be noted that when the liquid level drops quickly ( If the value is larger, it means that the liquid is decreasing rapidly. In this case, the judgment factor should be reduced. , improve the detection sensitivity and prevent the residual liquid film from causing the false judgment that the liquid is still there. If the value is larger, it indicates that there may be air bubble interference, and the judgment factor should be increased. , enhance anti-interference ability and avoid misjudging bubbles as liquid; if the liquid level drop rate is close to 0 ( ),but Maintain Nearby, to avoid over-sensitivity leading to false alarms; if the capacitance does not fluctuate, then Not affected by interference, ensuring detection accuracy;

[0080] The judgment factor adjustment formula refers to the viscosity effect in fluid mechanics. The residual effect of the boundary layer formed by viscous liquid on the wall causes the liquid level to drop slowly. The idea here is similar to the adaptive threshold adjustment algorithm (such as the gradient threshold method). It also refers to the noise filtering principle in signal processing, similar to the adaptive gain adjustment strategy of the dynamic Kalman filter, and adjusts the detection parameters according to the noise intensity.

[0081] The beneficial effects relative to the prior art are:

[0082] Compared with existing liquid level detection methods, this solution has significant improvements and enhancements, especially in high-viscosity liquid detection and bubble interference suppression. The specific advantages are as follows:

[0083] For high-viscosity liquids such as fat emulsions, the liquid adheres to the infusion container or sensor surface, forming a residual liquid film. This causes the capacitance value to decrease slowly, leading the system to mistakenly judge that the liquid level has not decreased, affecting the calculation of the actual remaining infusion volume. Dynamically reducing the judgment factor: When the system detects a rapid drop in the liquid level, it automatically lowers the judgment factor, increasing sensitivity and making it easier for the system to detect the actual drop in the liquid level. This solves the measurement lag caused by the residual liquid film effect and improves the detection accuracy of high-viscosity liquids. It ensures that the system can respond promptly when the liquid level drops, avoiding the misjudgment that the infusion is still continuing. It is particularly suitable for viscous liquids such as fat emulsions and albumin, improving the versatility of liquid level detection.

[0084] Dynamically increase the judgment factor. When large capacitance fluctuations are detected, the judgment factor is automatically increased, making the system less sensitive to short-term capacitance fluctuations and reducing false positives caused by bubbles. Standard deviation analysis distinguishes normal liquid level changes from short-term fluctuations caused by bubbles, ensuring system stability even in complex infusion environments. Enhanced bubble immunity prevents false level detections caused by bubble adhesion. This improves the accuracy of remaining liquid volume calculations, reduces false alarms, and enhances infusion monitoring reliability.

[0085] By comprehensively calculating the rate of liquid level drop and capacitance fluctuation, an intelligent adjustment judgment factor is implemented. This improves sensitivity in low-noise environments and enhances anti-interference capabilities in noisy environments, allowing the system to adapt to different infusion environments. This ensures a controlled and stable adjustment range, preventing measurement errors caused by over-adjustment. Combined with intelligent adjustment of the judgment factor, this method achieves more accurate liquid level detection, making ICU infusion monitoring more reliable. By optimizing liquid level calculation accuracy, false alarms are reduced, and the accuracy of nurse wristband alarms is improved, reducing the burden on medical staff.

[0086] For example, ICU patients receiving continuous dopamine infusion

[0087] Sensor array: 10 capacitive sensors are attached vertically along the infusion bag (with a spacing of 2 cm);

[0088] Real-time detection:

[0089] The initial liquid volume is 200 mL and the liquid level is 20 cm.

[0090] The infusion rate was 2 mL / min, and the expected remaining time was 100 minutes.

[0091] Warning trigger:

[0092] When the liquid level drops to 4cm (40mL remaining), the nurse's wristband vibrates to alert you.

[0093] When the liquid level approaches 0 cm, the system links the infusion pump to stop running to prevent air embolism.

[0094] The current liquid level drop rate Multiply by the dripping time t to get the total height change of the liquid level drop , the cross-sectional area of ​​the preset infusion container is , the maximum liquid level height at the start of infusion Total height change minus the drop in liquid level The difference is multiplied by the cross-sectional area of ​​the infusion container to obtain the remaining liquid volume.

[0095] Methods for obtaining accurate residual fluid volume include:

[0096] During the preset time period of g, the remaining liquid volume data is collected and smoothed by sliding window filtering to eliminate shaking interference; the total height change threshold of the liquid level drop is preset, and when the total height change of the liquid level drop When the total height change of the liquid level drop is greater than or equal to the preset threshold, it is determined that the total height change of the liquid level drop is disturbed by shaking, and the remaining liquid volume data is abnormal. The abnormal remaining liquid volume data will be eliminated, and the remaining liquid volume will be continuously updated to obtain accurate remaining liquid volume data.

[0097] Methods for obtaining patient vital sign data include:

[0098] A protocol conversion adapter is deployed at the edge node of the infusion stand, which is connected to the ICU central monitoring terminal through the ICU intranet. The real-time data stream of the ICU central monitoring terminal is parsed into structured data, and the open API of the ICU central monitoring system is called to authenticate the patient's vital signs data. Patient vital signs data include heart rate, blood pressure, blood oxygen saturation, respiratory rate, body temperature, hemodynamic data, brain function and metabolic indicators. Hemodynamic data include central venous pressure, pulmonary artery wedge pressure and the amount of blood pumped by the heart per minute. Brain function includes intracranial pressure and bispectral index. Metabolic indicators include blood glucose, lactate and electrolytes.

[0099] Methods for obtaining the deviation index include:

[0100] Build and train a deviation index prediction model using the deep learning library TensorFlow. The deviation index prediction model consists of an input layer, an LSTM layer, and an output layer. The input layer is used to input historical accurate remaining fluid volume data, and the output layer is used to output the remaining infusion time. The deviation index prediction model is an LSTM model.

[0101] Define the model's loss function and use the L2 regularized mean square error loss function to measure the difference between the predicted remaining infusion time and the actual remaining infusion time; use the trained precise remaining liquid volume data to predict the current precise remaining liquid volume data to obtain the remaining infusion time.

[0102] Methods for dynamically sorting the bed list include:

[0103] The types of infused drugs include antibiotics, nutritional support drugs, chemotherapy drugs, analgesics, hormone drugs, antivirals, anticoagulants, immunoglobulins, antiallergic drugs, and detoxification drugs;

[0104] Obtain the infusion drug type and dynamically sort the bed list on the nurse wristband interface based on the remaining infusion time; sort the bed list on the nurse wristband interface based on the patient's remaining infusion time and infusion drug type, and define sorting rules. The sorting rules include priority for the shortest remaining time, drug priority, bed distance, and patient health status.

[0105] Prioritize patients with the shortest remaining time: Prioritize patients with the shortest remaining time, helping nurses prioritize infusions that are nearing completion. Medication priority: Prioritize medications based on their type. For example, chemotherapy drugs, emergency medications, antibiotics, and other medications may require a higher priority. Prioritize medications based on expert experience.

[0106] Bed distance: In actual operation, the nurse wristband takes the distance to the bed into consideration and prioritizes the bed closest to the nurse, facilitating timely treatment;

[0107] Patient health status: Patients are given higher priority based on their clinical status (patient vital sign data), such as those with unstable vital signs, to ensure timely intervention.

[0108] Methods for dynamic transmission adjustment include:

[0109] Real-time perception of the current network status, obtaining bandwidth utilization, link round-trip delay, and packet loss rate, and classifying network status into levels of normal, congested, and faulty. Based on the network status level, different adjustment strategies are triggered to dynamically adjust the transmission of vibration alarm information and the sorted bed list.

[0110] surface Network status classification

[0111] Network status level Bandwidth utilization Link round-trip delay Packet loss rate normal <60% <50ms <0.1% congestion 60%-80% 50ms-100ms 0.1%-1% Fault >80% >100ms >1%

[0112] If the network status level is normal, the data is transmitted through the 5G slice primary path, the LoRaWAN backup path is dormant, and the vibration alarm information is transmitted to the ICU central monitoring terminal. The sorted bed list is updated every 5 seconds through the ZigBee mesh network;

[0113] If the network status level is congested, the sorted bed list is first compressed using Delta coding (transmitting only the changed parts) and then updated via the ZigBee mesh network. The "urgent frame" preemption mechanism of the time-sensitive network is triggered, interrupting the transmission of low-priority data, and automatically activating the LoRaWAN backup path to transmit the vibration alarm information to the ICU central monitoring terminal.

[0114] If the network status level is faulty, the 5G slice is shut down and all data is transmitted via a hybrid ZigBee+LoRaWAN network. The vibration alarm is split into N LoRa data packets (50 bytes each), using Fountain code with zero rate encoding to ensure that the information can be reassembled once any 50% of the packets arrive. If the local gateway is paralyzed, the data is relayed via the ZigBee mesh network of the adjacent ward.

[0115] The circular LED array displays the bed location in polar coordinates (e.g., 12 o'clock represents bed 1), and the color of the LED strip distinguishes the urgency of the vibration alarm information.

[0116] surface Urgency classification

[0117] Light strip color Urgency Vibration Mode red High risk (DI>3) 3 consecutive short vibrations yellow Medium risk (1 <DI≤3) 2-second interval vibration green Presentation (liquid volume <10%) Single slight vibration

[0118] For example: Scenario: ICU multiple devices concurrently alarm

[0119] Event: Five infusion pumps triggered red alarms simultaneously (DI>3); the network entered a congested state (bandwidth utilization 75%, RTT=85ms);

[0120] Dynamic adjustment process: Preemptive transmission: The 5G URLLC slice interrupts the bed list data being transmitted and prioritizes sending alarm information; Path diversion: Three alarms are transmitted via the 5G main path, and two are switched to LoRaWAN; LED linkage: The LED array of the central monitoring terminal highlights five directions, and the nurse's wristband receives graded vibration prompts;

[0121] Results: All alarms reached the nurse's wristband within 120ms; one alarm that was not responded to after 2 minutes was automatically forwarded to the adjacent ward for support and processing by the nurses in the adjacent ward;

[0122] Advantages summary: Intelligent adaptability: Dynamically adjusts strategies based on network status and data priority, balancing efficiency and reliability; Collaborative early warning: The cross-ward alarm forwarding mechanism breaks down information silos and improves emergency response capabilities; Human-computer interaction optimization: LED direction indication + vibration grading significantly reduces nurses' cognitive load.

[0123] In this embodiment, by dynamically reducing the judgment factor, when the system detects a rapid drop in the liquid level, the judgment factor is automatically reduced, the sensitivity is improved, and the system can more easily detect the actual drop in the liquid level. It solves the measurement lag problem caused by the residual effect of the liquid film and improves the detection accuracy of high-viscosity liquid medicines; ensures that the system can respond in time when the liquid level drops, avoiding the misjudgment that the infusion is still continuing; it is particularly suitable for viscous liquid medicines such as fat emulsions and albumin, and improves the versatility of liquid level detection. The judgment factor is dynamically increased. When a large capacitance fluctuation is detected, the judgment factor is automatically increased, making the system insensitive to capacitance fluctuations in a short period of time, thereby reducing misjudgments caused by bubbles; using standard deviation analysis, it can distinguish between normal liquid level changes and short-term fluctuations caused by bubbles, so that the system can remain stable in complex infusion environments. Enhance the ability to resist bubble interference and avoid false liquid level detection caused by bubble attachment. Improve the accuracy of the remaining liquid volume calculation, reduce false alarms, and improve the reliability of infusion monitoring.

[0124] By comprehensively calculating the rate of liquid level drop and capacitance fluctuation, an intelligent adjustment judgment factor is implemented. This improves sensitivity in low-noise environments and enhances anti-interference capabilities in noisy environments, allowing the system to adapt to different infusion environments. This ensures a controlled and stable adjustment range, preventing measurement errors caused by over-adjustment. Combined with intelligent adjustment of the judgment factor, this method achieves more accurate liquid level detection, making ICU infusion monitoring more reliable. By optimizing liquid level calculation accuracy, false alarms are reduced, and the accuracy of nurse wristband alarms is improved, reducing the burden on medical staff.

[0125] Example 2

[0126] See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of Example 1. A mobile infusion safety monitoring and remote early warning method based on edge computing is provided, including:

[0127] S1. Collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the built-in ADC module of the infusion stand edge node. It is distributed vertically along the infusion container, dynamically monitors the remaining liquid volume, and uses a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It is also synchronously connected to the ICU central monitoring terminal to obtain patient vital sign data.

[0128] S2. Predict the remaining time of infusion based on accurate remaining fluid volume data; obtain the infusion drug type and, combined with the remaining infusion time, dynamically sort the bed list on the nurse wristband interface; enable zero call mode for general wards, cancel all local sound and light alarms, and retain only the vibration alarm information from the nurse wristband;

[0129] S3: Use slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and define a three-level QoS strategy for dynamic transmission adjustment;

[0130] S4. The bed position is displayed through a circular LED array, and the urgency of the vibration alarm information is distinguished by the color of the LED light strip. If the nurse in this ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

[0131] Since the electronic device introduced in this embodiment is the electronic device used to implement the mobile infusion safety monitoring and remote warning system and method based on edge computing in the embodiment of this application, based on the mobile infusion safety monitoring and remote warning system and method based on edge computing introduced in the embodiment of this application, technical personnel in this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be introduced in detail here. As long as technical personnel in this field implement the electronic device used by the mobile infusion safety monitoring and remote warning system and method based on edge computing in the embodiment of this application, they all fall within the scope of protection of this application.

[0132] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0133] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. The mobile infusion safety monitoring and remote warning system based on edge computing is characterized by: include: The data stream monitoring unit collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the ADC module built into the edge node of the infusion stand. It is distributed vertically along the infusion container to dynamically monitor the remaining liquid volume, using a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It also simultaneously connects to the ICU central monitoring terminal to obtain patient vital sign data. The edge intelligent analysis unit predicts the remaining infusion time based on accurate remaining fluid volume data; obtains the infusion drug type and, combined with the remaining infusion time, dynamically sorts the bed list on the nurse wristband interface; and enables zero-call mode for general wards, canceling all local sound and light alarms and retaining only the vibration alarm information from the nurse wristband; The low-latency communication unit uses slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and defines a three-level QoS policy for dynamic transmission adjustment; The alarm interaction unit displays the bed location through a circular LED array and distinguishes the urgency of the vibration alarm information through the color of the LED light strip. If the nurse in the ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

2. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 1 is characterized in that: The method for collecting infusion drip rate data of patients in the ICU ward includes: Using electromagnetic interference-resistant infrared array sensors, adapted to the ICE electromagnetic environment, to collect infusion drip rate data for patients in the ICU ward, the electromagnetic interference-resistant infrared array sensors are installed around the drip bucket of the infusion tube. The infrared array sensors use an n×n infrared transmitter-receiver array layout. The array layout rules are defined, including the transmitter, receiver, array arrangement, and vertical spacing. The transmitter and receiver are located on either side of the drip bucket of the infusion tube, forming a penetrating optical path. The transmitter uses a narrow-band infrared LED for transmission, which is focused by a collimating lens to form a parallel beam. The receiving end uses a PIN photodiode with a built-in preamplifier and spectral response filter. The array is arranged as p×p detection points, forming a grid-like three-dimensional detection network covering the entire droplet passing area. The center of the grid is aligned with the droplet outlet. The vertical spacing is defined as the distance between adjacent transmitting and receiving pairs, and is designed to be m millimeters. Infusion drip rate data includes basic drip rate data and infusion environment data; basic drip rate data includes instantaneous drip rate, average drip rate, drip rate fluctuation and drip rate change rate; infusion environment data includes ward temperature, drip time and infusion tube pressure.

3. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 2 is characterized in that: The method for converting into a dripping speed digital signal comprises: The receiving end of the deployed electromagnetic interference-resistant infrared array sensor detects the light intensity signal reaching the receiving end after being blocked or scattered by droplets when passing through the infusion tube funnel. The light intensity signals detected at different detection points are weighted averaged to extract the optimal dripping rate signal. An anti-interference filtering algorithm is added to the front end of the built-in ADC module in the edge node of the infusion stand to optimize the signal's anti-interference capability. The ADC module converts the optimal drip rate signal after anti-interference at a preset sampling rate greater than vHz, and collects timestamp information to convert the optimal drip rate signal (analog voltage signal) into a digital drip rate signal.

4. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 3 is characterized in that: The method for dynamically monitoring the residual liquid volume comprises: Arrange K capacitive liquid level sensors vertically along the infusion container to form a liquid level detection array, which is attached to the outer wall of the infusion container. The liquid level detection array covers the entire liquid level range from the bottom to the top of the container; The initial capacitance value of each capacitive liquid level sensor measured in the absence of liquid is recorded as , the saturated capacitance value of each capacitive liquid level sensor measured when the infusion container is full is recorded as , construct the capacitance-liquid level relationship model of each capacitive liquid level sensor: ;in, Indicates the capacitance value at the current liquid level; Indicates the current liquid level; Indicates the total height of the infusion container; Each capacitive sensor measures the capacitance value of the liquid in real time, and the real-time capacitance value measured by each capacitive liquid level sensor is recorded as ; Determine whether there is liquid at any liquid level height. If , it is determined that there is liquid at this liquid level height; Indicates the judgment factor; calculates the current liquid level height through the highest triggered capacitance sensor position; The judgment factor is dynamically adjusted through the judgment factor adjustment formula. The judgment factor adjustment formula is: ;in, Represents the judgment factor after dynamic adjustment; Indicates the current liquid level drop rate; Indicates the maximum rate of liquid level drop; Indicates the standard deviation of the current capacitance value; Indicates the maximum standard deviation of capacitance value; represents the rate weight coefficient; represents the capacitance fluctuation weight coefficient; The current liquid level drop rate Multiply by the dripping time t to get the total height change of the liquid level drop , the cross-sectional area of ​​the preset infusion container is , the maximum liquid level height at the start of infusion Total height change minus the drop in liquid level The difference is multiplied by the cross-sectional area of ​​the infusion container to obtain the remaining liquid volume.

5. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 4 is characterized in that: The method for obtaining the precise amount of remaining liquid includes: During the preset time period g, the remaining liquid volume data is collected and smoothed by sliding window filtering to eliminate shaking interference; The total height change threshold of the liquid level drop is preset. When the total height change of the liquid level drop is greater than or equal to the preset threshold, it is determined that the total height change of the liquid level drop is disturbed by shaking, and the remaining liquid volume data is abnormal. The abnormal remaining liquid volume data will be eliminated, and the remaining liquid volume will be continuously updated to obtain accurate remaining liquid volume data.

6. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 5 is characterized in that: The method for obtaining patient vital sign data comprises: A protocol conversion adapter is deployed at the edge node of the infusion stand, which is connected to the ICU central monitoring terminal through the ICU intranet. The real-time data stream of the ICU central monitoring terminal is parsed into structured data, and the open API of the ICU central monitoring system is called to authenticate the patient's vital signs data; patient vital signs data include heart rate, blood pressure, blood oxygen saturation, respiratory rate, body temperature, hemodynamic data, brain function and metabolic indicators.

7. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 6 is characterized in that: The method for obtaining the deviation index includes: Build and train a deviation index prediction model using the deep learning library TensorFlow. The deviation index prediction model consists of an input layer, an LSTM layer, and an output layer. The input layer is used to input historical accurate remaining fluid volume data, and the output layer is used to output the remaining infusion time. The deviation index prediction model is an LSTM model. Define the model's loss function and use the L2 regularized mean square error loss function to measure the difference between the predicted remaining infusion time and the actual remaining infusion time; use the trained precise remaining liquid volume data to predict the current precise remaining liquid volume data to obtain the remaining infusion time.

8. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 7 is characterized in that: The method for dynamically sorting the bed list includes: The types of infused drugs include antibiotics, nutritional support drugs, chemotherapy drugs, analgesics, hormone drugs, antivirals, anticoagulants, immunoglobulins, antiallergic drugs, and detoxification drugs; Obtain the infusion drug type and dynamically sort the bed list on the nurse wristband interface based on the remaining infusion time; sort the bed list on the nurse wristband interface based on the patient's remaining infusion time and infusion drug type, and define sorting rules. The sorting rules include priority for the shortest remaining time, drug priority, bed distance, and patient health status.

9. The mobile infusion safety monitoring and remote warning system based on edge computing according to claim 8 is characterized in that: The method for performing dynamic transmission adjustment includes: Real-time perception of the current network status, obtaining bandwidth utilization, link round-trip delay, and packet loss rate, and classifying network status into levels of normal, congested, and faulty. Based on the network status level, different adjustment strategies are triggered to dynamically adjust the transmission of vibration alarm information and the sorted bed list. If the network status level is normal, the data is transmitted through the 5G slice primary path, the LoRaWAN backup path is dormant, and the vibration alarm information is transmitted to the ICU central monitoring terminal. The sorted bed list is updated every 5 seconds through the ZigBee mesh network; If the network status level is congested, the sorted bed list is first compressed using Delta coding and then updated through the ZigBee mesh network. The preemptive mechanism of the time-sensitive network is triggered, interrupting the transmission of low-priority data, and automatically activating the LoRaWAN backup path to transmit the vibration alarm information to the ICU central monitoring terminal. If the network status level is failure, the 5G slice will be shut down, and all data will be transmitted through a mixed ZigBee and LoRaWAN network. The vibration alarm will be split into N LoRa data packets, and Fountain code will be used for rate-free encoding to ensure that the information can be reassembled when any half of the packet arrives. If the local gateway is paralyzed, the data will be relayed through the ZigBee mesh network of the adjacent ward.

10. A mobile infusion safety monitoring and remote early warning method based on edge computing, used to implement the mobile infusion safety monitoring and remote early warning system based on edge computing according to any one of claims 1 to 9, characterized in that: include: S1. Collects infusion drip rate data from ICU patients and converts it into a digital drip rate signal through the built-in ADC module of the infusion stand edge node. It is distributed vertically along the infusion container, dynamically monitors the remaining liquid volume, and uses a compensation algorithm to eliminate sloshing interference to obtain accurate remaining liquid volume data. It is also synchronously connected to the ICU central monitoring terminal to obtain patient vital sign data. S2. Predict the remaining time of infusion based on accurate remaining fluid volume data; obtain the infusion drug type and, combined with the remaining infusion time, dynamically sort the bed list on the nurse wristband interface; enable zero call mode for general wards, cancel all local sound and light alarms, and retain only the vibration alarm information from the nurse wristband; S3: Use slicing technology to transmit vibration alarm information and sorted bed lists to the ICU central monitoring terminal, and define a three-level QoS strategy for dynamic transmission adjustment; S4. The bed position is displayed through a circular LED array, and the urgency of the vibration alarm information is distinguished by the color of the LED light strip. If the nurse in this ward does not respond to the vibration alarm information for more than T minutes, the vibration alarm information will be automatically forwarded to the wristband of the adjacent ward through the ICU central monitoring terminal.

Images