Drip early warning and constant temperature regulation infusion safety control method

Abstract

This invention relates to the field of medical infusion monitoring technology, specifically to an infusion safety control method with drip warning and constant temperature regulation. The method includes: acquiring a continuous video stream of liquid dripping from the drip chamber of the infusion tubing and converting it into a sequence of image frames; performing contour enhancement and background separation processing on the image frames; extracting isolated foreground regions of the droplets inside the drip chamber and calculating their pixel area; determining the start of a dripping event based on the appearance of a new isolated foreground region; determining the completion of droplet detachment and marking the end of dripping based on the pixel area change rate of consecutive preset frames before the region disappears; recording the actual duration of a single drip; and accumulating the measured drip frequency within a set time period; comparing the measured values ​​with a preset safe infusion frequency range; and triggering an abnormal infusion rate warning when the measured values ​​continuously exceed the range. This method can accurately identify droplet image features, accurately determine the dripping cycle, improve the accuracy of infusion rate detection, and achieve reliable early warning of infusion abnormalities.

Description

Technical Field

[0001] This invention relates to the field of medical infusion monitoring technology, and in particular to an infusion safety control method with drip early warning and constant temperature regulation. Background Technology

[0002] Existing infusion monitoring and temperature control technologies mostly use photoelectric sensing, gravity sensing, or conventional image acquisition methods to monitor droplet falling. They determine the falling behavior by the physical signal changes caused by the droplets, obtain the infusion rate parameters by counting the number of drops per unit time, and complete the abnormal prompts based on preset thresholds. They also achieve basic safety control of the infusion process in conjunction with conventional temperature control structures. However, there is still a lack of a refined image analysis and accurate judgment system for the dynamic changes of droplets.

[0003] Existing detection methods are susceptible to interference from factors such as ambient light, air bubbles in the tubing, and the structure of the drip chamber. The accuracy of signal acquisition and image recognition is significantly limited, and there is a lack of refined analytical methods for the complete dynamic process of droplet formation, growth, and detachment. Conventional image acquisition does not include contour enhancement and background separation processing, making it difficult to effectively extract isolated foreground regions of droplets inside the drip chamber. Furthermore, there is a lack of corresponding quantification methods for the pixel area of ​​these regions, making it difficult to accurately determine dripping events based on image parameters. Current technologies mostly determine dripping behavior using simple signals, and the identification of the dripping cycle relies on basic physical signals. The recording accuracy of the actual duration of a single drip is insufficient, and the calculated drip frequency deviates from the actual infusion state. Therefore, the accuracy of abnormal infusion rate warnings is insufficient to meet the needs of clinical safety management.

[0004] This invention aims to achieve independent extraction and numerical quantification of the effective image area of ​​droplets within the drip chamber, complete the full-cycle determination of the dripping event through the dynamic change characteristics of the droplet image area, and achieve accurate early warning of abnormal infusion rate based on precise dripping frequency calculation, thus making up for the accuracy shortcomings of existing technologies in droplet identification, dripping cycle determination, and rate monitoring. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and propose an infusion safety control method with drip warning and constant temperature regulation.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an infusion safety control method with drip warning and constant temperature regulation, comprising: Acquire a continuous video stream of liquid dripping at the dripping location of the infusion tubing, and convert the liquid dripping video stream into a sequence of image frames; Contour enhancement and background separation processing are performed on the sequence of image frames to extract the isolated foreground region of the droplet inside the droplet in each frame image, and the image pixel area of ​​the isolated foreground region is calculated. When a newly emerging isolated foreground region is detected, a dripping event is determined to have started, and the process of recording the change of the image pixel area of ​​the isolated foreground region corresponding to the dripping event in consecutive frames is started until the isolated foreground region disappears. Based on the rate of change of image pixel area in consecutive preset frames before the disappearance of the isolated foreground region, it is determined whether the droplet has completed the process of detaching from the upper port of the drip chamber. When it is determined that the detachment has been completed, a dripping event is marked as ended. Record the time interval between the start and end of the dripping event as the actual duration of a single drip, and accumulate the actual duration of all single drips within a set time period to calculate the measured value of the liquid dripping frequency within the set time period. The measured value of the liquid dripping frequency is compared with the preset safe infusion frequency range. When the measured value continuously exceeds the safe infusion frequency range, an abnormal infusion rate warning signal is triggered.

[0007] As a further aspect of the present invention, contour enhancement and background separation processing are performed on the sequence of image frames to extract the isolated foreground region of the droplet inside the droplet in each image frame, including: The input sequence of image frames is converted to grayscale, and a Gaussian filtering algorithm is used to smooth the grayscale image and suppress image noise. An edge detection operator is applied to the smoothed image to highlight the edge contours of all objects inside the drip chamber, thus obtaining an initial edge image. In the initial configuration phase, a background image of the dripping chamber in a state without dripping is acquired, and the background image of the dripping chamber is subtracted from the current frame of the sequence image frame to obtain a difference image; The differential image is segmented by thresholding, and regions with pixel values ​​greater than the dynamic threshold are initially identified as a set of potential foreground pixels; The distribution locations of the initially identified potential foreground pixels are matched and fused with the contour locations in the initial edge image obtained by the edge detection operator. Isolated noise areas caused by changes in light are eliminated, and finally, the isolated foreground regions and their contours of the droplets inside the droplet are determined.

[0008] As a further aspect of the present invention, determining whether the droplet has completed the process of detaching from the upper port of the droplet based on the rate of change of image pixel area in consecutive preset frames before the disappearance of the isolated foreground region includes: After the dripping event begins, the image pixel area of ​​the isolated foreground region is continuously tracked, and the difference in the image pixel area between adjacent frames is calculated; When the area of ​​the image pixels is continuously increased, it is determined that the droplet is in the growth and falling stage; When it is detected that the area of ​​the image pixels begins to decrease continuously after reaching the maximum value, the area decay frame sequence is recorded. The area decay frame sequence contains multiple frames of image data that decrease continuously from the maximum area. Analyze the rate of decrease of the image pixel area in the area decay frame sequence. If the rate of decrease is continuously higher than the preset detachment rate threshold, it is determined that the droplet is in the process of rapid tearing and detachment. When the image pixel area of ​​the isolated foreground region decreases to below the preset minimum foreground area threshold, it is determined that the droplet has completely detached from the upper port of the dropper, that is, the detachment process is completed, and the drop event is marked as ended.

[0009] As a further aspect of the present invention, the actual duration of all single drops within the cumulative set time period is used to calculate the measured value of the liquid drop frequency within the set time period, including: Set the length of the time window used to calculate the frequency, and slide the time window on the time axis; Within each time window, identify all marked dripping events and obtain the actual duration corresponding to each dripping event; The total time consumed by the droplet falling within the time window is obtained by summing the actual duration of all single drops within the time window. The average drop interval within the time window is calculated by subtracting the total time consumed by the droplet's fall from the length of the time window and then dividing by the total number of drop events recorded within the time window. Divide sixty seconds by the average drip interval to obtain the measured value of the liquid drip frequency corresponding to the time window.

[0010] As a further aspect of the present invention, after calculating the measured value of the liquid dripping frequency within a set time period, the method further includes: The real-time temperature of the liquid is acquired by a thin-film temperature sensor wrapped around the outside of the infusion tubing and recorded as the tubing temperature measurement value. Read the preset target temperature value of the infusion fluid in the infusion parameter settings; At the beginning of the constant temperature control cycle, the measured value of the pipeline temperature is compared with the target temperature value of the liquid, and the temperature deviation at the current moment is calculated. The temperature deviation is input into a preset temperature increment control lookup table, which defines the control signal duty cycle adjustment range corresponding to different temperature deviation ranges. Based on the duty cycle adjustment range of the control signal obtained from the query, a drive command is generated for the semiconductor heating and cooling chip wrapped around the infusion tubing. The drive command is used to adjust the power output duty cycle of the semiconductor heating and cooling chip during the constant temperature control cycle.

[0011] As a further aspect of the present invention, the process of the thin-film temperature sensor acquiring the real-time temperature of the liquid includes: The thin-film temperature sensor is tightly wound around the outer wall of the infusion tubing, with its sensing surface completely in contact with the outer wall of the tubing. The thin-film temperature sensor collects the temperature of the outer wall of the infusion tubing it contacts at a fixed sampling period as the original sample value; A moving average filter is applied to multiple continuously collected raw sample values ​​to remove temperature jumps caused by environmental airflow or instantaneous contact changes. The filtered temperature value is combined with the sensor's factory calibration curve to calculate the pipeline temperature measurement value reflecting the liquid temperature inside the infusion tubing, and then output to the outside. Meanwhile, the thin-film temperature sensor continuously monitors the thermal resistance between itself and the hose. When an abnormal increase in thermal resistance is detected, the temperature value of the sampling period may be invalid.

[0012] As a further aspect of the present invention, after generating the drive command for the semiconductor heating / cooling chip wrapped around the infusion tubing, the method further includes: At the end of the constant temperature control cycle, the updated pipeline temperature measurement value collected by the thin-film temperature sensor is read again. Calculate the change between the updated pipeline temperature measurement value and the pipeline temperature measurement value of the previous cycle, and use it as the actual temperature change. Read the theoretical expected temperature change value corresponding to the duty cycle of the control signal used in the previous cycle from the temperature increment control lookup table; Compare the difference between the actual temperature change and the theoretical expected temperature change. If the absolute value of the difference exceeds the allowable threshold, mark the current heating and cooling power transfer efficiency as abnormal. When a preset number of consecutive constant temperature control cycles are detected as abnormal heating and cooling power transfer efficiency, an inspection prompt will be initiated to check the contact status between the semiconductor heating and cooling chip and the infusion tubing, as well as the attachment status of the thin-film temperature sensor.

[0013] As a further aspect of the present invention, after triggering the abnormal infusion rate warning signal, the infusion equipment linkage control based on the warning is also included: Identify the specific type of the abnormal infusion rate warning signal that was triggered, including warnings of excessively fast drip rate and warnings of excessively slow drip rate; When a warning of slow drip rate is detected, an adjustment command is automatically generated for the peristaltic pump or infusion clamp in the infusion line. The adjustment command includes a preset small opening increase to attempt to increase the fluid flow rate. After executing the adjustment command, a targeted monitoring process is initiated to re-acquire the liquid dripping video stream within the new set time period and calculate the updated measured value of the liquid dripping frequency. Determine whether the measured value of the updated liquid dripping frequency has returned to the safe infusion frequency range. If it has not returned, repeat the steps of generating a new small opening increase and adjusting until the preset maximum number of adjustments is reached or the flow rate returns to normal. When an alarm for excessively fast drip rate is detected, an instruction is immediately generated to completely shut down the peristaltic pump or infusion clamp in the infusion line, and an emergency alarm requiring manual confirmation is triggered.

[0014] As a further aspect of the present invention, after immediately generating an instruction to completely shut down the peristaltic pump or infusion clamp in the infusion line when an excessively fast drip rate warning is detected, and triggering an emergency alarm requiring manual confirmation, the method further includes: Record the measured value of the liquid dripping frequency when the excessive dripping rate warning is triggered, as well as the historical measured value sequence of the liquid dripping frequency within a preset time period before the trigger; After the emergency alarm is manually confirmed and handled, a post-event review and analysis will be initiated. The sequence image frames corresponding to the triggered warning are invoked, and the accuracy of the determination process for the end of the dripping event is analyzed, and the accuracy of the calculation of the actual duration of a single drip is verified. At the same time, check the historical records of control commands for the peristaltic pump or infusion clamp in the infusion line before the warning is triggered to confirm whether there are any abnormal or unexpected opening increase commands. Generate an early warning event analysis report, which includes drip rate data, image analysis summary, control command history, and preliminary judgment on the effectiveness of the early warning, for reference in system maintenance and parameter calibration.

[0015] As a further aspect of the present invention, the input sequence of image frames is converted to grayscale, and a Gaussian filtering algorithm is used to smooth the grayscale image and suppress image noise, including: Receive a continuous video stream of liquid dripping from the dripping location of the infusion tubing, and convert the liquid dripping video stream into a sequence of image frames to obtain the original image frame sequence; Apply a luminance and chrominance separation algorithm to each color image in the original image frame sequence, extract its luminance component, complete the grayscale conversion, and obtain the corresponding grayscale image frame sequence. A Gaussian kernel function is constructed, and a two-dimensional Gaussian filter kernel is generated based on the preset kernel size and standard deviation parameters. A two-dimensional convolution operation is performed on each grayscale image in the grayscale image frame sequence, and the image pixels are weighted and averaged to obtain a smoothed image frame sequence. After smoothing, the local pixel value standard deviation is calculated for each frame of the image frame sequence. Based on the local pixel value standard deviation, it is compared with a preset noise threshold to identify and mark potential residual noise regions, thus completing the image noise suppression process.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The video stream of continuously collected liquid dripping at the dripping location of the infusion tubing is converted into a sequence of image frames. Contour enhancement and background separation processing are performed on the sequence of image frames to extract the isolated foreground region of the droplet inside the dripping tub in each image frame. The image pixel area of ​​the isolated foreground region is calculated, which can eliminate interference from irrelevant image information such as the dripping tub itself and the external environment, isolate the influence of image signals from non-droplet areas, accurately locate the effective image range corresponding to the droplet, realize the independent extraction and numerical quantization of droplet image features, ensure the stability of droplet region recognition, and provide accurate image quantization parameters for the determination of dripping events.

[0017] The dripping event begins with the emergence of a new isolated foreground region. The changes in the image pixel area of ​​this isolated foreground region are recorded across consecutive frames. Based on the rate of change in the image pixel area of ​​consecutive preset frames before the isolated foreground region disappears, the process of the droplet detaching from the upper end of the drip chamber is determined, and the dripping event ends. The time interval between the start and end of the dripping event is recorded to obtain the actual duration of a single drip. The measured value of the liquid dripping frequency is calculated by accumulating the actual durations of all single drips within a set time period. This allows for a complete definition of the complete time cycle of a single drip, reducing errors caused by dynamic changes in droplet shape in determining the start and end of the dripping event, and improving the accuracy of dripping duration recording and frequency calculation. The measured value of the liquid dripping frequency is compared with a preset safe infusion frequency range. When the measured value continuously exceeds the safe infusion frequency range, an abnormal infusion rate warning signal is triggered. This allows for real-time identification of abnormal changes in the infusion rate, improving the accuracy and timeliness of abnormal infusion rate detection, and enhancing the reliability of infusion monitoring. Attached Figure Description

[0018] Figure 1 The flowchart is a method for infusion safety control with drip warning and constant temperature regulation according to the present invention. Figure 2 A flowchart for determining the process of droplet detachment from the upper port of the dropping chamber based on the rate of change of image pixel area; Figure 3 The pixel area change curve is shown for the entire process of droplet growth and shedding. Figure 4 For temperature calibration and isothermal target tracking curve; Figure 5 Monitoring process for automatic adjustment when the drip rate is too slow. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides an infusion safety control method with drip warning and constant temperature regulation, the specific method including: A camera fixed at the drip chamber of the infusion tubing acquires a continuous stream of video footage of the dripping liquid. The image processing unit converts this video stream into a sequence of image frames. Subsequently, contour enhancement and background separation are performed on the image frames to extract isolated foreground regions of the droplets inside the drip chamber in each frame, and the pixel area of ​​these isolated foreground regions is calculated. When the system detects a newly appearing isolated foreground region, a dripping event is declared to have begun, and the system starts recording the change in the pixel area of ​​the corresponding isolated foreground region across consecutive frames until the isolated foreground region disappears from the image. Based on the rate of change in pixel area across a preset number of frames before the isolated foreground region disappears, the system determines whether the droplet has completed its detachment from the drip chamber. When detachment is confirmed, the dripping event is marked as ended. The system records the time interval between the start and end of a dripping event as the actual duration of a single drip, and accumulates the actual durations of all single drips within a set time period to calculate the measured value of the liquid dripping frequency within that time period. Finally, the measured value of the liquid dripping frequency is compared with the preset safe infusion frequency range. When the measured value continues to exceed the safe infusion frequency range, the system triggers an abnormal infusion rate warning signal.

[0022] In one embodiment of the present invention, the input sequence of image frames is converted to grayscale, and a Gaussian filtering algorithm is used to smooth the grayscale images to suppress image noise. A continuous video stream of liquid dripping from the drip chamber of the infusion tubing is received and converted into a sequence of image frames to obtain an original image frame sequence. A luminance and chrominance separation algorithm is applied to each color image frame in the original image frame sequence to extract its luminance component, completing the grayscale conversion to obtain the corresponding grayscale image frame sequence. A Gaussian kernel function is constructed, and a two-dimensional Gaussian filter kernel is generated based on preset kernel size and standard deviation parameters. A two-dimensional convolution operation is performed on each grayscale image frame in the grayscale image frame sequence, and a weighted average is applied to the image pixels to obtain a smoothed image frame sequence. After smoothing, the local pixel value standard deviation is calculated for each frame of the image frame sequence. Based on the comparison between the local pixel value standard deviation and a preset noise threshold, potential residual noise regions are identified and marked to complete the suppression of image noise. An edge detection operator is applied to the smoothed image to highlight the edge contours of all objects inside the drip chamber, obtaining an initial edge image. In the initial configuration phase, a background image of the dripping bucket under no-droplet-fall conditions is acquired. This background image is subtracted from the current frame of the image sequence to obtain a difference image. Thresholding is applied to the difference image, and regions with pixel values ​​greater than a dynamic threshold are preliminarily identified as potential foreground pixel sets. The distribution locations of these preliminarily identified potential foreground pixel sets are matched and fused with the contour locations in the initial edge image obtained through edge detection operators. Isolated noise regions caused by lighting variations are eliminated, ultimately determining the isolated foreground regions and their contours within the dripping bucket.

[0023] In specific implementation, contour enhancement and background separation processing are performed on the input sequence of image frames to extract isolated foreground regions of droplets inside the drip chamber. This processing includes the following steps: receiving a continuously acquired video stream of liquid dripping from the drip chamber location of the infusion line and converting the liquid dripping video stream into a sequence of image frames to obtain an original image frame sequence; applying a luminance and chrominance separation algorithm to each color image in the original image frame sequence to extract its luminance component and complete the grayscale conversion to obtain the corresponding grayscale image frame sequence; constructing a Gaussian kernel function based on preset kernel size and standard deviation parameters to generate a two-dimensional Gaussian filter kernel, performing a two-dimensional convolution operation on each grayscale image in the grayscale image frame sequence, and performing a weighted average of the image pixels to obtain a smoothed image frame sequence; after smoothing, calculating the local pixel value standard deviation for each frame of the smoothed image frame sequence, comparing the local pixel value standard deviation with a preset noise threshold, identifying and marking potential residual noise regions to complete the image noise suppression processing. In some embodiments, an edge detection operator is applied to the smoothed image frame sequence to highlight the edge contours of all objects inside the dripping bucket to obtain an initial edge image; during the initial configuration stage, a background image of the dripping bucket in a dropless state is acquired, and the background image of the dripping bucket is subtracted from the current frame of the sequence image frame to obtain a difference image; threshold segmentation is performed on the difference image to initially identify regions with pixel values ​​greater than a dynamic threshold as a set of potential foreground pixels; the distribution positions of the initially identified set of potential foreground pixels are matched and fused with the contour positions in the initial edge image obtained by the edge detection operator to remove isolated noise regions caused by light changes, and finally the isolated foreground regions and contours of the droplets inside the dripping bucket are determined.

[0024] Optionally, the standard deviation of local pixel values ​​can be calculated using the following formula: in: The standard deviation of pixel values ​​in a local region. This represents the total number of pixels within a local area. Indicates the first term within a local region grayscale value of each pixel. This represents the average grayscale value of pixels within a local area. It can be understood that by comparing the standard deviation of local pixel values ​​with a preset noise threshold, detailed areas and noisy areas in an image can be distinguished. It can also be understood that the luminance and chrominance separation algorithm converts a color image from the RGB color space to the YUV color space and extracts the Y component as the luminance component for grayscale conversion. In some embodiments, the Gaussian kernel function is constructed based on a two-dimensional Gaussian distribution formula; the kernel size and standard deviation parameters are preset by the system and adjusted according to the image resolution. In specific implementations, the edge detection operator can be the Sobel operator, the Canny operator, or other operators suitable for extracting image edges. The dynamic threshold is determined based on statistical calculations of the global pixel value distribution of the difference image. The matching and fusion process is achieved by calculating the spatial overlap between the potential foreground pixel set and the contour positions in the initial edge image; when the overlap is higher than a preset threshold, the region is confirmed as a valid isolated foreground region.

[0025] In one embodiment of the present invention, see [reference] Figure 2 After the droplet event begins, the system continuously tracks the image pixel area of ​​the isolated foreground region and calculates the difference in image pixel area between adjacent frames. When a continuous increase in image pixel area is detected, it is determined that the droplet is in the growth and falling stage. When a continuous decreasing trend in image pixel area is detected after reaching its maximum value, an area decay frame sequence is recorded. This sequence contains multiple frames of image data that continuously decrease from the maximum area. The rate of decrease in image pixel area in the area decay frame sequence is analyzed. If the rate of decrease is consistently higher than a preset detachment rate threshold, it is determined that the droplet is in the rapid detachment process. When the image pixel area of ​​the isolated foreground region decreases to below a preset minimum foreground area threshold, it is determined that the droplet has completely detached from the upper port of the dropper, thus completing the detachment process, and the droplet event is marked as ending.

[0026] In specific implementation, the process of determining whether a droplet has completed its detachment from the upper port of the dripping container based on the rate of change of image pixel area in consecutive preset frames before the isolated foreground region disappears includes the following operations: After the dripping event begins, the system continuously tracks the image pixel area of ​​the isolated foreground region and calculates the difference in image pixel area between adjacent frames. When a continuous increase in image pixel area is detected, it is determined that the droplet is in the growth and falling stage. When a continuous decreasing trend is detected after the image pixel area reaches its maximum value, an area decay frame sequence is started to be recorded. The area decay frame sequence contains multiple frames of image data that continuously decrease from the maximum area. The rate of decrease of image pixel area in the area decay frame sequence is analyzed. If the rate of decrease is continuously higher than a preset detachment rate threshold, it is determined that the droplet is in the rapid detachment process. When the image pixel area of ​​the isolated foreground region decreases to below a preset minimum foreground area threshold, it is determined that the droplet has completely detached from the upper port of the dripping container, and the dripping event is marked as ended.

[0027] In some embodiments, the change process of image pixel area in consecutive frames can be illustrated by a specific frame sequence example: Assume that in a dripping event, the recorded image pixel area values ​​of the isolated foreground region in five consecutive frames are 150 pixels, 320 pixels, 500 pixels, 480 pixels, and 120 pixels, respectively. Calculating the difference between adjacent frames yields a change sequence of +170 pixels, +180 pixels, -20 pixels, and -360 pixels. From this sequence, it can be observed that the image pixel area increases continuously in the first three frames, reaching a maximum of 500 pixels in the third frame, and then begins to decrease after the third frame. The area decay frame sequence data starting from the third frame are 500 pixels, 480 pixels, and 120 pixels. The system analyzes the decrease rate of the area decay frame sequence and calculates the area change rate between the last two frames. Optionally, the formula for calculating the area change rate is: in: This represents the rate at which the image pixel area decreases from frame t-1 to frame t, in pixels per second. This represents the image pixel area of ​​the isolated foreground region in frame t-1; This represents the image pixel area of ​​the isolated foreground region in frame t; and These represent the timestamps of frame t and frame (t-1), respectively. It can be understood that if the timestamps are calculated for multiple consecutive frames... If all values ​​are higher than the detachment rate threshold, the condition for rapid detachment is met. In practice, the preset minimum foreground area threshold is a fixed pixel value set based on the image resolution and the size of the droplet area. When the image pixel area is lower than this value, the droplet body is considered to have detached from the field of view. It can be understood that the detachment rate threshold is pre-calibrated based on the typical speed of the droplet's physical detachment process combined with the video frame rate.

[0028] In some embodiments, the area decay frame sequence is defined starting from the frame where the image pixel area reaches a local maximum and begins a continuous decreasing trend. A continuous decreasing trend means that the image pixel area decreases for at least three consecutive frames. In a specific implementation, the system continuously calculates and caches the image pixel area values ​​within a sliding window to detect the maximum value and the starting point of the decreasing trend. Optionally, during the droplet growth and falling phases, even if the overall image pixel area shows an increasing trend, there may be slight fluctuations in a single frame. The system avoids misjudgment by setting an area fluctuation tolerance; a decreasing trend is only considered to have started when the decrease in image pixel area exceeds the tolerance for two consecutive frames. In a specific implementation, after determining that the droplet is in a rapid detachment process, the system continues to monitor until the image pixel area is below the minimum foreground area threshold. During this process, even if the rate of decrease in a certain frame is temporarily lower than the detachment rate threshold, the determination process continues as long as the image pixel area continues to decrease and does not recover.

[0029] In one embodiment of the invention, a time window length for calculating the frequency is set, and this time window is slid along the time axis. Within each time window, all marked drop events are identified, and the actual duration corresponding to each drop event is obtained. The actual durations of all single drops within the time window are summed to obtain the total time consumed by the droplet's fall within the time window. The total time consumed by the droplet's fall is subtracted from the time window length, and then divided by the total number of drop events recorded within the time window to calculate the average drop interval within the time window. Sixty seconds is divided by the average drop interval to calculate the measured value of the liquid drop frequency corresponding to the time window.

[0030] In practice, the actual duration of all single drops within a set time period is accumulated, and the measured value of the liquid drop frequency within that time period is calculated. This includes the following steps: First, the length of the time window used for frequency calculation is set, and the time window is slid across the time axis. Within each time window, all marked drop events are identified, and the actual duration corresponding to each drop event is obtained. The actual durations of all single drops within the time window are summed to obtain the total time consumed by the droplets falling within that time window. The total time consumed by the droplets is subtracted from the time window length, and then divided by the total number of drop events recorded within the time window to calculate the average drop interval within that time window. Finally, sixty seconds is divided by the average drop interval to calculate the measured value of the liquid drop frequency corresponding to the time window.

[0031] In some embodiments, the time window length is set to 10 seconds and slides on the time axis in 1-second increments. Optionally, three complete dripping events are recorded within a specific time window. After identifying the dripping events, the actual duration corresponding to each dripping event is obtained, as shown in Table 1.

[0032] Table 1: Example Table of Drop Event Duration within Time Window It can be understood that, by summing the actual durations of all individual drops within the time window, the total time consumed by the droplet's descent is 0.3 seconds + 0.6 seconds + 0.6 seconds = 1.5 seconds. In practical implementation, the average drop interval within the time window is calculated using the following formula: in: This represents the calculated average drop interval, in seconds. This indicates the preset time window length, in seconds. This represents the total time taken for the droplet to fall within the time window, expressed in seconds. This represents the total number of dripping events recorded within the time window. It can be understood that substituting the example data into the formula gives the average dripping interval. The measured value of the liquid dripping frequency is calculated by dividing sixty seconds by the average dripping interval, i.e., 60 / 2.833≈21.2 drops / minute.

[0033] In some embodiments, the sliding process of the time window is continuous. In a specific implementation, after the time window slides forward by 1 second, the new time window may include newly ended dripping events, while removing older dripping events that are no longer within the window's time range. Optionally, the system timestamps each identified dripping event; the sliding of the time window and event identification are independent and parallel processes. In a specific implementation, the choice of time window length is related to the expected sensitivity of liquid dripping frequency monitoring. A shorter window can respond more quickly to changes in drip rate, while a longer window can smooth out instantaneous fluctuations to obtain a more stable frequency value. The measured value of the liquid dripping frequency is dynamically updated as the sliding window moves, forming a continuous frequency monitoring curve.

[0034] See Figure 3This study presents the dynamic evolution of pixel area in an isolated foreground region during a single drop event, providing core data support for determining the droplet growth stage, identifying the detachment process, and timing the drop event. Specifically, the curve fully covers the entire cycle of a droplet from generation, growth, maximum filling, to breakage and detachment: Droplet growth stage (0s~0.5s interval): The pixel area starts from 0 and rises continuously and rapidly, corresponding to the generation and continuous growth process of the droplet at the upper end of the droplet container. The system determines that the droplet is in the growth and falling stage by monitoring the continuous increasing trend of the pixel area, and marks the start time of the drop event. Droplet maximum filling stage (0.4s~0.6s interval): The pixel area reaches its peak (approximately 510 pixels) and remains stable for a short period of time, corresponding to the critical state where the droplet has completed growth and is about to break off. The system starts recording the area decay frame sequence from this peak. The droplet detachment stage (0.6s~1.1s): The pixel area rapidly and continuously decays from its peak, corresponding to the process of the droplet breaking off under gravity, completely detaching from the upper end of the drip chamber, and falling. The system analyzes the rate of pixel area reduction during this stage. When the rate of reduction consistently exceeds a preset detachment rate threshold, the system determines that the droplet is in a rapid detachment process; when the pixel area decays to below a preset minimum foreground area threshold (approximately 10 pixels), the system determines that the droplet has completely detached, marks the end of the drop event, and calculates the actual duration of a single drop. The post-detachment stage (after 1.1s): The pixel area drops back to 0 and remains stable, corresponding to the droplet completely leaving the drip chamber's monitoring field of view. The system waits for the next isolated foreground area to appear to trigger the monitoring of the next drop event. The morphological characteristics of the curve directly serve the core patented algorithm: by classifying the different physical states of the droplet through the trend of pixel area changes, the area decay rate is used as the core basis for detachment determination, while providing a precise time interval for calculating the drop frequency. This is a key visual data carrier for achieving real-time monitoring and abnormal early warning of infusion rates.

[0035] In one embodiment of the invention, a thin-film temperature sensor wrapped around the outside of an infusion tubing acquires the real-time temperature of the liquid and records it as a tubing temperature measurement. The thin-film temperature sensor is tightly wrapped around the outer wall of the infusion tubing, with its sensing surface completely in contact with the tubing's outer wall. The thin-film temperature sensor acquires the temperature of the outer wall of the infusion tubing it contacts at a fixed sampling period, using this as the raw sample value. Multiple continuously acquired raw sample values ​​are filtered using a moving average to eliminate temperature jumps caused by ambient airflow or instantaneous contact changes. The filtered temperature value is then combined with the sensor's factory calibration curve to calculate a tubing temperature measurement reflecting the liquid temperature inside the infusion tubing, which is then output. Simultaneously, the thin-film temperature sensor continuously monitors its thermal resistance in contact with the tubing; if an abnormally high thermal resistance is detected, the temperature value for that sampling period is marked as potentially invalid.

[0036] The system reads the preset target liquid temperature value from the infusion parameter settings. At the start of the isothermal control cycle, it compares the measured tubing temperature with the target liquid temperature value to calculate the current temperature deviation. This temperature deviation is then input into a preset temperature increment control lookup table, which defines the control signal duty cycle adjustment range for different temperature deviation ranges. Based on the retrieved control signal duty cycle adjustment range, a drive command is generated for the semiconductor heating and cooling element wrapped around the infusion tubing. This drive command adjusts the power output duty cycle of the semiconductor heating and cooling element during the isothermal control cycle.

[0037] At the end of the isothermal control cycle, the updated pipeline temperature measurement value collected by the thin-film temperature sensor is read again. The change between the updated pipeline temperature measurement value and the pipeline temperature measurement value of the previous cycle is calculated as the actual temperature change. The theoretical expected temperature change value corresponding to the duty cycle of the control signal used in the previous cycle is read from the temperature increment control lookup table. The difference between the actual temperature change value and the theoretical expected temperature change value is compared. If the absolute value of the difference exceeds the allowable threshold, the current heating / cooling power transfer efficiency is marked as abnormal. When a preset number of consecutive isothermal control cycles are marked as abnormal heating / cooling power transfer efficiency, an inspection prompt is initiated to check the contact status of the semiconductor heating / cooling element and the infusion tubing, as well as the attachment status of the thin-film temperature sensor.

[0038] In practice, the real-time temperature of the liquid is acquired by a thin-film temperature sensor wrapped around the outside of the infusion tubing and recorded as the tubing temperature measurement value. The thin-film temperature sensor acquires the temperature of the outer wall of the infusion tubing it contacts at a fixed sampling period as the raw sample value. Optionally, the fixed sampling period is set to 2 seconds, and an exemplary sequence of raw sample values ​​is 23.1°C, 23.3°C, 22.8°C, 23.0°C, and 23.2°C over 5 consecutive sampling periods. A moving average filter is applied to the multiple continuously acquired raw sample values. A moving average with a window size of 5 is used to filter the above sequence, eliminating temperature jumps caused by ambient airflow or instantaneous contact changes, resulting in a filtered temperature value of 23.08°C. The filtered temperature value is then combined with the sensor's factory calibration curve to calculate the tubing temperature measurement value reflecting the temperature of the liquid inside the infusion tubing, which is then output. In some embodiments, the factory calibration curve provides a mapping relationship between the outer wall temperature of the tubing and the internal liquid temperature. For example, a filtered 23.08°C will be mapped to a tubing temperature measurement of 24.0°C after calibration. Simultaneously, the thin-film temperature sensor continuously monitors the thermal resistance between itself and the tubing. When an abnormal increase in thermal resistance exceeding a set threshold is detected, the temperature value for the current sampling period is marked as potentially invalid. In specific implementations, the thin-film temperature sensor is tightly wound around the outer wall of the infusion tubing, with its sensing surface completely in contact with the tubing wall to ensure efficient heat transfer. The preset target liquid temperature value in the infusion parameter settings is read, for example, a target liquid temperature value set to 25.0°C. At the start of the isothermal control cycle, the tubing temperature measurement value is compared with the target liquid temperature value, and the current temperature deviation is calculated, for example, the current deviation is 25.0°C - 24.0°C = 1.0°C. It can be understood that the temperature deviation is the basic input for subsequent control. The temperature deviation is input into a preset temperature increment control lookup table, which defines the duty cycle adjustment range of the control signal corresponding to different temperature deviation ranges. See Table 2.

[0039] Table 2: Temperature Increment Control Lookup Table Based on the duty cycle adjustment range of the control signal obtained from the query, a drive command is generated for the semiconductor heating and cooling chip wrapped around the infusion tubing. This drive command adjusts the power output duty cycle of the semiconductor heating and cooling chip during the isothermal control cycle. In a specific implementation, if the current temperature deviation is +1.0°C, and the range found in the table above falls within "+0.5 < ΔT ≤ +2.0", the corresponding control signal duty cycle adjustment range ΔD is -5 percentage points. Assuming the power output duty cycle of the semiconductor heating and cooling chip in the previous control cycle was 40%, the new drive command will adjust the power output duty cycle to 35%. In some embodiments, the formula for calculating the power output duty cycle is: in: This indicates the new power output duty cycle of the semiconductor heating and cooling chip during the current constant temperature control cycle. This indicates the power output duty cycle of the semiconductor heating and cooling chip in the previous constant temperature control cycle. This indicates the duty cycle adjustment range of the control signal obtained from the temperature increment control lookup table.

[0040] At the end of the isothermal control cycle, the updated pipeline temperature measurement value collected by the thin-film temperature sensor is read again. The change between the updated pipeline temperature measurement value and the pipeline temperature measurement value of the previous cycle is calculated as the actual temperature change. The theoretical expected temperature change value corresponding to the duty cycle of the control signal used in the previous cycle is read from the temperature increment control lookup table. The difference between the actual temperature change and the theoretical expected temperature change value is compared. If the absolute value of the difference exceeds the allowable threshold, the current heating / cooling power transfer efficiency is marked as abnormal. When a preset number of consecutive isothermal control cycles are marked as abnormal heating / cooling power transfer efficiency, an inspection prompt is initiated on the contact status of the semiconductor heating / cooling element and the infusion tubing, as well as the attachment status of the thin-film temperature sensor. Optionally, the allowable threshold is set to 0.3°C, and the preset number of consecutive cycles is 3 cycles. In one example, the control command in the previous cycle was to increase the duty cycle by 5 percentage points, which corresponds to a theoretical expected temperature change of 0.5°C. However, the actual monitored updated pipeline temperature measurement only increased by 0.1°C. The difference between the actual temperature change and the theoretical expected temperature change is -0.4°C. Since the absolute value of 0.4°C is greater than the allowable threshold of 0.3°C, the heating and cooling power transfer efficiency in this cycle is marked as abnormal.

[0041] See Figure 4This demonstrates the closed-loop tracking performance of the calibrated pipeline temperature towards a target temperature of 25.0℃. Specifically, the solid line represents the actual pipeline temperature after being acquired by a thin-film temperature sensor, filtered by moving average, and converted from the factory calibration curve. The dashed line represents the preset target liquid temperature of 25.0℃, and the filled area is a visual representation of the temperature deviation. Analysis of the curve's timing characteristics reveals the following: Heating phase (0-60 seconds): The initial pipeline temperature is approximately 14.7℃. Under the PID incremental duty cycle control of the semiconductor heating and cooling chip, the temperature rises rapidly and gradually approaches the target value, demonstrating the system's rapid response characteristics. Constant temperature tracking phase (60-115 seconds): The pipeline temperature stabilizes near the target value of 25.0℃, with the maximum steady-state deviation controlled within ±0.5℃. This conforms to the control logic of not adjusting the duty cycle within a ±0.5℃ deviation range in the temperature incremental control lookup table, verifying the steady-state accuracy of the closed-loop control system. Cooling phase (115-120 seconds): The system performs a shutdown operation, and the pipeline temperature drops rapidly, completing a full constant temperature control cycle. The curves simultaneously validated two core functions of the system: first, the calibration accuracy of the thin-film temperature sensor, ensuring that the measured temperature values ​​in the tubing accurately reflect the temperature of the fluid inside the infusion tubing; and second, the effectiveness of the duty cycle incremental control strategy based on temperature deviation, which achieves precise tracking of the target temperature through table-based duty cycle adjustment, meeting the clinical needs for constant temperature control of infusion fluids. Furthermore, the smoothness of the curves also verified the effectiveness of moving average filtering in suppressing environmental noise and transient contact interference, ensuring the stability of temperature control.

[0042] In one embodiment of the invention, the specific type of the triggered abnormal infusion rate warning signal is identified, including excessively fast drip warning and excessively slow drip warning. When an excessively slow drip warning is identified, an adjustment command is automatically generated for the peristaltic pump or infusion clamp in the infusion line. The adjustment command includes a preset small opening increase to attempt to increase the fluid flow rate. After executing the adjustment command, a targeted monitoring process is initiated to re-acquire the fluid dripping video stream within a new set time period and calculate the updated measured value of the fluid dripping frequency. It is determined whether the updated measured value of the fluid dripping frequency has returned to the safe infusion frequency range. If it has not returned, the steps of generating a new small opening increase and adjusting are repeated until the preset maximum number of adjustments is reached or the flow rate returns to normal. When an excessively fast drip warning is identified, an instruction to completely shut down the peristaltic pump or infusion clamp in the infusion line is immediately generated, and an emergency alarm requiring manual confirmation is triggered.

[0043] Record the measured value of the liquid dripping frequency when the excessive dripping rate warning is triggered, as well as the historical measured value sequence of the liquid dripping frequency within a preset time period before the trigger. After manual confirmation and handling of the emergency alarm, initiate subsequent warning event review analysis. Retrieve the sequence image frames corresponding to the triggering of the warning, focusing on analyzing the accuracy of the determination process for the end of the dripping event and verifying the correctness of the calculation of the actual duration of a single drip. Simultaneously, check the historical control command records of the peristaltic pump or infusion clamp in the infusion line before the warning is triggered to confirm whether there are any abnormal or unexpected opening increase commands. Generate a warning event analysis report, which includes dripping rate data, image analysis summary, control command history, and preliminary judgment conclusions on the effectiveness of the warning, for system maintenance and parameter calibration reference.

[0044] In practical implementation, after triggering the abnormal infusion rate warning signal, the system also includes infusion device linkage control based on the warning. It identifies the specific type of the triggered abnormal infusion rate warning signal, including excessively fast drip warnings and excessively slow drip warnings. When an excessively slow drip warning is identified, an adjustment command is automatically generated for the peristaltic pump or infusion clamp in the infusion line. The adjustment command includes a preset small opening increase to attempt to increase the fluid flow rate. In some embodiments, the preset small opening increase can be an increase of 5 rpm in the peristaltic pump speed or a 2% increase in the opening percentage of the infusion clamp. After executing the adjustment command, a targeted monitoring process is initiated. Within a new set time period, the fluid dripping video stream is re-acquired, and the updated measured value of the fluid dripping frequency is calculated. It is determined whether the updated measured value of the fluid dripping frequency has returned to the safe infusion frequency range. If it has not, the steps of generating a new small opening increase and adjusting are repeated until the preset maximum number of adjustments is reached or the flow rate returns to normal. In practical implementation, the preset maximum number of adjustments can be 5 times, and the safe infusion frequency range is preset to 20 to 40 drops / minute. Optionally, in an example scenario, when the initial slow drip warning is triggered, the measured drip rate is 15 drops / minute. The system first generates and executes an adjustment command to increase the infusion opening by 2%. Within the new monitoring time window, the updated measured drip rate is calculated to be 18 drops / minute. Since 18 drops / minute is still below the lower limit of the safe infusion frequency range of 20 drops / minute, the system generates a second adjustment command to increase the infusion opening by 2%. After this second adjustment, the measured frequency is 22 drops / minute, which is now back to the safe range, and the adjustment cycle stops. The formula for determining whether the updated measured drip rate has returned to the safe range can be understood as: in: This represents the measured value of the updated liquid droplet frequency. This indicates the lower limit of the safe infusion frequency range. This indicates the upper limit of the safe infusion frequency range. When an excessively fast drip warning is detected, an instruction is immediately generated to completely shut down the peristaltic pump or infusion clamp in the infusion line, and an emergency alarm requiring manual confirmation is triggered.

[0045] Record the measured value of the liquid dripping frequency when the excessive dripping rate warning is triggered, as well as the historical measured value sequence of the liquid dripping frequency within a preset time period before the trigger. After manual confirmation and handling of the emergency alarm, initiate subsequent warning event review analysis. In some embodiments, the preset time period can be 60 seconds before the trigger. The system records the measured value of the liquid dripping frequency calculated once per second, forming a historical measured value sequence containing 60 data points. For example, the sequence data gradually increases from the normal 30 drops / minute to 45 drops / minute, 50 drops / minute, and finally reaches 65 drops / minute at the trigger time. Call up the sequence image frame corresponding to the triggering of the warning, focusing on analyzing whether the determination process of the end of the dripping event is accurate and verifying whether the actual duration of a single drip is calculated correctly. At the same time, check the historical control command history of the peristaltic pump or infusion clamp in the infusion line before the warning is triggered to confirm whether there are any abnormal or unexpected opening increase commands. In specific implementations, the historical control command history is stored in the form of timestamps and command content. During the review, all commands within a period of time before the warning is triggered are retrieved. Optionally, if a series of unexpected and significant opening increases are found in the historical records shortly before the warning is triggered, this may indicate equipment malfunction or control logic failure. A warning event analysis report is generated, containing drip rate data, an image analysis summary, control command history, and a preliminary judgment on the effectiveness of the warning, for system maintenance and parameter calibration reference. The warning event analysis report is stored in a structured document format. The drip rate data section includes the measured value and historical sequence at the time of triggering; the image analysis summary describes the inspection results of keyframes; the control command history lists relevant operation records; and the preliminary judgment is based on a comprehensive analysis of the aforementioned information, such as "The warning is effective; the abnormal increase in drip rate is suspected to be caused by external interference."

[0046] See Figure 5The diagram presents the closed-loop control logic of the system based on visual drip rate detection and actuator linkage. The horizontal axis represents the number of adjustments, and the vertical axis represents the corresponding monitored values. It includes two core monitoring curves and two safety threshold lines: the solid dotted line represents the measured drip rate (drops / minute), the solid box line represents the infusion clamp opening (%), the lower dashed line represents the lower limit of the safe drip rate (20 drops / minute), and the upper dashed line represents the upper limit of the safe drip rate (40 drops / minute). The specific control process is as follows: Initial abnormality trigger: When the number of adjustments is 0, the measured drip rate is 15 drops / minute, which is lower than the safety lower limit of 20 drops / minute. The system triggers a slow drip rate warning, at which point the initial opening of the infusion clamp is 10%. Closed-loop adjustment phase: The system adjusts according to preset small opening increments (+2% each time): After the first adjustment, the infusion clamp opening increases to 12%, and the measured drip rate recovers to 18 drops / minute, still not returning to the safe range; after the second adjustment, the infusion clamp opening increases to 14%, and the measured drip rate recovers to 22 drops / minute, successfully entering the safe range of 20-40 drops / minute, and the adjustment logic meets the stop condition. Anomaly protection and locking: After the second adjustment, the system detects an abnormal drop in drip rate (when adjusting 3-5 times, the measured drip rate drops to 0 drops / minute). At this time, the infusion clamp opening has been locked at a stable value of 14%, the system triggers a second anomaly warning, stops further adjustment, and waits for manual intervention, verifying the system's safety protection mechanism under extreme conditions. The monitoring chart visually verifies the core logic of automatic adjustment when the drip rate is too slow in the patent: the actual drip rate is obtained in real time through the visual drip rate detection module, and after being compared with the safety threshold, the opening of the infusion clamp is adjusted incrementally to form a closed-loop control of "detection-adjustment-re-detection". At the same time, protection is triggered in extreme scenarios such as abnormal drip rate returning to zero, so as to ensure the safety and stability of the infusion process.

[0047] In the actual operation of the infusion safety control method, a three-color indicator light group (red, yellow, and green) can be integrated into the casing of the infusion monitoring terminal. Simultaneously, a non-contact water level sensor patch made of flexible printed circuitry is attached to the outer bottom of the medication bag. The water level sensor patch continuously senses the liquid level at the bottom of the medication bag using capacitive detection and synchronously sends the real-time liquid level signal to the control unit. The control unit compares the received liquid level signal with three preset threshold ranges. When the liquid level is sufficient, the green indicator light remains constantly lit, indicating that the current remaining medication volume is normal. When the liquid level drops to the preset remaining medication threshold, for example, about 50 ml remaining, the control unit switches to a constantly lit yellow indicator light to remind nursing staff to monitor the medication progress. When the liquid level approaches the empty bag warning line, for example, about 10 ml remaining, the control unit immediately switches to a high-frequency flashing red indicator light and simultaneously calls the built-in voice broadcast module to play a prompt such as "The patient's medication is about to run out; please replace it promptly." The loudness of the prompt can be automatically adjusted according to the ambient volume of the ward.

[0048] The light and voice prompt function is linked to the abnormal infusion rate warning signal described in the claims. When the system detects an excessively fast drip rate warning and generates a command to completely shut down the peristaltic pump, in addition to triggering an emergency alarm requiring manual confirmation, it can also simultaneously illuminate a red indicator light and play an emergency alarm voice, prompting nursing staff to immediately handle the situation. When the system detects an excessively slow drip rate warning and automatically generates an adjustment command to slightly increase the opening, if the flow rate returns to the normal range after adjustment, the green indicator light remains constantly on. If the flow rate still cannot return to the normal range after adjustment, the yellow indicator light flashes continuously accompanied by intermittent voice prompts, reminding nursing staff to check whether the tubing is blocked or the needle has shifted. During routine rounds, nursing staff do not need to approach and check the specific drip rate values ​​or remaining medication percentages on the screen one by one. They can quickly judge the current infusion status simply by observing the light color from a distance. Green indicates stable operation, yellow indicates that attention is needed or there is a slight abnormality, and red indicates an emergency requiring immediate attention. Combined with multiple reminders from voice prompts, this effectively reduces the workload of rounds, avoids missed or misjudgments due to distraction, and makes the feedback from infusion monitoring more intuitive and efficient. The water level sensing patch is made of medical-grade flexible material. When applied, it does not require alteration of the original drug bag structure and will not come into direct contact with the drug. The sensing signal is transmitted to the control unit wirelessly, without interfering with the video stream acquisition and image processing process described in the claims. The power supply and control of the three-color indicator light and the voice module are uniformly scheduled by the infusion monitoring terminal. It shares the same control logic with the original constant temperature control and drip rate warning functions, together forming a complete infusion safety monitoring system.

[0049] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for infusion safety control with drip warning and constant temperature regulation, characterized in that, The method includes: Acquire a continuous video stream of liquid dripping at the dripping location of the infusion tubing, and convert the liquid dripping video stream into a sequence of image frames; Contour enhancement and background separation processing are performed on the sequence of image frames to extract the isolated foreground region of the droplet inside the droplet in each frame image, and the image pixel area of ​​the isolated foreground region is calculated. When a newly emerging isolated foreground region is detected, a dripping event is determined to have started, and the process of recording the change of the image pixel area of ​​the isolated foreground region corresponding to the dripping event in consecutive frames is started until the isolated foreground region disappears. Based on the rate of change of image pixel area in consecutive preset frames before the disappearance of the isolated foreground region, it is determined whether the droplet has completed the process of detaching from the upper port of the drip chamber. When it is determined that the detachment has been completed, a dripping event is marked as ended. Record the time interval between the start and end of the dripping event as the actual duration of a single drip, and accumulate the actual duration of all single drips within a set time period to calculate the measured value of the liquid dripping frequency within the set time period. The measured value of the liquid dripping frequency is compared with the preset safe infusion frequency range. When the measured value continuously exceeds the safe infusion frequency range, an abnormal infusion rate warning signal is triggered.

2. The infusion safety control method with drip warning and constant temperature regulation according to claim 1, characterized in that, The sequence of image frames is subjected to contour enhancement and background separation processing to extract the isolated foreground region of the droplet inside the droplet in each frame, including: The input sequence of image frames is converted to grayscale, and a Gaussian filtering algorithm is used to smooth the grayscale image and suppress image noise. An edge detection operator is applied to the smoothed image to highlight the edge contours of all objects inside the drip chamber, thus obtaining an initial edge image. In the initial configuration phase, a background image of the dripping chamber in a state without dripping is acquired, and the background image of the dripping chamber is subtracted from the current frame of the sequence image frame to obtain a difference image; The differential image is segmented by thresholding, and regions with pixel values ​​greater than the dynamic threshold are initially identified as a set of potential foreground pixels; The distribution locations of the initially identified potential foreground pixels are matched and fused with the contour locations in the initial edge image obtained by the edge detection operator. Isolated noise areas caused by changes in light are eliminated, and finally, the isolated foreground regions and their contours of the droplets inside the droplet are determined.

3. The infusion safety control method with drip early warning and constant temperature regulation according to claim 2, characterized in that, Based on the rate of change of image pixel area in consecutive preset frames before the disappearance of the isolated foreground region, determining whether the droplet has completed the process of detaching from the upper port of the dropper includes: After the dripping event begins, the image pixel area of ​​the isolated foreground region is continuously tracked, and the difference in the image pixel area between adjacent frames is calculated; When the area of ​​the image pixels is continuously increased, it is determined that the droplet is in the growth and falling stage; When it is detected that the area of ​​the image pixels begins to decrease continuously after reaching the maximum value, the area decay frame sequence is recorded. The area decay frame sequence contains multiple frames of image data that decrease continuously from the maximum area. Analyze the rate of decrease of the image pixel area in the area decay frame sequence. If the rate of decrease is continuously higher than the preset detachment rate threshold, it is determined that the droplet is in the process of rapid tearing and detachment. When the image pixel area of ​​the isolated foreground region decreases to below the preset minimum foreground area threshold, it is determined that the droplet has completely detached from the upper port of the dropper, that is, the detachment process is completed, and the drop event is marked as ended.

4. The infusion safety control method with drip warning and constant temperature regulation according to claim 3, characterized in that, The actual duration of all single drops within the cumulative set time period is used to calculate the measured value of the liquid drop frequency within the set time period, including: Set the length of the time window used to calculate the frequency, and slide the time window on the time axis; Within each time window, identify all marked dripping events and obtain the actual duration corresponding to each dripping event; The total time consumed by the droplet falling within the time window is obtained by summing the actual duration of all single drops within the time window. The average drop interval within the time window is calculated by subtracting the total time consumed by the droplet's fall from the length of the time window and then dividing by the total number of drop events recorded within the time window. Divide sixty seconds by the average drip interval to obtain the measured value of the liquid drip frequency corresponding to the time window.

5. The infusion safety control method with drip early warning and constant temperature regulation according to claim 4, characterized in that, After obtaining the measured value of the liquid dripping frequency within the set time period through calculation, the method further includes: The real-time temperature of the liquid is acquired by a thin-film temperature sensor wrapped around the outside of the infusion tubing and recorded as the tubing temperature measurement value. Read the preset target temperature value of the infusion fluid in the infusion parameter settings; At the beginning of the constant temperature control cycle, the measured value of the pipeline temperature is compared with the target temperature value of the liquid, and the temperature deviation at the current moment is calculated. The temperature deviation is input into a preset temperature increment control lookup table, which defines the control signal duty cycle adjustment range corresponding to different temperature deviation ranges. Based on the duty cycle adjustment range of the control signal obtained from the query, a drive command is generated for the semiconductor heating and cooling chip wrapped around the infusion tubing. The drive command is used to adjust the power output duty cycle of the semiconductor heating and cooling chip during the constant temperature control cycle.

6. The infusion safety control method with drip early warning and constant temperature regulation according to claim 5, characterized in that, The process by which the thin-film temperature sensor acquires the real-time temperature of the liquid includes: The thin-film temperature sensor is tightly wound around the outer wall of the infusion tubing, with its sensing surface completely in contact with the outer wall of the tubing. The thin-film temperature sensor collects the temperature of the outer wall of the infusion tubing it contacts at a fixed sampling period as the original sample value; A moving average filter is applied to multiple continuously collected raw sample values ​​to remove temperature jumps caused by environmental airflow or instantaneous contact changes. The filtered temperature value is combined with the sensor's factory calibration curve to calculate the pipeline temperature measurement value reflecting the liquid temperature inside the infusion tubing, and then output to the outside. Meanwhile, the thin-film temperature sensor continuously monitors the thermal resistance between itself and the hose. When an abnormal increase in thermal resistance is detected, the temperature value of the sampling period may be invalid.

7. The infusion safety control method with drip early warning and constant temperature regulation according to claim 6, characterized in that, After generating the drive command for the semiconductor heating / cooling chip wrapped around the infusion tubing, the method further includes: At the end of the constant temperature control cycle, the updated pipeline temperature measurement value collected by the thin-film temperature sensor is read again. Calculate the change between the updated pipeline temperature measurement value and the pipeline temperature measurement value of the previous cycle, and use it as the actual temperature change. Read the theoretical expected temperature change value corresponding to the duty cycle of the control signal used in the previous cycle from the temperature increment control lookup table; Compare the difference between the actual temperature change and the theoretical expected temperature change. If the absolute value of the difference exceeds the allowable threshold, mark the current heating and cooling power transfer efficiency as abnormal. When a preset number of consecutive constant temperature control cycles are detected as abnormal heating and cooling power transfer efficiency, an inspection prompt will be initiated to check the contact status between the semiconductor heating and cooling chip and the infusion tubing, as well as the attachment status of the thin-film temperature sensor.

8. The infusion safety control method with drip early warning and constant temperature regulation according to claim 7, characterized in that, After triggering the abnormal infusion rate warning signal, the system also includes infusion device linkage control based on the warning: Identify the specific type of the abnormal infusion rate warning signal that was triggered, including warnings of excessively fast drip rate and warnings of excessively slow drip rate; When a warning of slow drip rate is detected, an adjustment command is automatically generated for the peristaltic pump or infusion clamp in the infusion line. The adjustment command includes a preset small opening increase to attempt to increase the fluid flow rate. After executing the adjustment command, a targeted monitoring process is initiated to re-acquire the liquid dripping video stream within the new set time period and calculate the updated measured value of the liquid dripping frequency. Determine whether the measured value of the updated liquid dripping frequency has returned to the safe infusion frequency range. If it has not returned, repeat the steps of generating a new small opening increase and adjusting until the preset maximum number of adjustments is reached or the flow rate returns to normal. When an alarm for excessively fast drip rate is detected, an instruction is immediately generated to completely shut down the peristaltic pump or infusion clamp in the infusion line, and an emergency alarm requiring manual confirmation is triggered.

9. The infusion safety control method with drip early warning and constant temperature regulation according to claim 8, characterized in that, When a warning of excessively rapid drip rate is detected, an instruction is immediately generated to completely shut down the peristaltic pump or infusion clamp in the infusion line, and an emergency alarm requiring manual confirmation is triggered. This also includes: Record the measured value of the liquid dripping frequency when the excessive dripping rate warning is triggered, as well as the historical measured value sequence of the liquid dripping frequency within a preset time period before the trigger; After the emergency alarm is manually confirmed and handled, a post-event review and analysis will be initiated. The sequence image frames corresponding to the triggered warning are invoked, and the accuracy of the determination process for the end of the dripping event is analyzed, and the accuracy of the calculation of the actual duration of a single drip is verified. At the same time, check the historical records of control commands for the peristaltic pump or infusion clamp in the infusion line before the warning is triggered to confirm whether there are any abnormal or unexpected opening increase commands. Generate an early warning event analysis report, which includes drip rate data, image analysis summary, control command history, and preliminary judgment on the effectiveness of the early warning, for reference in system maintenance and parameter calibration.

10. The infusion safety control method with drip early warning and constant temperature regulation according to claim 9, characterized in that, The input sequence of image frames is converted to grayscale, and a Gaussian filtering algorithm is used to smooth the grayscale image and suppress image noise, including: Receive a continuous video stream of liquid dripping from the dripping location of the infusion tubing, and convert the liquid dripping video stream into a sequence of image frames to obtain the original image frame sequence; Apply a luminance and chrominance separation algorithm to each color image in the original image frame sequence, extract its luminance component, complete the grayscale conversion, and obtain the corresponding grayscale image frame sequence. A Gaussian kernel function is constructed, and a two-dimensional Gaussian filter kernel is generated based on the preset kernel size and standard deviation parameters. A two-dimensional convolution operation is performed on each grayscale image in the grayscale image frame sequence, and the image pixels are weighted and averaged to obtain a smoothed image frame sequence. After smoothing, the local pixel value standard deviation is calculated for each frame of the image frame sequence. Based on the local pixel value standard deviation, it is compared with a preset noise threshold to identify and mark potential residual noise regions, thus completing the image noise suppression process.

Images