Transfusion monitoring management method based on central monitoring

Abstract

The invention relates to the technical field of infusion monitoring management based on central monitoring, and discloses an infusion monitoring management method based on central monitoring. Collecting the residual volume at regular time through a liquid level sensor, calculating the instantaneous flow velocity by combining the volume difference and time difference of adjacent samples, setting the order of a fractional derivative, and recursively solving a generalized binomial coefficient to obtain a derivative value; constructing a sliding window according to a preset time window, extracting a median and a median absolute deviation, and adaptively generating upper and lower limit thresholds; and comparing the derivative and the threshold value mark abnormity in real time, and regularly summarizing to generate a monitoring report. According to the integrated process, the problems of noise sensitivity, lag and missing detection of traditional monitoring are solved, fractional order derivatives strengthen tiny drift perception, robust statistics restrains interference, a self-adaptive threshold gives consideration to sensitivity and robustness, automatic high-frequency sampling and periodic reporting improve monitoring precision and response speed, manual burden is relieved, and management is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of infusion monitoring management based on central monitoring, and in particular to an infusion monitoring management method based on central monitoring. Background Art

[0002] With the development of medical information technology and remote monitoring technology, infusion monitoring and management based on a central monitoring platform has become an important part of hospital intensive care, emergency treatment and general ward care. Existing technologies mainly rely on liquid level sensors, weight sensors or pressure sensors to perform simple differential or linear calibration on the real-time readings of infusion containers, and then determine infusion abnormalities using a fixed threshold or sliding average method. Specifically, the system usually collects changes in container volume or weight at preset time intervals and calculates the instantaneous flow rate through traditional first-order difference; then, it compares the upper and lower thresholds set by experience. Once the flow rate exceeds this range, it is determined to be an abnormality such as tube blockage, interruption or leakage.

[0003] First-order differences are slow to respond to small fluctuations in infusion flow rate, making it difficult to promptly capture subtle flow rate fluctuations caused by patient movement, minor tubing distortion, or syringe pump failure. Furthermore, traditional sliding average filters often have delayed responses, delaying the detection of true sudden changes and impacting clinical response time. In clinical settings, infusion lines are often subject to interference from vibration, patient movement, and pump oscillation, resulting in significant reading noise. Using fixed thresholds or simple mean addition and subtraction multipliers can easily lead to numerous false positives and false negatives, increasing nursing workload and undermining system credibility. Existing thresholds are often determined empirically or by preset multipliers, unable to dynamically adjust to real-time data fluctuations, making it difficult to balance sensitivity and stability. Factors such as patient variation, varying drug viscosity, and varying tubing resistance can all cause baseline flow rate variations, making static thresholds incompatible with multiple scenarios. Traditional methods rely solely on integer-order differentials and mean statistics, lacking in-depth feature extraction of the flow rate process. Furthermore, most algorithms employ "black-box" threshold determination, making it difficult to provide clear mathematical interpretation and clinical reference. Existing central monitoring systems usually only provide overall flow rate trend charts and simple event logs, lacking in-depth analysis of flow rate distribution characteristics (such as skewness and kurtosis) within short-term windows, making it difficult to support the accurate identification and positioning of infusion abnormalities under complex conditions.

[0004] To this end, this case aims to propose a centralized monitoring and management method for infusion monitoring. With a centralized monitoring platform at its core, liquid level sensors deployed along the infusion lines continuously collect the remaining volume of the infusion container at preset intervals and calculate the instantaneous flow rate in real time based on the difference between adjacent data. To address the subtle fluctuations and cumulative trends during the infusion process, fractional derivatives with "long memory" properties are introduced. Derivative weights are efficiently calculated recursively. Combined with robust statistical indicators within a sliding window—the median and its absolute deviation—a dynamic anomaly threshold is constructed to provide accurate and early warning of abnormal flow rates. Summary of the Invention

[0005] The present invention provides an infusion monitoring and management method based on central monitoring, which helps solve the problems mentioned in the above background technology.

[0006] The present invention provides the following technical solution: a method for managing infusion monitoring based on central monitoring, comprising: The liquid level sensor collects the remaining volume data of the infusion container in sequence at preset fixed time intervals, and calculates the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samplings; The order of the fractional derivative is set, the generalized binomial coefficient is calculated recursively, and it is combined with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at that moment; Determine the set of continuous fractional derivative values ​​contained in each sliding window according to the preset statistical window duration and sampling interval; In each sliding window, the fractional derivative values ​​are sorted in ascending order, and the value in the middle is selected as the median of the window; Calculate the absolute difference between each fractional derivative value in each sliding window and the window median, sort these differences in ascending order, and take the middle value as the median absolute deviation; Based on the window median and median absolute deviation, the upper and lower thresholds of the fractional derivative are determined according to the preset multiplication coefficients; The fractional derivative value calculated in real time is compared with the dynamic abnormal threshold. The value exceeding the threshold is marked as abnormal, and the value within the threshold range is marked as normal. The occurrence time and abnormal identification of each abnormal event detected are recorded, and all records are summarized into the central database at the end of the predetermined period to generate a periodic monitoring report.

[0007] Optionally, the method of sequentially collecting the remaining volume data of the infusion container at preset fixed time intervals by a liquid level sensor and calculating the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samples specifically includes: The initial sampling time is recorded as , the constant sampling interval is recorded as ; The liquid level sensor at each moment Direct reading of the remaining volume in the container ;in, is the sample index; For each , calculate the Subsampled instantaneous infusion rate .

[0008] Optionally, the order of the fractional derivative is set by recursively calculating the generalized binomial coefficient and combining it with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at the moment, specifically including: Set the order of fractional derivatives ; Construct a recursive calculation of the generalized binomial coefficient: ; in, The order is , the number of items is The generalized binomial coefficient of ; where is the order index of the generalized binomial coefficient; For each Construct fractional derivatives: ;in, For the The subsampled fractional derivative value.

[0009] Optionally, determining the set of continuous fractional derivative values ​​contained in each sliding window according to a preset statistical window duration and sampling interval specifically includes: Set the statistical window length to ; Calculate the number of sampling points in the window ; Construct the latest A set of fractional derivative values : .

[0010] Optionally, within each sliding window, the fractional derivative values ​​are arranged in ascending order, and the value in the middle position is selected as the median of the window, specifically including: Pair Collection The elements in are arranged in non-decreasing order as ;in, For collection The number of elements in Calculate the The median of the fractional derivatives in the time window , specifically: .

[0011] Optionally, the step of calculating the absolute differences between each fractional derivative value in each sliding window and the window median, arranging these differences in ascending order, and taking the middle value as the median absolute deviation specifically includes: Construct a set of absolute deviations: ;in, For the The set of absolute deviations at the moment; is the absolute value function; Get the Median absolute deviation of the time .

[0012] Optionally, the determining of the upper and lower thresholds of the fractional derivative based on the window median and the median absolute deviation according to a preset multiplication coefficient specifically includes: Set the standard deviation equivalent conversion constant ;in, is the inverse function of the Gaussian error function; Set the The upper threshold of the abnormal fractional derivative at the moment ; Set the The lower limit threshold of the fractional derivative anomaly at the moment .

[0013] Optionally, the real-time calculated fractional derivative value is compared with a dynamic abnormality threshold, and a value exceeding the threshold is marked as abnormal, and a value within the threshold is marked as normal, specifically including: For each , judge and generate abnormal marks: ; when When Sudden changes in flow rate occurred during the sub-sampling; when When The flow rate was normal during the second sampling.

[0014] Optionally, the occurrence time and abnormality identification of each abnormal event detected are recorded, and all records are aggregated into a central database at the end of a predetermined period to generate a periodic monitoring report, specifically including: whenever , generating the Records : ; All The sequence is stored in the central database; By statistical time window The cycle summarizes all , and output abnormal monitoring report.

[0015] The present invention has the following beneficial effects: 1. The liquid level sensor is coupled with the system clock, and the remaining volume of the container is automatically collected at fixed intervals. The instantaneous flow rate is obtained by the ratio of the data difference and the time difference between the front and back. Without relying on the flow estimation of the infusion pump itself or manual transcription, high-precision flow rate data can be obtained with an external sensor. Through high-frequency sampling, any sudden fluctuations during infusion can be captured, avoiding missed diagnosis due to average rate masking short-term abnormalities. The traditional monitoring method solves the problem of slow response to occasional rate changes, tedious and error-prone manual recording. Compared with the prior art, this scheme no longer uses fixed threshold or static rate evaluation method, but dynamically updates the flow rate index in real time, improving monitoring sensitivity and system response speed.

[0016] 2. Fractional derivative is introduced into infusion flow rate analysis, and the "long memory" feature of fractional calculus is used to capture small cumulative changes, rather than being limited to traditional integer order difference. The algorithm efficiently calculates the generalized binomial weight through recursion, without pre-storing the full coefficient, significantly reducing the calculation and storage overhead. Fractional derivative can consider both historical flow rate changes and current trends, and is more sensitive to slight but continuous abnormalities; by recursively updating the coefficients, real-time online operation can be realized on embedded or edge devices, without relying on powerful computing power. In reality, subtle flow rate drift during infusion is often a sign of early pipeline blockage or liquid leakage. Integer order method is easily masked by noise or overreacted, while this scheme accurately identifies such implicit risks using fractional dynamic differentiation. Compared with traditional algorithms that rely only on first or second order difference, fractional derivative considers both historical dependence and current responsiveness, with significantly superior compatibility and robustness.

[0017] 3. A configurable sliding time window is introduced to flexibly adjust the window length and sampling frequency according to the clinical scene requirements, realizing local review and dynamic rolling of historical data. By customizing the sliding window length, the best balance between emphasizing fast response and resisting occasional noise can be achieved; the decoupling design of window length and sampling interval allows flexible deployment under different device performance and monitoring needs; in clinical emergency or intensive care scenarios that require second-level response, shorter windows can be selected to enhance agility, while longer windows can be used in regular ward monitoring to suppress short-term jitter. In reality, fixed windows often struggle to balance response speed and false alarm rate. This scheme solves the limitations of traditional algorithms in various application scenarios, efficiently filtering noise while capturing real trend changes in time.

[0018] 4. Sort the fractional derivative values ​​within each sliding window in ascending order and select the median, replacing the commonly used arithmetic mean, to significantly improve the ability to resist outliers and extreme data. As a robust statistic, the median is insensitive to occasional measurement jitter or sudden data anomalies, effectively preventing false alarms caused by single-point deviations. It can also maintain the stability of the threshold in the presence of air bubbles, large amounts of jitter, or sensor jitter in the pipeline. The calculation process is simplified, and the core calculation can be completed by sorting and taking the median, making it easy to implement in resource-constrained environments. In reality, traditional monitoring relies on average or variance assessment, which can be easily pulled up or down by single abnormal data, affecting the accuracy of the threshold. This solution uses median statistics to maintain the simplicity of the algorithm while achieving a highly robust grasp of the central trend of the data distribution, effectively improving the reliability of anomaly identification.

[0019] 5. Calculate the absolute difference between each fractional derivative value and the median of the window in each sliding window, and take the median again as the absolute deviation indicator to measure the local data dispersion. As a robust dispersion measure, the median absolute deviation does not rely on the Gaussian distribution assumption and is also applicable to asymmetric and non-normally distributed data. It replaces the standard deviation in the presence of extreme outliers and can more accurately reflect the typical deviation level. The calculation method based on absolute difference has low computational complexity and is suitable for real-time online processing. The practical problem is that infusion monitoring data often exhibits nonlinear fluctuations and multi-source noise, and the standard deviation is easily affected by outliers and misjudged. The absolute deviation algorithm of this scheme reduces the risk of false alarms and provides a stable discreteness basis for subsequent threshold construction. Compared with traditional technologies that rely only on variance or mean square error, it has more anti-interference and reliability advantages.

[0020] 6. Based on the sum of the absolute deviations between the median and the median of the sliding window, a dynamic threshold with symmetrical upper and lower limits is constructed, and the balance between sensitivity and robustness can be flexibly adjusted through a multiplier coefficient. The dynamic threshold can be adaptively adjusted in real time according to the data characteristics, fully considering the noise level and trend changes of the current monitoring environment; the safety factor can be preset for different wards or clinical scenarios to implement personalized monitoring strategies; the threshold construction method does not rely on the global statistical model, avoiding overfitting of long historical data. In reality, fixed thresholds will cause omissions or false alarms due to different fluctuations in flow velocity. The adaptive dynamic threshold of this solution effectively solves the problem that traditional threshold presets are difficult to take into account various abnormal scenarios. Compared with the conventional method of using only static upper and lower limits, it can better balance sensitivity and reliability and reduce the cost of subsequent manual intervention.

[0021] 7. The fractional derivative value calculated in real time is immediately compared with the dynamic threshold generated at the corresponding moment to achieve real-time marking and response to abnormal fluctuations. In the event of a threshold violation, an alarm can be triggered immediately or a preset disposal process can be started to ensure early risk intervention; by quickly locating the abnormal flow rate point, an accurate basis can be provided for subsequent cause analysis and troubleshooting; the entire process does not rely on offline batch operations to ensure the real-time and stability of the system. The real problem is that traditional alarm systems often miss diagnoses or false alarms due to delays or improper threshold settings. This solution greatly improves the capture rate of minor but persistent anomalies and reduces false triggers caused by occasional noise by combining fractional-order refined indicators with dynamic thresholds, achieving efficient and reliable protection of infusion safety.

[0022] 8. Automatically record the time and identification of each abnormal event detected, and summarize them in the central database after the preset period to generate a structured and visual monitoring report. The entire process is automated without manual intervention, reducing the burden of repetitive record-keeping for medical staff; the summary report includes the time distribution, frequency statistics and severity assessment of abnormal events, providing an intuitive basis for clinical quality improvement and risk management; data based on central storage can also be used for long-term trend analysis and algorithm optimization. In reality, manual logs are prone to omissions or difficult to perform batch analysis. This solution achieves a closed-loop data system for the infusion monitoring process through automated recording and periodic reporting, providing stable and reliable data support for the continuous optimization of monitoring strategies and clinical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0024] 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.

[0025] Example, see Figure 1 , a central monitoring-based infusion monitoring management method, comprising: The liquid level sensor collects the remaining volume data of the infusion container in sequence at preset fixed time intervals, and calculates the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samplings; The order of the fractional derivative is set, the generalized binomial coefficient is calculated recursively, and it is combined with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at that moment; Determine the set of continuous fractional derivative values ​​contained in each sliding window according to the preset statistical window duration and sampling interval; In each sliding window, the fractional derivative values ​​are sorted in ascending order, and the value in the middle is selected as the median of the window; Calculate the absolute difference between each fractional derivative value in each sliding window and the window median, sort these differences in ascending order, and take the middle value as the median absolute deviation; Based on the window median and median absolute deviation, the upper and lower thresholds of the fractional derivative are determined according to the preset multiplication coefficients; The fractional derivative value calculated in real time is compared with the dynamic abnormal threshold. The value exceeding the threshold is marked as abnormal, and the value within the threshold range is marked as normal. The occurrence time and abnormal identification of each abnormal event detected are recorded, and all records are summarized into the central database at the end of the predetermined period to generate a periodic monitoring report.

[0026] By continuously and regularly collecting the remaining volume in the infusion container, calculating the instantaneous flow rate by combining two adjacent sampling values ​​with the corresponding time interval, and then introducing fractional derivatives to reflect the historical cumulative change trend of the flow rate, the entire process establishes an integrated monitoring chain from data collection, derivative calculation, window statistics, robust indicator extraction, dynamic threshold construction, to abnormality marking and report output. This solves the problem of traditional infusion monitoring that relies solely on built-in pump feedback or manual transcription of flow rate data, which is susceptible to occasional noise, monitoring lag, and hidden risks of missed detection. It also avoids the defect that threshold setting based on average rate can easily mask short-term mutations and cause serious missed or false alarms. Automated collection and real-time calculations have greatly improved the frequency and accuracy of data sampling; the "long memory" characteristics of fractional derivatives have enhanced sensitivity to persistent small drifts; the sliding window combined with robust statistics such as the median and median absolute deviation has not only suppressed occasional abnormal interference, but also ensured timely response to real trend changes; dynamic threshold adaptive adjustment allows the monitoring system to flexibly set alarm boundaries according to the actual fluctuation level, effectively balancing sensitivity and robustness; periodic reports provide clinical staff with objective and intuitive monitoring basis through automated summary and visualization, reducing the burden of manual recording and optimizing ward management. While improving the accuracy and response speed of infusion monitoring, the overall solution also provides reliable data support for centralized monitoring of multiple wards and subsequent quality improvement, significantly enhancing infusion safety and management efficiency.

[0027] The method of sequentially collecting the remaining volume data of the infusion container at preset fixed time intervals by the liquid level sensor and calculating the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samples specifically includes: The initial sampling time is recorded as , the constant sampling interval is recorded as ; The liquid level sensor at each moment Direct reading of the remaining volume in the container ;in, Index the samples; establish a uniform time grid for subsequent rate and derivative calculations; The function will index Mapped to real time; For each , calculate the Subsampled instantaneous infusion rate ; Obtain the instantaneous infusion rate by differential derivation; is the first-order difference quotient, indicating the The instantaneous flow rate at a moment.

[0028] A liquid level sensor automatically reads the remaining volume of the infusion container at preset fixed time intervals and calculates the instantaneous infusion flow rate based on the volume difference between two consecutive samples and the corresponding time interval. This solves the delay and error issues commonly encountered in traditional monitoring methods, which rely on the pump's built-in flow rate estimation or nurses' regular meter readings. In practice, simply installing an external liquid level sensor in the infusion line and configuring a unified sampling clock can obtain high-precision, equidistant volume data. The instantaneous flow rate is then obtained by dividing the volume change by the sampling interval, completely bypassing the device's inherent feedback errors and the arbitrariness of manual operation. The sampling frequency can be flexibly adjusted to meet the monitoring needs of different wards. The adjacent difference method is highly sensitive to short-term flow rate fluctuations and can promptly capture early anomalies. It also avoids omissions in manual recording and drift in mechanical feedback, improving the reliability and integrity of the original monitoring data and providing a solid data foundation for subsequent anomaly identification and early warning strategies.

[0029] The order of the fractional derivative is set, and the generalized binomial coefficient is calculated recursively, and the coefficient is combined with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at the moment, specifically including: Set the order of fractional derivatives ;Define the required fractional derivative order for sensitivity enhancement; Dimensionless, directly determines the derivative weight distribution; Construct a recursive calculation of the generalized binomial coefficient: ; in, The order is , the number of items is The generalized binomial coefficient of ; where is the order index of the generalized binomial coefficient; the generalized binomial coefficient is calculated step by step, supporting the summation of fractional derivatives; The coefficient of the generalized binomial is calculated recursively without pre-storing all values; For each Construct fractional derivatives: ;in, For the The fractional derivative value of the sampling is taken at a certain time; the fractional derivative of the instantaneous velocity series is taken at a certain time to capture the trend of small mutations.

[0030] Based on the instantaneous flow velocity, fractional derivatives are introduced. The weight coefficients required for fractional differentiation are recursively generated and combined with the instantaneous flow velocity data at each moment to produce fractional derivative values ​​that reflect the cumulative trend of historical flow velocity changes. This overcomes the limitation of traditional integer-order differentiation, which focuses only on the difference between the current or a few adjacent points and is unable to capture long-term accumulated drift. The fractional-order derivative has a "long memory" property, allowing it to consider both historical paths and current velocity changes within the same time window, thus providing enhanced tracking and amplification of persistent but small flow velocity deviations. In terms of implementation, the recursive calculation of generalized weight coefficients eliminates the need to store a large number of historical values, significantly reducing computational and storage burdens and making it suitable for online operation in embedded or resource-constrained devices. This method provides earlier and more reliable warnings for hidden faults such as initial pipeline blockages, leaks, and pump head obstructions. It also mitigates the over-amplification of transient noise and the slow response to slow drift that can occur with integer-order methods. It improves the sensitivity and stability of the entire warning chain, providing more targeted abnormality alerts for clinical and equipment maintenance personnel, reducing misdiagnosis and secondary inspection workloads.

[0031] The step of determining a set of continuous fractional derivative values ​​contained in each sliding window according to a preset statistical window duration and sampling interval specifically includes: Set the statistical window length to ; Define the scope of historical data review and balance response speed and robustness; Dimensionless parameter, directly determines the number of subsequent sampling points; Determined at the current moment How long to look ahead to collect fractional derivative sequences Used to calculate the median, , threshold calculation; through the formula The time window length Convert to the number of sample points included , ensuring that the sliding window can cover the most recent All data within seconds; When it is larger, more historical points are involved in the statistics, the median and It is insensitive to short-term fluctuations and can filter out occasional noise. After a real mutation occurs, it is necessary to wait for more new data to enter the window before the median and threshold can move to a new level, which increases the alarm delay. When it is small, only data in a very short interval are used, the median and Keep up with the latest changes and quickly respond to sudden changes in flow rate; due to the small number of samples, occasional noise or measurement jitter is more likely to affect the median and threshold, which may cause false alarms. Adjust according to the trade-off between "quick response" and "false alarm rate" in clinical or monitoring scenarios: if the timeliness of mutation detection is high, it is recommended to ; High requirements on system stability and noise resistance, it is advisable The optimal value can be verified through historical data playback experiments (offline simulation) to find a balance between timely detection of anomalies and avoiding excessive false alarms.

[0032] Calculate the number of sampling points in the window ;Convert the time window length into the number of sample points for easy discrete processing; the floor function ensures that sufficient historical data is included; Construct the latest A set of fractional derivative values : ; Collect local historical derivative values ​​for statistical feature extraction; Set Represents the data samples within the sliding window.

[0033] A sliding window mechanism based on a preset statistical time window is proposed. Fractional derivative data within the corresponding time range are aggregated into a statistical subset based on a set duration, and the window is continuously updated as the sampling progresses. This resolves the contradiction between static global statistics and the inability to balance monitoring response speed and noise suppression. The sliding window can dynamically configure the window length according to different clinical scenarios: a short window facilitates second-level response and timely captures acute flow velocity mutations; a long window enhances robustness, filters occasional jitter and sensor noise, and reduces false alarm rates. In terms of implementation, the time window length is mapped to the number of discrete sampling points, ensuring that each statistic is based on data from the most recent period. The window rolling mechanism is simple and efficient, facilitating real-time online calculations. It achieves dual adaptation to both oscillatory and mutational anomalies, meeting the different sensitivity requirements of intensive care units and conventional wards. It provides an adjustable historical data support point for subsequent robust statistics and threshold construction, enabling personalized and scenario-based optimization of monitoring strategies.

[0034] In each sliding window, the fractional derivative values ​​are arranged in ascending order, and the value in the middle position is selected as the median of the window, specifically including: Pair Collection The elements in are arranged in non-decreasing order as ;in, For collection The number of elements in the median; prepare an ordered sequence for median calculation; sort function rearranges the data in ascending order; Calculate the The median of the fractional derivatives in the time window , specifically: ; Use the median to measure the central tendency, which has strong anti-interference ability; half-split function, take the middle value or average.

[0035] The fractional derivative values ​​in each sliding window are arranged in ascending order and the value in the middle of the sequence is selected as the median of the window to measure the central trend of the period. This solves the problem that the arithmetic mean is easily disturbed by extreme anomalies, resulting in a shift in the central trend. The use of the median can effectively ignore occasional spikes or extreme deviations, so that the central trend measurement remains stable for a small number of abnormal samples. In the presence of bubbles, sensor jitter or instantaneous impact on the pipeline, the median will not be deviated by a single extreme value; when deployed on a large scale in multiple wards, a highly robust central trend indicator can be obtained while ensuring simplicity of calculation; the subsequent threshold construction and deviation measurement input are simplified, providing a more reliable reference basis for dynamic anomaly judgment, thereby further reducing the false alarm rate and improving the system's ability to identify real anomalies.

[0036] The calculation of the absolute difference between each fractional derivative value in each sliding window and the window median, and arranging these differences in ascending order, taking the middle value as the median absolute deviation, specifically includes: Construct a set of absolute deviations: ;in, For the The set of absolute deviations at the moment; is the absolute value function; it measures the deviation of each point from the center; the absolute value function eliminates the sign and retains the magnitude; Get the Median absolute deviation of the time ;by It measures the dispersion of data and is more robust than the standard deviation; the median is recalculated to extract typical deviations.

[0037] The absolute difference between each fractional derivative value and the median of the window is calculated in each sliding window, and the median of the difference sequence is taken again as the median absolute deviation indicator to measure the dispersion. This solves the problem that the dispersion measurement based on standard deviation or variance is easily affected by non-normal distribution and outliers. The median absolute deviation does not require the assumption of data distribution and is also insensitive to extreme values, and can truly reflect the typical deviation level. When the infusion flow rate data presents multi-source noise and nonlinear fluctuations, the obtained dispersion indicator can better reflect the normal range; it provides a robust and accurate deviation quantification for the adaptive construction of upper and lower thresholds; and the calculation process only requires two median operations and several absolute difference calculations. The algorithm has low complexity and is suitable for real-time online applications, further improving the robustness and real-time performance of anomaly detection.

[0038] The upper and lower thresholds of the fractional derivative are determined based on the window median and the median absolute deviation according to a preset multiplication coefficient, specifically including: Set the standard deviation equivalent conversion constant ;in, is the inverse function of the Gaussian error function; Equivalent conversion to standard deviation scale; inverse error function generates standard normal corresponding factor; Set the The upper threshold of the abnormal fractional derivative at the moment ; Set the The lower limit threshold of the fractional derivative anomaly at the moment ; Construct a symmetric threshold interval for anomaly determination; a linear combination function combines the center and the dispersion.

[0039] Based on the absolute deviation between the median and the median of the sliding window, symmetrical dynamic upper and lower thresholds are constructed according to the preset multiplier coefficient for subsequent abnormality judgment. This solves the problem that static fixed thresholds are difficult to maintain adaptability in different fluctuating environments. The dynamic threshold can be adaptively adjusted according to the characteristics of real-time data. It can not only automatically expand the threshold range according to the current noise amplitude to prevent excessive false alarms, but also timely narrow the threshold range to enhance sensitivity when there is continuous drift or abnormal trends. It realizes the personalization and scenario-based setting of thresholds, and can flexibly adjust the multiplier coefficient according to the characteristics of the ward and risk preferences; it avoids the tediousness of frequent manual calibration of traditional thresholds; it strikes a balance between system noise resistance and timely warning, reduces the dual risks of missed alarms and false alarms, and provides a reliable and easy-to-maintain threshold strategy for clinical infusion monitoring.

[0040] The real-time calculated fractional derivative value is compared with the dynamic abnormality threshold, and the value exceeding the threshold is marked as abnormal, and the value within the threshold is marked as normal, specifically including: For each , judge and generate abnormal marks: ; when When Sudden changes in flow rate occurred during the sub-sampling; when When The flow rate was normal during the second sampling; Instant judgment Whether the subsampling is normal; indicator function, converting the logical judgment into binary value.

[0041] The fractional derivative values ​​obtained by real-time calculation are immediately compared with the dynamic thresholds generated at the corresponding moment. Those exceeding the threshold range are marked as abnormal, and those within the range are marked as normal. This solves the lag problem of traditional batch offline judgment or manual review, and realizes a full-link real-time closed loop from data collection to abnormality judgment. The instant comparison of each sample can trigger an alarm or response plan within the shortest delay to ensure the safety of infusion. It improves the timeliness of early risk intervention; provides accurate timestamps and abnormality intensity information for subsequent cause investigation; cooperates with the central monitoring system to realize parallel monitoring of multiple wards and centralized alarm management, reducing the pressure of on-site manual inspections; ensures the active controllability of the entire infusion monitoring process and improves the level of clinical safety.

[0042] The time of occurrence and abnormality identification of each abnormal event detected are recorded, and all records are aggregated into a central database at the end of a predetermined period to generate a periodic monitoring report, specifically including: whenever , generating the Records : ;Record complete exception information for easy tracing and analysis; Tuple organization function, package various indicators; All Sequentially store data in a central database; establish long-term exception logs to support subsequent audits; insert functions into the database to ensure persistence; By statistical time window The cycle summarizes all , and output exception monitoring reports; regularly present system exception overviews to assist decision-making; aggregate and format functions to generate easy-to-read reports.

[0043] The system automatically records the occurrence time and abnormal identification of each abnormal event detected, and summarizes all records into the central database after the preset period to generate a structured and visual periodic monitoring report. This solves the problems of untimely manual recording, omissions, and difficulty in batch analysis, and realizes automated archiving and centralized management of the entire process. The periodic report not only provides a statistical summary and time distribution of abnormal events, but also includes a severity assessment, providing sufficient data support for the clinical team to carry out quality improvement and risk control. It reduces the burden of repeated records on medical staff; through visual reports, the system operation and abnormal situation are intuitively presented, making it easier for management to quickly grasp the overall monitoring status; historical data can also be used for long-term trend analysis and monitoring strategy optimization, promoting the continuous improvement and refined management of infusion safety monitoring.

[0044] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0045] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for infusion monitoring and management based on central monitoring, characterized in that: include: The liquid level sensor collects the remaining volume data of the infusion container in sequence at preset fixed time intervals, and calculates the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samplings; The order of the fractional derivative is set, the generalized binomial coefficient is calculated recursively, and it is combined with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at that moment; Determine the set of continuous fractional derivative values ​​contained in each sliding window according to the preset statistical window duration and sampling interval; In each sliding window, the fractional derivative values ​​are sorted in ascending order, and the value in the middle is selected as the median of the window; Calculate the absolute difference between each fractional derivative value in each sliding window and the window median, sort these differences in ascending order, and take the middle value as the median absolute deviation; Based on the window median and median absolute deviation, the upper and lower thresholds of the fractional derivative are determined according to the preset multiplication coefficients; The fractional derivative value calculated in real time is compared with the dynamic abnormal threshold. The value exceeding the threshold is marked as abnormal, and the value within the threshold range is marked as normal. The occurrence time and abnormal identification of each abnormal event detected are recorded, and all records are summarized into the central database at the end of the predetermined period to generate a periodic monitoring report.

2. The method for managing infusion monitoring based on central monitoring according to claim 1, characterized in that: The method of sequentially collecting the remaining volume data of the infusion container at preset fixed time intervals by the liquid level sensor and calculating the instantaneous infusion flow rate based on the volume difference and time difference between two adjacent samples specifically includes: The initial sampling time is recorded as , the constant sampling interval is recorded as ; The liquid level sensor at each moment Direct reading of the remaining volume in the container ;in, is the sample index; For each , calculate the Subsampled instantaneous infusion rate .

3. The method for managing infusion monitoring based on central monitoring according to claim 2, characterized in that: The order of the fractional derivative is set, and the generalized binomial coefficient is calculated recursively, and the coefficient is combined with the instantaneous infusion flow rate data at the corresponding moment to obtain the fractional derivative value at the moment, specifically including: Set the order of fractional derivatives ; Construct a recursive calculation of the generalized binomial coefficient: ； in, The order is , the number of items is The generalized binomial coefficient of ; where is the order index of the generalized binomial coefficient; For each Construct fractional derivatives: ;in, For the The fractional derivative value of the subsample.

4. The method for managing infusion monitoring based on central monitoring according to claim 3, characterized in that: The step of determining a set of continuous fractional derivative values ​​contained in each sliding window according to a preset statistical window duration and sampling interval specifically includes: Set the statistical window length to ; Calculate the number of sampling points in the window ; Construct the latest A set of fractional derivative values : 。 5. The method for managing infusion monitoring based on central monitoring according to claim 4, characterized in that: In each sliding window, the fractional derivative values ​​are arranged in ascending order, and the value in the middle position is selected as the median of the window, specifically including: Pair Collection The elements in are arranged in non-decreasing order as ;in, For collection The number of elements in Calculate the The median of the fractional derivatives in the time window , specifically: 。 6. The method for managing infusion monitoring based on central monitoring according to claim 5, characterized in that: The calculation of the absolute difference between each fractional derivative value in each sliding window and the window median, and arranging these differences in ascending order, taking the middle value as the median absolute deviation, specifically includes: Construct a set of absolute deviations: ;in, For the The set of absolute deviations at the moment; is the absolute value function; Get the Median absolute deviation of the time .

7. The method for managing infusion monitoring based on central monitoring according to claim 6, characterized in that: The upper and lower thresholds of the fractional derivative are determined based on the window median and the median absolute deviation according to a preset multiplication coefficient, specifically including: Set the standard deviation equivalent conversion constant ;in, is the inverse function of the Gaussian error function; Set the The upper threshold of the abnormal fractional derivative at the moment ; Set the The lower limit threshold of the fractional derivative anomaly at the moment .

8. The method for managing infusion monitoring based on central monitoring according to claim 7, characterized in that: The real-time calculated fractional derivative value is compared with the dynamic abnormality threshold, and the value exceeding the threshold is marked as abnormal, and the value within the threshold is marked as normal, specifically including: For each , judge and generate abnormal marks: ； when When Sudden changes in flow rate occurred during the sub-sampling; when When The flow rate was normal during the second sampling.

9. The method for managing infusion monitoring based on central monitoring according to claim 8, characterized in that: The time of occurrence and abnormality identification of each abnormal event detected are recorded, and all records are aggregated into a central database at the end of a predetermined period to generate a periodic monitoring report, specifically including: whenever , generating the Records : ； All The sequence is stored in the central database; By statistical time window The cycle summarizes all , and output abnormal monitoring report.

Images