Intelligent infusion process monitoring system based on precise infusion monitoring data

Abstract

The application discloses a kind of infusion process intelligent monitoring systems based on accurate infusion monitoring data, belong to biomedical engineering technical field, including infusion initialization and multi-source data acquisition module, infusion quality score and continuous curve analysis module, abnormal preliminary judgment and confidence calculation module, active perturbation control module and early warning and decision module.The application reaches the technical effect of accurately identifying early abnormal risk in infusion process, accurately distinguishing different types of abnormalities and quickly disposing abnormal events by constructing the infusion quality score of continuous curve, actively applying small disturbance and introducing confidence dynamic management and hierarchical disposal mechanism, solve the problems of single alarm mode, lack of confidence management, difficult to balance safety and nursing efficiency, frequent false alarm, missed judgment or unable to determine the root cause of abnormality in the prior art.

Description

Technical Field

[0001] This invention relates to the field of biomedical engineering technology, and in particular to an intelligent monitoring system for the infusion process based on precise infusion monitoring data. Background Technology

[0002] Intravenous infusion is one of the most common and fundamental treatment methods in current clinical practice, with a large number of patients relying on it to complete their treatment. Traditionally, the infusion process relies primarily on manual monitoring by medical staff, who observe the drip rate, fluid level, and patient complaints to determine if the infusion is normal, whether the IV fluid needs to be changed, or if the needle needs to be removed. In addition, current technology has developed various infusion monitoring devices to monitor the infusion process, including those using photoelectric sensors, weight sensors, or fluid level detectors at the drip chamber to automatically monitor the drip rate and remaining medication volume, and to promptly alert medical staff to intervene when abnormalities occur.

[0003] For example, the intelligent infusion detection and control system disclosed in Chinese invention patent CN109692381B includes: an infusion detection subsystem, a server, and a nurse terminal. The infusion detection subsystem includes a drug level detection module, a control module, a wireless transmission module, and a positioning module. The drug level detection module is used to detect the height of the remaining drug in the infusion bottle or bag. The control module is used to control the wireless transmission module to send the drug level to the server. The control module is also used to control the positioning module to send positioning information when the detected drug level is less than a preset value. The server includes a route planning module, which is used to plan the nurse service route based on the positioning information and the drug level information.

[0004] The above-mentioned technology has at least the following technical problems: Most existing intelligent infusion monitoring and control systems still trigger warnings based on whether the instantaneous drip rate and drug level exceed thresholds. They lack comprehensive analysis of continuous time series data such as drip rate, level, and weight, making it difficult to identify trend anomalies. For phenomena like slowing drip rates or stagnant fluid levels, they can only passively provide vague alerts, unable to actively intervene in the infusion circuit and observe differences before and after intervention to differentiate warnings. They lack mechanisms to verify the nature of anomalies by applying minor perturbations to the infusion circuit and observing differences in response before and after the perturbation, failing to effectively distinguish between mechanical blockages and occasional sensor errors. Furthermore, most existing technologies employ a single alarm mode, alarming and notifying medical staff for different risk levels, making it difficult to balance safety and nursing workload. Simultaneously, existing intelligent infusion systems generally only display the infusion status to medical staff through data signals and a terminal interface, lacking rapid verification methods based on specific visuals under remote conditions. They cannot supplement high-risk alarms with visual evidence, making it difficult to support emergency decision-making. Summary of the Invention

[0005] To address the technical problems of existing technologies, such as the inability to perform intelligent adjustment and early warning, the lack of comprehensive analysis and differentiation of anomaly types leading to frequent false alarms and the inability to identify the root cause of anomalies, this invention provides an intelligent monitoring system for the infusion process based on accurate infusion monitoring data. The technical solution is as follows: An intelligent monitoring system for the infusion process based on precise infusion monitoring data is provided. The system includes: an infusion initialization and multi-source data acquisition module, an infusion quality scoring and continuous curve analysis module, an anomaly preliminary judgment and confidence calculation module, an active perturbation control module, and an early warning and decision-making module.

[0006] The infusion initialization and multi-source data acquisition module is used to establish the correlation information of the infusion process and collect accurate infusion monitoring data corresponding to the infusion process; The infusion quality scoring and continuous curve analysis module is used to process precise infusion monitoring data, generate at least one type of continuous time-series curve, and output a quality score. The anomaly initial judgment and confidence calculation module is used to determine candidate anomaly segments based on continuous time series curves and quality scores, and output anomaly type identifiers and corresponding initial confidence levels. The active perturbation control module is used to output a control sequence for the infusion regulation actuator when the preset trigger conditions are met, and to acquire the corresponding accurate infusion monitoring data before and after the control sequence is applied. The early warning and decision-making module is used to extract response fingerprints and update confidence levels based on accurate infusion monitoring data before and after the action of the control sequence, and output anomaly judgment results as well as risk classification information and treatment strategy information associated with the anomaly judgment results.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. The intelligent monitoring system for infusion process based on precise infusion monitoring data provided by this invention achieves real-time monitoring of the infusion status by simultaneously collecting multi-source data such as drug drip rate, weight, and infrared dripping during the infusion process, and constructs an infusion quality score with a continuous curve. This enables early and accurate identification of trend abnormalities such as slow drip rate decline and stagnant liquid level, as well as potential abnormal risks during the infusion process, effectively avoiding the problem of insufficient alarm accuracy caused by relying solely on instantaneous threshold monitoring in existing technologies.

[0008] 2. This invention actively applies minute disturbances within a safe range when an anomaly is detected by driving an electric tube clamp and collecting response fingerprints such as drug droplet and weight changes before and after the disturbance in real time. This enables active verification and identification of anomalies, and thus achieves accurate differentiation of different anomaly types such as mechanical blockage, slow leakage at the side, end-of-infusion state, and occasional sensor fluctuations. This effectively solves the problems in the prior art where it is difficult to determine the nature of anomalies based on static signals, leading to frequent false alarms, missed detections, and inability to identify the root cause of anomalies.

[0009] 3. By introducing a dynamic confidence management and tiered handling mechanism based on quality scoring and response fingerprints, and automatically pushing information to medical staff in necessary high-confidence, high-risk abnormal situations, combined with remote video verification, differentiated intervention strategies such as prompting for review, suggesting speed adjustment, automatic speed reduction, or pausing are executed according to different confidence levels. This enables rapid and accurate handling of abnormal events, thereby improving the safety and nursing efficiency during infusion. It effectively solves the problems of existing technologies, such as single alarm methods, lack of confidence management and remote visual verification, and difficulty in balancing safety and nursing efficiency. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the structure of an intelligent monitoring system for the infusion process based on precise infusion monitoring data provided in an embodiment of this application; Figure 2 This is a data flow diagram of an intelligent monitoring system for the infusion process based on precise infusion monitoring data provided in an embodiment of this application. Figure 3 A flowchart illustrating the business method of the intelligent monitoring system for infusion process based on precise infusion monitoring data provided in this application embodiment; Figure 4 This is a schematic diagram of the event package model iteration of the early warning and decision-making module provided in the embodiments of this application.

[0012] Figure 5-(a) is a schematic diagram of the drop rate changes before, during and after perturbation under normal conditions provided in the embodiments of this application.

[0013] Figure 5-(b) is a schematic diagram of the drop rate changes before, during and after perturbation in the case of suspected blockage provided in the embodiments of this application.

[0014] Figure 5-(c) is a schematic diagram of the drop rate changes before, during and after perturbation in the case of suspected leakage provided in the embodiments of this application.

[0015] Figure 5-(d) is a schematic diagram of the droplet rate changes before, during and after perturbation under abnormal flow conditions provided in the embodiments of this application. Detailed Implementation

[0016] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0017] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] like Figure 1 The diagram shown is a structural schematic of the intelligent monitoring system for the infusion process based on precise infusion monitoring data provided in Embodiment 1 of the present invention. Figure 3 The diagram shown is a flowchart of the business method of the intelligent monitoring system for the infusion process based on precise infusion monitoring data provided in this application embodiment. The system structure and the business methods of each system are as follows: The infusion initialization and multi-source data acquisition module is used to complete basic hardware self-tests and configuration loading after the platform's electronic equipment is powered on, binding patient identity information with the corresponding infusion solution and related medical orders. It performs initialization and self-calibration on infusion-related sensors, including a drug weight sensor and an infrared drip sensor, and performs zero-opening self-calibration on the electric tube clamp. The initialization and self-calibration processes for the sensors and tube clamp need to be performed by technicians according to the actual situation and specific requirements; this embodiment of the invention does not provide specific explanations or constraints on this process. When the device, sensors, and tube clamp all pass the self-test, the device status is set to infusion ready, indicating that infusion operations can now be performed on patients.

[0020] In currently publicly available and widely used scenarios, infrared drip sensors typically rely on a clip that attaches to the outside of the drip chamber in the infusion pathway. The clip contains an infrared emitting tube and a receiving tube, forming a through-beam optical path that passes through the dripping area to monitor changes in drip rate. Medication weight sensors are usually positioned between the infusion stand and the infusion bag, calculating the remaining amount by sensing weight changes. In practical applications, different weight sensors need to be selected based on the type of infusion container. For example, in this embodiment of the invention using an infusion bottle as an example, a tray-type weight sensor is selected, where the infusion container is placed on a tray for weighing; if the infusion container is an infusion bag, a hook-type weight sensor is selected.

[0021] During the infusion process, the patient's condition is monitored at each time point t on the edge processor side. k Collect and record precise infusion monitoring data corresponding to the infusion process: Infrared drip sensor records the time t of each drug drop passing through the drip chamber. drop (k), and denot the number of drops falling within this time period as N. drop (k); The drug solution weight sensor records the total weight including the infusion bottle, drug solution, and hook, denoted as W(k); The system records the current opening degree C of the electric tube clamp. open (k).

[0022] The collected data undergoes unified preprocessing. If at a certain time point t... k Values ​​recorded exceeding the sensor's range, or differences from preceding and following time periods exceeding physically acceptable limits, are marked as outliers. These outliers are removed and replaced with nearest-neighbor interpolation. The values ​​W(k) recorded by the liquid weight sensor are smoothed using a sliding window averaging method. Median filtering is applied to the pulse signals generated when the infrared drip sensor detects liquid dripping to remove jitter, facilitating accurate identification of the dripping pulses. For each time point t... k Calculate the average drip rate v(k) during this period: The unit of drip rate is drops per minute, and ∆t represents the time point t between adjacent time points. k and t k-1 The difference. The calculated v(k) is assigned to the sliding window T. w The midpoint of the sliding window is used to obtain a drip rate variation curve v(t) that changes over time. In this embodiment, the sliding window value is 30s. It should be noted that the sliding window value is not fixed and can be adjusted by those skilled in the art according to actual conditions. This embodiment of the invention does not impose any constraints on this.

[0023] The infusion quality scoring and continuous curve analysis module is used to process precise infusion monitoring data, generate continuous time-series curves, and output quality scores. For the collected drug weight data W(k), within a sliding window T... wEach time point t within k Calculate the net weight of the medicine solution: Where W0 is the unloaded mass of the infusion bottle and its holder obtained through preset calibration; its specific value can be obtained from the corresponding manufacturer. The calculated W'(k) is assigned to the sliding window T. w The midpoint of the curve is used to obtain a curve W(t) showing the change in residual drug weight over time. To facilitate subsequent consistency verification, linear fitting is used within a sliding window to calculate the curve W(t) at each time point t. k slope S W(k) : Where W(k) represents W(t) at time t. k The weight, W(k-1) represents W(t) at time point t. k-1 The weight. Theoretically, when the drug solution density and drip rate are constant, the slope is linearly related to the drip rate change curve v(t). In practical applications, a small hysteresis τ is generally allowed in the system, i.e., v(t-τ) is aligned with S. W(t) To reduce errors, if the type of medication changes during the infusion process due to special reasons (such as changing the medication midway or the density of the medication in the later stages being different), the system no longer assumes that the ratio between the change in medication weight and the drip rate is constant. Instead, it monitors the corresponding ratio change between the drip rate change trend and the weight change trend for switching detection. If this corresponding ratio stably jumps to another level within several consecutive sliding windows, the system determines that the current infusion process has entered a new medication stage. After the determination is made, the system automatically updates the parameter baseline corresponding to this stage (such as the weight-drip rate correspondence coefficient, quality score reference range, etc.) and re-records the baseline opening, baseline drip rate, and baseline weight change rate for this stage, thereby ensuring that subsequent consistency checks are still based on the consistency of the "same medication stage".

[0024] Next, the infusion quality score is calculated. In each sliding window T... w Within this framework, the drip rate stability index, multi-sensor consistency index, and signal integrity index are calculated respectively. The specific calculation methods are as follows.

[0025] The drip rate stability index is expressed as the rate of dripping within a sliding window T. w The stability of the internal drip rate. The variance σ of the drip rate change curve v(t) calculated from the infusion initialization and multi-source data acquisition module is statistically analyzed. v 2 And calculate the drip rate stability index F1(k): Where, σ max2 This represents the maximum expected variance. F1(k) ranges from [0,1], and the closer the value is to 1, the more stable the drip rate.

[0026] The multi-sensor consistency index represents the degree of difference between the actual change data detected by the drug solution weight sensor and the theoretical change data calculated based on the infrared drip sensor. The sliding window T is estimated based on the drip rate change curve v(t) calculated by the infusion initialization and multi-source data acquisition module. w Theoretical weight consumption ∆W est : Calculate the measured weight change ∆W from the above weight curve: Where W(k) represents the value at time t. k The measured weight, W(k-1), represents the weight at time point t. k-1 The actual measured weight is used. Calculate ∆W. est The relative deviation Err between ∆W and ∆W: Where ε represents the preset maximum weight change. Calculate the multi-sensor consistency index F2(k): Among them, Err max This represents the preset maximum relative deviation. The value of F2(k) ranges from [0,1]. The closer the value is to 1, the closer the theoretical data is to the actual data.

[0027] Signal integrity metric, representing the current sliding window T w The effective proportion of each data point in the internal statistics. After removing invalid data such as missing sampling points and sampling points judged as outliers, the proportion of remaining valid data to all sampled data is denoted as P. 有效 The signal integrity index F3(k) is obtained as follows: The value of F3(k) ranges from [0,1]. The closer the value is to 1, the better the signal integrity and reliability.

[0028] The infusion quality score Q is obtained by weighted summing of the above-mentioned drip rate stability index, multi-sensor consistency index, and signal integrity index. Where w1 represents the weight of the drip rate stability index F1(k), w2 represents the weight of the multi-sensor consistency index F2(k), and w3 represents the weight of the signal integrity index F3(k). The specific values ​​of each weight need to be determined and adjusted by technical personnel according to the actual situation, and generally, w1+w2+w3=1 is sufficient.

[0029] The initial anomaly detection and confidence level calculation module is used to detect and identify candidate segments of abnormal drug solution changes by analyzing curve features. Within the first few sliding windows after the infusion begins, when no obvious anomalies are detected by the system, the slope of the reference drip rate and reference weight change is calculated. Where v represents the reference drip rate, S w This represents the slope of the reference weight change, and mean represents the average value calculation. For each sliding window T... w Calculate the slope deviation of weight change D W(k) (i.e., rate of change characteristic), drip rate deviation D v(k) (i.e., stability characteristics) and the stopping drip ratio P(k) (i.e., persistence characteristics), the specific calculation formulas are as follows: Where ε is a very small real number to prevent the denominator from becoming 0 and rendering the expression meaningless, N represents the number of sub-segments into which the sliding window is further subdivided, and n nondrop ≥T blockmin This indicates that within the subdivided segments, the time during which no dripping occurs for a continuous period is greater than or equal to a preset minimum time threshold T that can be considered a blockage. blockmin The number of sub-segments. A manually set threshold θ for the slope deviation of weight change. W Drop rate deviation threshold θ v and the lower limit of quality score θ Q It should be noted that the specific value of the threshold needs to be manually set by technicians during operation based on the actual situation; this embodiment of the invention does not impose any restrictions on this. For ease of understanding by technicians, this embodiment of the invention uses θ. W =0.3, θ v =0.2, θ Q =0.6.

[0030] When a certain sliding window satisfies D W(k) >θ W And Q > θ Q Or D v(k) >θ v And Q > θ QWhen the P(k) value is close to 1, the sliding window is marked as a candidate window for liquid level anomaly. Temporally adjacent candidate windows are merged to form a continuous liquid level anomaly phase.

[0031] For each candidate window k of liquid level anomaly, calculate the anomaly intensity index A(k) and the blockage index B(k), and then combine them in a weighted manner to obtain the initial confidence level C0. The specific calculation formula is as follows: Where, ω W The deviation of the slope of the weight change is represented by D. W(k) The weight, ω v Indicates drip rate deviation D v(k) The weight, ω Q This represents the weight of the infusion quality score Q. The specific values ​​for each weight need to be determined and adjusted by technical personnel based on the actual situation, generally satisfying ω. W +ω v +ω Q =1 is sufficient; for example, in this embodiment, ω is set to 1. W =0.3, ω v =0.3, ω Q =0.4.

[0032] Where, ω stall Indicates the degree of stagnation of liquid level S stall The weight, ω zero This represents the weight of the drip cessation ratio P(k). The specific values ​​of each weight need to be determined and adjusted by technical personnel based on the actual situation, generally satisfying ω. stall +ω zero =1 is sufficient; for example, in this embodiment, ω is set to 1. stall =0.5, ω zero =0.5.

[0033] Therefore, the initial confidence level C0 can be obtained: Wherein, γ represents the weight of the anomaly intensity index A(k), with a value of 0.5. This value is chosen to balance the overall anomaly intensity and the reliability of the congestion event. The closer the value of C0 is to 1, the more likely this anomaly is to trigger the subsequent active perturbation control module. It should be noted that the values ​​in this embodiment are for reference only, and technicians can adjust them according to the actual situation during actual operation. This embodiment of the invention does not impose any constraints on this.

[0034] To detect suspected leaks and improve the accuracy of anomaly monitoring, a drip rate-weight consistency index is introduced. This applies to the sliding window T.w Given that the average drip rate within the window is v(k), let w be the weight of a single drop of medicine. drop Then, based on the drip rate, the theoretical weight reduction W of the medicine solution can be estimated. drop : The actual reduction in weight of the liquid medicine is denoted as W. meas The specific calculation formula is as follows: Among them, W start W represents the weight of the medicine at the point in time when the weight decrease begins to be counted within the sliding window. end The weight of the liquid is represented by W(t), indicating the point in time when the statistical weight decrease ends within the sliding window. Both are read from the denoised and smoothed liquid weight change curve W(t). Under normal ideal conditions, if the sensor is functioning properly and there is no leakage, then W... drop Numerical value and W meas The values ​​should be approximately equal.

[0035] Calculate the tendency to leak (L) seg The specific formula is as follows: Where, ω L The value represents the weight of the leakage deviation, ranging from (0,1). For ease of explanation, this embodiment uses a value of 0.5. ε is a very small real number to prevent the denominator from becoming 0 and rendering the expression meaningless. L seg The larger the value, the more significant the actual weight reduction is compared to the predicted drip rate, and the more likely it is to indicate a possible leak.

[0036] Artificially set the blockage confidence threshold C block Overall anomaly threshold C abn and leakage deviation threshold θ leak For ease of explanation, this embodiment uses C. block =0.6, C abn =0.4, θ leak =0.3. When C0 > C block When the blockage index B(k) is large and v(k) is significantly low, this abnormal stage is marked as a "suspected blockage" state; when C0≥C abn The blockage index B(k) is within the normal range and L seg >θ leak When C occurs, the abnormal stage is marked as a "suspected leak" state; when C abn <C0<C blockFurthermore, when the drip rate v(k) deviates significantly from the reference value, this abnormal stage is marked as an "abnormal flow rate" state. It should be noted that the aforementioned manually set threshold needs to be freely set by technicians based on actual conditions during operation; this embodiment does not impose any restrictions on this.

[0037] The active perturbation control module combines the judgment results from the initial anomaly assessment and confidence calculation module. When the system determines that the anomaly is a suspected leak or suspected blockage, the system controls an existing and technologically mature electric tube clamp to perform perturbation operations. The electric tube clamp generally consists of a motor (stepper motor or small servo motor), a lead screw and rack and pinion mechanism, and a pressure plate. During operation, the motor rotates, driving the lead screw, which moves the pressure plate linearly, thereby changing the degree of compression of the infusion tube and thus altering the effective cross-sectional area and drip rate within the tube. This is widely used in clinical medicine to regulate infusion rate. During the device initialization phase, the drive motor sequentially finds two mechanical limits: fully released and fully clamped, denoted as s. max and s min The current position of the clamp is denoted as s, and the opening percentage O is defined as follows: Based on this, technicians pre-configure the safe opening range according to the type of medication, infusion stage, and doctor's instructions. min O max ] and the corresponding safe drip rate range [v min ,v max Based on the anomaly type determination in the initial anomaly assessment and confidence level calculation module, a perturbation operation of the pipe clamp is implemented. For ease of explanation, this embodiment sets O... min =35%, O max =70%. Furthermore, the safe drip rate range is preset by medical staff or equipment administrators during the system configuration phase. This setting is based on factors including the target drip rate or administration time limit specified in the doctor's order, the drug's instructions, the infusion guidelines established by the hospital department, and individualized constraints such as the patient's weight, age, and comorbidities. In actual use, the target drip rate specified in the doctor's order can be adjusted accordingly. set As a baseline value, and with a given allowable deviation coefficient Δ, different tolerance zones are set according to the type of drug solution. The specific formula is as follows: For example, for a certain medication solution, a department specifies a drip rate of 30 drops / minute, with Δ set at 20%. The corresponding safe drip rate range is [24, 36] drops / minute. It should be noted that the specific values ​​in this embodiment are merely examples, and technicians need to adjust them according to actual conditions during operation. This embodiment does not impose any constraints on this.

[0038] Before performing the perturbation operation, the system records the current reference state, including the reference opening O. base The reference drip rate v base Reference weight change rate ∆W base And the baseline infrared dripping mode. When an anomaly marked as "suspected blockage" occurs, the clamp executes the first state, i.e., slightly open, with a target opening of 0. target Change to: When an anomaly marked as "suspected leakage" occurs, the clamp enters the second state, which is slightly closed, with a target opening of 0. target Change to: Where clip indicates truncation when the value exceeds the safe range, δ open δ represents the amount of change in the preset clamp angle. close The preset reduction in the clamp angle is indicated by a value of 2%-5% in this embodiment of the invention. Technicians can adjust this value according to actual conditions during operation; this embodiment does not impose any constraints on this. In determining a suspected leak, theoretically, when the clamp is closed, the drip rate should decrease and the rate of weight change should also decrease. If the rate of weight change decreases rapidly but the drip rate change is not linearly correlated with it, it is determined to be an external leak in the pipeline.

[0039] The electric pipe clamp operates according to the target opening O. target Perform active perturbation until the clamp opening reaches 0. targetThen, a timer is started and the clamp is held for T seconds, which is 5 seconds in this embodiment. During T seconds, the drip rate change curve v(t) and the remaining drug weight change curve W(t) are continuously collected, and the infrared drip sensor constantly records the pulse signal generated when the drug drips. The drip rate change data, remaining drug weight change data, and pulse signal collected during this time T, together with the same reference data before the perturbation, constitute the raw data of the "response fingerprint". As shown in Figure 5, the drip rate changes of the clamp before, during, and after the perturbation are shown under different conditions. The entire sampling time window is set to 120 seconds. Figure 5-(a) shows the drip rate change of the clamp under normal conditions when the perturbation is implemented. At this time, the clamp is in its first state, and the drip rate before and after the perturbation is around 30 drops / minute and within a safe range. During the perturbation, the drip rate slightly increases to 31 drops / minute, but is still within a safe range. After the perturbation is completed, the drip rate returns to the state before the perturbation. It should be noted that in actual application, the perturbation operation of the clamp is not triggered under normal circumstances. This embodiment is only used as an example for reference. Figure 5-(b) shows the drip rate change when the clamp implements perturbation under suspected blockage. At this time, the clamp is in the first state. Due to the suspected blockage, the drip rate is significantly lower than the normal safe range, only 5-6 drops / minute. When the clamp is in the first state, the blockage is relatively improved but still significantly lower than the normal safe drip rate range, at 7-8 drops / minute. After the perturbation is completed, the drip rate returns to the state before the perturbation. Figure 5-(c) shows the drip rate change when the clamp implements perturbation under suspected leakage. At this time, the clamp is in the second state. Due to the suspected leakage, the drip rate is significantly higher than the normal safe range, at about 45 drops / minute. When the clamp is in the second state, the leakage is relatively improved but still significantly higher than the normal safe range, at about 40 drops / minute. After the perturbation is completed, the drip rate returns to the state before the perturbation. Figure 5-(d) shows the drip rate change when the clamp implements perturbation under abnormal flow conditions. At this time, the clamp is in the first state. Due to abnormal reasons such as the failure of the infrared drip sensor, the current drip rate does not match the reference value. Therefore, the drip rate has many irregular discrete points and most drip rates are not within the safe range. When the clamp implements perturbation, the overall drip rate increases slightly, but it is still not within the safe range. After the perturbation is completed, the drip rate does not fully recover to the state before the perturbation, but shows a more irregular distribution. Therefore, this state needs to be closely monitored.

[0040] After T seconds, the system generates the target opening for rollback. Generally, it rolls back to the baseline opening of 0. base If the risk is deemed too high, i.e., severe leakage or complete blockage, the clamp will fully close and the event will be packaged and handed over to the early warning and decision-making module for processing. After the rollback is completed, the state machine returns from the clamp perturbation mode to the normal speed regulation mode, and the system resumes normal speed regulation.

[0041] Since perturbations of the tube clamp can also cause negative effects such as medication backflow or patient discomfort, a unified interruption condition should be set for safety assurance. During the tube clamp perturbation and response fingerprint acquisition process, the following abnormalities should be monitored in real time: whether the drip rate exceeds the safety limit, whether the weight change gradient is abnormal, whether the infrared drip signal exhibits a sudden abnormal pattern, whether an interruption command is issued from the nurse's end, and whether the motor current exceeds the safety threshold. If any of the above abnormalities are met, the system perturbation action should be stopped immediately, and the tube clamp should quickly retract to the zero position. base .

[0042] The early warning and decision-making module is used to perform response analysis on the data before and after the perturbation of the tube clamp, such as... Figure 4 The diagram shown is an iterative illustration of the event package model of the early warning and decision-making module provided in this embodiment. First, the system divides the infusion process into the pre-perturbation time, the perturbation duration T, and the recovery phase after the clamp retraction. Based on the raw response fingerprint data statistically obtained from the active perturbation control module, the system records the curve change amplitude, trend, and drip recovery rate before and after the clamp perturbation. The difference between the statistical value of the recovery phase and the baseline value before the perturbation is used to construct the curve change amplitude difference. Simultaneously, the window sequences of the same channel in the pre-perturbation window set and the recovery phase window set are trend estimated in chronological order to obtain the pre-perturbation trend quantity and the recovery phase trend quantity. The difference between the two is used to construct the trend difference. Similarly, for the drip rate, after the clamp retraction time, the system searches along the time axis for the time point when the drip rate re-enters the safe range before the perturbation and continuously meets the preset number of windows. The time interval between this time point and the retraction time is recorded as the drip rate recovery time, and the drip rate recovery rate is calculated accordingly. The difference between this recovery rate value and the reference recovery rate is used to construct the drip rate recovery rate difference. The combination of the differential features of these sliding window feature sets before and after the perturbation is the response fingerprint. Based on this response fingerprint, the anomaly type can be identified.

[0043] When the initial anomaly type is "suspected blockage," the system focuses on observing whether the drip rate and weight decrease recover significantly within a short period of time after the clamp is slightly loosened, and whether the liquid level and infrared drip pattern return to normal. If the recovery is significant and relatively stable, the anomaly type is corrected to "clamp too tight or brief compression," and the confidence level for blockage risk is lowered. If the drip rate is still close to stopping and weight change is still severely impaired during the perturbation and recovery phases, it is judged as "actual blockage," and the confidence level for blockage risk is increased accordingly.

[0044] When the initial anomaly type is "suspected leak," the system checks whether the drip rate and weight change decrease synchronously and whether their relationship becomes consistent again after slightly tightening the clamp. If the drip rate and weight change decrease synchronously as expected after tightening, it indicates that the previous anomaly was mostly a short-term fluctuation or a judgment error, and the risk level can be downgraded. If the drip rate decreases significantly or even almost stops after tightening, but the weight continues to decrease at an abnormal rate, or the relationship between drip rate and weight remains severely mismatched, then it is confirmed as a "real leak," and the confidence level of the leak risk is increased. Through the above rules, the system updates the initial confidence level in the anomaly initial judgment and confidence level calculation module to the final confidence level and obtains a clear anomaly type label, such as real blockage, real leak, excessive clamping, false alarm, etc.

[0045] In terms of system architecture, the bedside infusion monitoring terminal acts as an edge node, establishing a two-way secure communication channel with the central AI analysis platform via the hospital's internal 5G private network or Wi-Fi. The terminal preprocesses and compresses data collected from multiple sensors, such as drip rate, liquid level, weight, and infrared drip patterns, locally. For normal processes, it uploads summary data in the form of periodic heartbeat packets; for suspected abnormal processes, it packages and uploads high-resolution curve segments and key feature quantities. The AI ​​analysis platform deploys a rule engine and model services, and provides a backend configuration interface for managing basic parameters by department, bed, and medication type, including drip coefficient, dynamic threshold, risk grading standards, active perturbation amplitude and duration, alarm push strategy, and model version number. The information technology department or quality control center can uniformly maintain and distribute configurations through this interface.

[0046] After completing anomaly confirmation and confidence level update, the edge device uploads the anomaly type, final confidence level, response fingerprint, and key curve segments to the AI ​​analysis platform via 5G / Wi-Fi. The platform combines historical infusion trends, past events involving similar medications, and the current medication regimen to classify the event into risk levels and generate corresponding treatment strategies. Simultaneously, through a messaging channel with the nurse station terminal or mobile nursing terminal, it pushes the risk level, treatment suggestions, and options requiring manual confirmation to nursing staff in the form of pop-ups or task lists. The nursing staff's confirmation results and manual operation records are then sent back to the platform for closed-loop record keeping. If the risk level is normal or low, the platform only issues an automatic micro-adjustment command to the bedside terminal; if the risk level is medium or high, it simultaneously pushes a warning message to the nurse's terminal, issues an automatic speed adjustment or pause command, and marks the current event as "requiring manual confirmation."

[0047] Finally, based on the final strategy provided by the early warning and decision-making module, the bedside terminal automatically adjusts the clamp opening, maintains the current infusion status, or immediately pauses the infusion, and continues to monitor subsequent curves to verify the intervention effect. For all infusion events judged as abnormal, including suspected blockage, suspected leakage, and abnormal flow rate events, the system automatically generates an event package, including the original monitoring curve of this event, preliminary anomaly type and initial confidence level, active perturbation parameters, response fingerprint summary, final risk level, etc., and uploads it to the quality control center for model iteration. After being updated using machine learning-related models, it is redistributed for subsequent optimization of anomaly identification logic and active perturbation parameter configuration. Among them, machine learning-related models include support vector machine models, random forest models, etc. This embodiment does not restrict the selection of specific models, and technicians can freely choose according to the actual situation.

[0048] In Example 2, while remaining largely unchanged from Example 1, a video verification and behavior confirmation module is added to address the issue that simple alerts may lead to insufficient attention from healthcare workers in high-risk situations, resulting in significant risks. This module is used after the early warning and decision-making module determines the risk level based on the updated confidence level C from the edge and information such as historical trends and medication regimens provided by the AI ​​analysis platform. When the risk level is in the medium range and / or the algorithm confidence level is still in the transition range, the video verification and behavior confirmation module is triggered. The system calls the camera terminal bound to the current infusion route and pushes real-time video or recent video recordings, including "abnormality type, risk level," and the corresponding bed, to the nurse's end. The nurse can remotely verify the specific type and risk level of the current abnormality based on the video footage and send the manual confirmation result back to the early warning and decision-making module.

[0049] Based on the video verification results, the early warning and decision-making module modifies the final level and handling strategy for abnormal events: when the blockage is confirmed to be caused by the patient's brief compression of the tube or turning over, the event can be downgraded to a warning alert, and the automatic pause or significant speed adjustment operation can be canceled; when the video is manually verified and no abnormalities are found, but the model determines it to be a high-risk blockage or leakage, the automatic intervention strategy is upgraded. The video verification results are also uploaded as part of the event package to the quality control center and the model iteration module for subsequent optimization of the active perturbation strategy and the anomaly identification model.

[0050] like Figure 2The diagram shows a data flow diagram of the intelligent monitoring system for the infusion process based on precise infusion monitoring data provided in this application embodiment. First, the infusion monitoring terminal integrates components such as an infrared drip rate sensor, a drug weight sensor, and an electric tube clamp, responsible for real-time collection of multi-source monitoring raw data such as drip rate, fluid level / weight, and tube clamp opening at the bedside. After the collected data is timestamped, it is sent to the edge processor at the corresponding bedside via a wired network or 5G network, forming a "raw data" input stream. Next, the edge processor, acting as a local real-time computing node, preprocesses the upstream input raw data, including filtering, noise reduction, time alignment, and missing data completion. Based on this, it calculates continuous curve features and infusion quality scores, completing the initial anomaly judgment and initial confidence level C0 calculation. When the trigger condition is met, the edge processor controls the electric tube clamp to perform active perturbation, statistically analyzes the response fingerprint of the curve before and after the perturbation, and updates the anomaly type and confidence level accordingly. For each anomaly event, the edge processor constructs an event package, which includes at least the original monitoring curve, feature quantity, response fingerprint, and locally calculated anomaly type and confidence level. Subsequently, the edge processor uploads the "abnormal results" and corresponding event packets to the AI ​​analysis platform via the network. The AI ​​analysis platform receives abnormal results and event packets from multiple edge processors, and performs a fusion analysis combining patient basic information, historical infusion records, and past events involving similar medications to complete risk classification and treatment strategy decisions. On one hand, the analysis platform sends warning levels, suggested actions, and whether manual confirmation is required to medical staff; on the other hand, it sends statistical information and verification results of abnormal events to the quality control center and model iteration module for continuous optimization of thresholds and models. Medical staff can view key curve segments and event summaries on the terminal interface and perform remote confirmation or manual intervention for events marked "requiring manual confirmation." The medical staff's confirmation results, actual executed instructions, and related notes are transmitted back to the AI ​​analysis platform via the network as feedback data for subsequent strategy evaluation and model correction. In scenarios requiring visual verification, the AI ​​analysis platform or the quality control center triggers the video verification module to call the camera device of the corresponding bed, collect video segments within the abnormal time window, and send the video segments, associated with the corresponding event packets, to the quality control center and model iteration module. The video verification module primarily provides the function of "retrieving and uploading videos during abnormal pushes," providing real-world evidence for post-event review and model training. The quality control center and model iteration module centrally record and review accumulated event packages, video clips, and medical staff feedback information. Through statistical analysis and model training, they continuously optimize the abnormal judgment threshold, quality scoring model, and treatment strategy parameters. The optimized thresholds and model configurations are distributed to the AI ​​analysis platform in the form of "strategy feedback," and then the AI ​​analysis platform synchronously updates each edge processor, ensuring that the data processing logic and active perturbation parameters of the entire system remain consistent and continuously iteratively updated within the hospital.

[0051] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0052] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0053] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0054] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0055] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the solution, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0056] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent monitoring system for the infusion process based on precise infusion monitoring data, characterized in that, include: The infusion initialization and multi-source data acquisition module is used to establish the association information between the patient and the infusion device before the infusion begins, and to collect accurate infusion monitoring data corresponding to the infusion process during the infusion. The infusion quality scoring and continuous curve analysis module is used to preprocess the precise infusion monitoring data, generate at least one type of continuous time-series curve indexed by time, and output a quality score based on the precise infusion monitoring data and / or the continuous time-series curve. The anomaly initial judgment and confidence calculation module is used to determine at least one anomaly candidate segment based on the continuous time series curve and the quality score, assign an anomaly stage type identifier to the anomaly candidate segment and calculate the corresponding initial confidence to form confidence information. The active perturbation control module is used to generate and output a control sequence for the infusion adjustment actuator based on a preset control strategy when the anomaly type identifier and initial confidence level output by the anomaly initial judgment and confidence level calculation module meet the preset trigger conditions. This allows for the application of perturbation to the infusion process within a controlled range, and the acquisition of accurate infusion monitoring data before and after the control sequence is applied. The early warning and decision-making module is used to extract response fingerprints to characterize the dynamic response of the infusion process based on the accurate infusion monitoring data before and after the control sequence, update the confidence of the abnormal candidate segments, obtain the abnormal judgment result, and output the risk classification information and treatment strategy information associated with the abnormal judgment result.

2. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The precise infusion monitoring data includes at least one or more of the following: drip rate related data, weight related data, liquid level related data, and drip detection related data; The drip rate related data includes time-series drip rate information calculated based on drop count statistics or drop intervals; the weight related data includes load change information of the infusion device obtained based on weighing monitoring; the liquid level related data includes time-series information representing the height or remaining weight of the liquid; and the drip detection related data includes statistical information on drip event detection results or drip event sequences generated by optical detection or image detection devices.

3. The intelligent monitoring system for the infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The infusion quality scoring and continuous curve analysis module is used to perform at least one processing on the precise infusion monitoring data to generate the continuous time-series curve, and to determine the quality score based on the precise infusion monitoring data and / or the continuous time-series curve. The processing includes at least one of filtering, smoothing, time alignment, sampling rate unification, and feature parameter calculation. The quality score is used to represent the state level of the infusion process in a unified metric space.

4. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The anomaly initial judgment and confidence calculation module determines the anomaly candidate segment based on the segment characteristics of the continuous time series curve. The segment characteristics include one or more of the following: rate of change characteristics, stability characteristics, and persistence characteristics. The anomaly initial judgment and confidence calculation module is used to perform joint analysis of the segment features on multiple time-series channels, mark the time-series segments that meet the preset segment feature conditions, collect the marked time-series segments into anomaly candidate segment set, and generate the corresponding initial confidence based on the combination relationship between the segment features of each anomaly candidate segment and the quality score.

5. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The abnormal stage type includes at least one or more of the following: suspected blockage, suspected leakage, and abnormal flow rate; The preset triggering conditions include at least one of the following: the abnormal stage type belongs to a preset set and the initial confidence level is within a preset range, and / or at least one of the triggering condition sets constructed based on the number of abnormal occurrences, the duration of abnormal occurrences, and historical record information, to control whether to invoke the active perturbation control module.

6. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The control sequence output by the active perturbation control module includes at least one of the following: switching between the first and second states of the actuator control quantity, a hold period, and a backoff period; The first state and the second state switching is used to switch between different infusion adjustment states, the hold period is used to maintain the control quantity in the selected state to observe the infusion process response, and the rollback period is used to restore the actuator to the target or reference state after the observation is completed.

7. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The early warning and decision-making module is used to: after the active perturbation control module outputs the control sequence, acquire accurate infusion monitoring data in the sliding window before and after the control sequence takes effect, construct a first feature set corresponding to the sliding window before the action and a second feature set corresponding to the sliding window after the action, and determine the response fingerprint based on the differential features of the first feature set and the second feature set. The response fingerprint is used to update the confidence level and form the anomaly determination result. The anomaly determination result is used at least to distinguish different anomaly types or to confirm whether an anomaly is valid.

8. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 7, characterized in that: The differential features include at least one or a combination of the following: variation amplitude differential, variation trend differential, response delay differential, recovery rate differential, and cross-channel consistency differential. The differential features are mapped as response fingerprints characterizing the response mode of the infusion process under perturbation, for use in the generation and recording of the confidence update and anomaly determination results.

9. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that: The early warning and decision-making module is used to determine the risk classification information based on one or more of historical time series information, medication plan information, and abnormal event context information, and output the treatment strategy information. The risk grading information is used to indicate the risk level or priority of the abnormal event. The handling strategy information includes at least one or more of the following: early warning push information, speed adjustment information, pause information, and manual confirmation mark information. The handling strategy information can be used to drive the infusion regulation actuator and / or provide interactive operation for remote terminals.

10. The intelligent monitoring system for infusion process based on precise infusion monitoring data as described in claim 1, characterized in that, The early warning and decision-making module is used to generate event packages for abnormal events that meet preset conditions and store or report them. The event package includes at least one or more of the following: raw or digested time-series data, parameter digests of the control sequence, response fingerprints, confidence information, and handling strategy information; Furthermore, the system is also used to obtain reference information of image or video data associated with the abnormal event when the abnormal event meets preset conditions, and to associate the reference information with the event package for use in quality control analysis and model update processes.

Images