Operating room nursing infusion intelligent monitoring method and monitoring system thereof

Abstract

The invention relates to the technical field of infusion nursing, and discloses an intelligent monitoring method and system for nursing infusion in an operating room, and the method comprises the steps: collecting the pressure intensity, flow velocity and liquid level height of an infusion pipeline, and obtaining physiological parameters; in a non-transfusion state, collecting N frames of drip chamber images to establish a coordinate system and a background template; m frames of images are obtained in the infusion process, a binary foreground difference image is constructed, a connected domain is extracted, barycentric coordinates are calculated, and whether the images are the same liquid drop track or not is judged through the barycentric distance; counting the number of liquid drops and calculating the flow velocity by combining the drop ratio; constructing infusion indexes including a time sequence consistency score and a pressure diagnosis factor; risk vectors are constructed through patient parameters and infusion indexes, and normalization processing is carried out to generate dynamic risk mapping; and designing a risk scoring algorithm, setting grades according to risk values, and taking different treatment measures.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of infusion nursing, in particular to a surgical nursing infusion intelligent monitoring method and a monitoring system thereof. BACKGROUND

[0002] In modern surgical procedures, intravenous infusion, as an important means to maintain patient vital signs and deliver drugs and nutrients, its safety and real-time monitoring has become one of the key links of operating room nursing management. Traditional infusion monitoring relies on manual observation, such as nursing staff observing the number of drops, checking the liquid level of the infusion bottle, or regularly patrolling to judge the infusion. This method is not only inefficient and susceptible to subjective errors, but also cannot achieve real-time response to sudden abnormal situations such as pipe blockage, liquid leakage, and patient pathological changes, which poses a great potential risk.

[0003] With the development of image processing technology, sensor technology and intelligent algorithm, it is possible to apply multi-source data fusion to infusion monitoring. Some current research attempts to use image recognition technology to identify infusion dripping conditions or use pressure sensors to detect infusion status, but there are generally the following problems: first, it is difficult to accurately extract the liquid drop motion trajectory in the image recognition process, especially under complex lighting or background interference conditions, which is prone to recognition errors; second, there is a lack of dynamic fusion analysis of multiple parameters such as drop speed, pressure, and liquid level, making it difficult to accurately assess the stability and safety of infusion; third, it fails to combine patient's own physiological parameters for personalized risk assessment, resulting in delayed monitoring response or high false alarm rate.

[0004] Therefore, there is an urgent need for an intelligent monitoring method that can integrate image recognition and physical parameter monitoring, and combine real-time physiological state of the patient, which can not only accurately extract liquid drop features and flow rate changes during infusion, but also dynamically assess infusion risk levels according to pressure changes, liquid level, and patient physiological indicators, thereby providing a more scientific, reliable, and real-time infusion monitoring means for operating room nursing. SUMMARY

[0005] The present application provides a surgical nursing infusion intelligent monitoring method and a monitoring system thereof, which solves the problems mentioned in the background.

[0006] The present application provides the following technical solution: a surgical nursing infusion intelligent monitoring method, comprising:

[0007] Collecting a set of infusion parameters D at time t t , including:

[0008] Using a pressure sensor to collect the pressure of the liquid in the infusion pipeline at time t, denoted as P t ;

[0009] Calculate the liquid flow rate in the infusion pipeline, denoted as F t ;

[0010] Obtain the liquid level of the infusion bottle, denoted as L t ;

[0011] Construct a set of patient physiological state parameters H t , including:

[0012] The patient's heart rate HR, systolic blood pressure SBP, and diastolic blood pressure DBP;

[0013] Collect image data of the drip chamber when infusion is not performed, and construct a coordinate system and background template for the image data;

[0014] Obtain image data during infusion, and construct a binary foreground difference image;

[0015] Extract connected domains from the binary foreground difference image and calculate the barycentric coordinates of each connected domain;

[0016] Distinguish and judge the connected domains according to the barycentric coordinates between adjacent connected domains;

[0017] Count the number of different connected domains and calculate the liquid flow rate in combination with the drop ratio;

[0018] Construct an algorithm and calculate infusion indicators, including timing consistency score and pressure diagnostic factor;

[0019] Construct a risk vector from patient physiological parameters and infusion indicators, perform normalization processing to generate a dynamic risk map;

[0020] Construct a risk assessment algorithm to calculate a risk score, set risk levels, and take corresponding measures for different risk levels.

[0021] Optionally, the collecting of image data of the drip chamber when infusion is not performed, and the construction of a coordinate system and background template for the image data, includes:

[0022] Continuously collect N frames of images of the infusion drip chamber when infusion is not performed, denoted as initial images;

[0023] Take the pixel point located at the midpoint in the initial image as the coordinate origin of the initial image, and draw a vertical upward ray through the coordinate origin as the y-axis;

[0024] Take the ray in the initial image that is perpendicular to the y-axis and horizontally to the right as the x-axis, forming the coordinate system of the initial image;

[0025] Obtain the gray value of each initial image at pixel point (x, y), and construct a gray value sequence S x,y , specifically:

[0026] S x,y ={f1(x,y),f2(x,y),……,f N (x,y)};

[0027] Among them, f N (x,y) represents the gray value at pixel (x,y) in the Nth frame image;

[0028] The grayscale value sequence S x,y Sort the sequences in ascending order to obtain an ordered sequence S′. x,y Specifically:

[0029] S′ x,y =sort(S) x,y )={s1,s2,……,s N};

[0030] Among them, sort(S) x,y ) represents the sorting function, i.e., for S x,y Sort the grayscale values ​​within the range from smallest to largest, s N S′ x,y The gray value of the Nth pixel (x, y) in the image, where s1 ≤ s2 ≤ ... ≤ s N ;

[0031] The background modeling algorithm is constructed as follows:

[0032]

[0033] Where B(x,y) represents the grayscale value of the constructed background template at pixel (x,y), s k Indicates at S′ x,y The k-th element in.

[0034] Optionally, acquiring image data during the infusion process and constructing a binary foreground difference image includes:

[0035] Once the infusion begins, M frames of images of the infusion drip chamber are continuously acquired and recorded as infusion images;

[0036] The pixel located at the midpoint of the infusion image is taken as the origin of the coordinate system of the infusion image. A vertical ray is drawn through the origin as the vertical axis, i.e., the y-axis.

[0037] The ray perpendicular to the vertical axis and pointing horizontally to the right in the infusion image is used as the horizontal axis, i.e., the x-axis, to form the coordinate system of the infusion image;

[0038] The foreground difference extraction algorithm is constructed as follows:

[0039]

[0040] wherein D s_i (x,y) represents the binary foreground mask image of the s_i-th infusion image, 1 and 0 represent the values of the foreground mask respectively, f s_i (x,y) represents the gray value of the s_i-th infusion image at the pixel point (x,y), δ s_i (x,y) represents the adaptive differential threshold value of the s_i-th infusion image at the pixel point (x,y), and is specifically as follows:

[0041] δ s_i (x,y) = μ s_i + 2σ s_i ;

[0042] wherein μ s_i represents the average gray value of the s_i-th infusion image, σ s_i represents the gray standard deviation of the s_i-th infusion image;

[0043] The average gray value is specifically:

[0044]

[0045] wherein H represents the vertical pixel number of the infusion image, and W represents the horizontal pixel number of the infusion image;

[0046] The gray standard deviation is specifically:

[0047]

[0048] Optionally, the extracting the connected domains and calculating the barycentric coordinates of each connected domain through the binary foreground difference image comprises:

[0049] labeling the connected domains of the binary foreground mask image corresponding to the s_i-th infusion image;

[0050] extracting the pixel set corresponding to each connected domain, denoted as R l_k ;

[0051] wherein R l_k represents the pixel point set of the l_k-th connected domain;

[0052] in the order of acquisition of the infusion image where the connected domain is located, the barycentric coordinates of each connected domain are calculated in turn, and the specific process is as follows:

[0053]

[0054] wherein represents the barycentric coordinates of the l_k-th connected domain in the c-th infusion image, c is less than or equal to M, and |R l_k| represents the total number of pixel points in the l_kth connected domain.

[0055] Optionally, the distinguishing judgment of the key point coordinates between adjacent connected domains includes:

[0056] According to the acquisition order of the infusion image, the distance of the key point coordinates between adjacent connected domains is calculated in turn, specifically:

[0057] If It is determined that the l_k+1th connected domain is the same target continuation track of the l_kth connected domain, that is, the same connected domain, τ d represents the maximum reachable distance threshold;

[0058] Otherwise, it is determined that the l_k+1th connected domain is not the same target continuation track as the l_kth connected domain, that is, different connected domains;

[0059] The above-mentioned maximum reachable distance threshold τ d = v p × Δt is calculated.

[0060] Where Δt represents the inter-frame interval time of the infusion image acquisition, v p represents the pixel speed of the center of gravity coordinates of the same connected domain, specifically:

[0061]

[0062] Where c_n is the pixel point base, and the value range is [5, 10].

[0063] Optionally, the number of different connected domains is counted and combined with the drop ratio to calculate the liquid flow rate, including:

[0064] The number of connected domains is obtained as the number of droplets, denoted as new_k.

[0065] The drop ratio of the current drop model is obtained, denoted as r drip .

[0066] The actual flow rate F t of the liquid at the current time is calculated, specifically as follows:

[0067]

[0068] Where T m represents the time when M frames of infusion images are collected.

[0069] Optionally, the algorithm is constructed and the infusion indicators, including the timing consistency score and the pressure diagnostic factor, are calculated, including:

[0070] The timing consistency recognition algorithm is designed, specifically as follows:

[0071]

[0072] wherein S PCL (t) is the time consistency score at time t, and time_i is in the range of [1, T m -t];

[0073] represents the pressure change rate at t-time_i;

[0074] represents the flow rate change rate at t-time_i;

[0075] represents the liquid level change rate at t-time_i;

[0076] sign(*) is a sign function, and returns 1 when is greater than 0, returns 0 when is equal to 0, and returns -1 when is less than 0;

[0077] A differential pressure tracking diagnosis algorithm is designed to calculate a pressure diagnosis factor D FTD (t), specifically as follows:

[0078]

[0079] wherein P t+1 and F t+1 represent the pressure and the liquid flow rate at t+1, respectively.

[0080] Optionally, the risk vector is constructed by the patient physiological parameters and the infusion indicators, and normalized processing is performed to generate a dynamic risk mapping, including:

[0081] The risk vector is constructed, denoted as Risk t = [S PCL (t), D FTD (t), H t ];

[0082] Each item in the risk vector is defined as r Risk_i (t) ∈ Risk t ;

[0083] The maximum value and the minimum value of each item r Risk_i (t) in the risk vector between time [t, T m ] are obtained, and are denoted as and

[0084] For each item rRisk_i (t) performing dimensionless processing, specifically:

[0085]

[0086] wherein, Risk_i represents the i-th risk indicator in the risk vector after dimensionless processing, and ε is the denominator protection coefficient;

[0087] Let the risk vector after dimensionless processing be

[0088] Optionally, the risk assessment algorithm is constructed to calculate a risk score and set a risk level, and corresponding measures are taken for different risk levels, including:

[0089] A risk score algorithm is designed, specifically:

[0090]

[0091] wherein, Risk represents the comprehensive risk score, Risk represents the maximum value in the vector at time t;

[0092] A risk score level level is set, specifically:

[0093]

[0094] When Level = 1, the risk level is set to normal and continues to be monitored;

[0095] When Level = 2, the risk level is set to medium risk, and the nursing staff is reminded to check;

[0096] When Level = 3, the risk level is set to high risk, and the infusion operation is immediately terminated and an alarm is sounded.

[0097] A monitoring system for implementing the operating room nursing infusion intelligent monitoring method, comprising:

[0098] A data collection module collects a set of infusion parameters in the operating room for subsequent judgment of the risk level of the infusion;

[0099] An image acquisition module acquires images before and during infusion to monitor whether the infusion process is abnormal in real time;

[0100] An image data processing module processes the acquired image data to identify falling droplets;

[0101] A data calculation module calculates the pixel moving speed of the connected domain focus coordinates, and further calculates the flow speed of the liquid according to the movement of the connected domain focus coordinates.

[0102] A risk judgment module calculates the risk comprehensive score of the infusion in real time through the risk vector, and judges the risk level of the infusion.

[0103] The present application has the following advantages:

[0104] 1. When infusion is not performed, N frames of drop chamber images are continuously collected as an initial image sequence, and a two-dimensional image coordinate system is constructed with the image center pixel as the coordinate origin; subsequently, the gray values of the pixel points at the same position in each frame of image are extracted to form a gray sequence, and the sequence is sorted; according to the sorting result, the kth smallest gray value is selected as the background gray value of the corresponding pixel point in the background template, thereby completing the construction of the static background template. By collecting multiple frames of images of the drop chamber region under the non-infusion state and constructing the gray background template, the background noise and environmental interference factors are effectively excluded, and the accuracy and robustness of the subsequent foreground difference image are improved; at the same time, the establishment of the standardized coordinate system provides a unified spatial reference for the liquid drop motion trajectory recognition and the connected domain barycenter positioning, which helps to realize the fusion calculation of image analysis and physical monitoring data, thereby improving the reliability and precision of the whole infusion monitoring system.

[0105] 2. After the infusion starts, M frames of infusion drop chamber images are continuously collected, and a coordinate system is established with the image center pixel; an adaptive difference threshold is constructed through the gray value, the average gray value and the gray standard deviation of each frame of image; the current frame of image is subjected to difference processing by using the threshold, and the corresponding binary foreground mask image is generated. This method can effectively extract the liquid drop motion foreground region, and lay a foundation for the subsequent connected domain recognition and liquid drop trajectory tracking. By introducing the adaptive difference threshold, this method overcomes the problem of low recognition accuracy of the traditional fixed threshold under different illuminations or backgrounds, and can more accurately extract the liquid drop image foreground; at the same time, the foreground extraction based on image statistical features improves the robustness and real-time performance of the liquid drop recognition, and helps to realize the accurate detection and intelligent monitoring of the liquid drop trajectory in the infusion process.

[0106] 3、The surgical nursing infusion intelligent monitoring method provided by the application, through processing the binary foreground difference image, first marks the connected domain in each frame of infusion image, and extracts the corresponding pixel point set. Subsequently, according to the time sequence of image acquisition, the center of gravity coordinates of each connected domain are calculated in turn to obtain the motion trajectory of the liquid drop in the image sequence. The method can accurately extract the key feature points in the liquid drop falling process through pixel-level spatial analysis, and provide basic data support for the subsequent dynamic tracking of the connected domain and drop speed identification. This can significantly improve the accuracy and robustness of drop identification, and even in the case of light changes or background interference, the liquid drop area can be stably extracted and its spatial position can be accurately calculated. At the same time, through the time sequence calculation of the center of gravity coordinates, the continuity judgment basis for the liquid drop trajectory analysis can be provided, the identification error caused by the change of the liquid drop shape in multiple images can be effectively avoided, and high-reliability data support for liquid flow rate monitoring and infusion abnormality early warning can be provided, thereby improving the intelligence and safety of infusion monitoring in the surgical nursing process.

[0107] 4、The application provides a target distinguishing judgment method based on adjacent connected domain center of gravity coordinates, which realizes the continuity identification of the liquid drop motion trajectory by calculating the distance between the center of gravity coordinates of the connected domains in adjacent frames of the infusion image in turn, and combining a preset farthest reachable distance threshold. The threshold is dynamically calculated according to the frame interval time and the pixel speed, so that the judgment basis has time consistency and physical rationality, thereby effectively distinguishing the continuous trajectory of the same liquid drop and the independent trajectory of different liquid drops. The method significantly improves the accuracy and robustness of drop identification by constructing a connected domain tracking mechanism based on coordinate changes, avoiding the misidentification problem caused by drop speed changes or image jitter in traditional image processing. At the same time, the method has good time matching ability and dynamic adaptability, and can be applied to infusion scenes with different drop speeds and sampling frequencies, providing a reliable data basis for subsequent flow rate calculation and infusion risk identification.

[0108] 5、The surgical nursing infusion intelligent monitoring method provided by the application extracts the number of connected domains in the infusion process through image processing means, regards the number of liquid drops, and calculates the actual liquid flow rate per unit time by combining a preset drop number ratio parameter. The process realizes the quantitative expression of the flow rate by acquiring consecutive image frames and counting the number of recognized connected domains, and has strong real-time and accuracy. In this way, the liquid flow rate can be accurately measured without additional hardware devices, overcoming the large drop speed estimation error and the dependence on manual observation in the traditional infusion monitoring method, especially suitable for scenes such as operating rooms that require high nursing safety and monitoring accuracy. At the same time, the method has good adaptability and expandability, and can be flexibly applied to various infusion systems combined with different types of drop parameters, improving the intelligent level and risk prevention and control ability of the infusion process.

[0109] 6、The application constructs key indicators in the infusion process, time sequence consistency score and pressure diagnosis factor, which can dynamically reflect the change trend and mutual relationship of pressure, flow rate and liquid level in the infusion process. Based on this, the patient physiological parameters and infusion indicators are further fused to construct a risk vector, and a dynamic risk mapping is generated through normalization processing, realizing multi-dimensional and real-time quantitative evaluation of infusion risk. The method can accurately capture the small changes of infusion state, combined with physiological data such as patient heart rate and blood pressure, improve the individualization and accuracy of risk assessment; through dimensionless processing, the interference caused by the dimension difference of data is effectively avoided, and the robustness and stability of the algorithm are enhanced; overall, the intelligent level of operating room infusion monitoring is significantly improved, early warning and dynamic management of potential abnormalities are realized, patient fluid safety is ensured, and the work burden of nursing personnel is reduced.

[0110] 7、The application proposes a kind of operating room nursing infusion intelligent monitoring method based on the construction risk assessment algorithm, realizes the scientific division of risk grade in infusion process by calculating the comprehensive score of dynamic risk vector. The method sets up three risk grades of normal, medium risk and high risk, and takes corresponding monitoring and intervention measures for different risk grades, including continuing to monitor, reminding nursing staff to check and urgently terminating infusion and alarming, to ensure infusion safety. The method can evaluate the potential risk in infusion process in real time and accurately, and comprehensively judge combined with multi-dimensional dynamic data, improve the intelligence and accuracy of infusion monitoring, effectively reduce human negligence and missed report, ensure patient medication safety, and improve the overall quality and efficiency of surgical nursing. BRIEF DESCRIPTION OF DRAWINGS

[0111] Figure 1 It is the flowchart of the application.

[0112] Figure 2 It is the system structure schematic diagram of the application. DETAILED DESCRIPTION

[0113] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0114] Embodiment one, refer to Figure 1 A kind of operating room nursing infusion intelligent monitoring method, comprising:

[0115] Collect infusion parameter set D at t time t , including:

[0116] acquire the pressure of the liquid in the infusion pipeline at time t, denoted as P t ;

[0117] calculate the flow rate of the liquid in the infusion pipeline, denoted as F t ;

[0118] acquire the liquid level of the infusion bottle, denoted as L t ;

[0119] construct a set of physiological state parameters of the patient, denoted as H t , including:

[0120] heart rate HR, systolic blood pressure SBP and diastolic blood pressure DBP of the patient;

[0121] acquire image data of the drip chamber when no infusion is performed, and construct a coordinate system and a background template of the image data;

[0122] acquire image data in the infusion process, and construct a binary foreground difference image;

[0123] extract connected domains from the binary foreground difference image and calculate the barycentric coordinates of each connected domain;

[0124] distinguish and judge the connected domains according to the barycentric coordinates between adjacent connected domains;

[0125] count the number of different connected domains and calculate the flow rate of the liquid in combination with the drop ratio;

[0126] construct an algorithm and calculate infusion indicators, including timing consistency score and pressure diagnosis factor;

[0127] construct a risk vector from the physiological parameters of the patient and the infusion indicators, perform normalization processing to generate a dynamic risk mapping;

[0128] construct a risk assessment algorithm to calculate a risk score, set risk levels, and take corresponding measures for different risk levels.

[0129] The acquisition of the image data of the drip chamber when no infusion is performed, and the construction of the coordinate system and the background template of the image data, include:

[0130] continuously acquire N frames of images of the infusion drip chamber when no infusion is performed, denoted as initial images;

[0131] respectively take the pixel point located at the midpoint in the initial image as the coordinate origin of the initial image, and make a vertical upward ray through the coordinate origin as the y-axis;

[0132] take the ray in the initial image that is perpendicular to the y-axis and horizontally to the right as the x-axis, forming a coordinate system of the initial image;

[0133] The gray value at pixel point (x, y) in each initial image is acquired respectively, and a gray value sequence S is constructed x,y , specifically:

[0134] S x,y ={f1(x, y), f2(x, y), …, f N (x, y)};

[0135] Wherein, f N (x, y) represents the gray value of the Nth image at pixel point (x, y);

[0136] The gray value sequence S x,y is sorted in ascending order to obtain an ordered sequence S' x,y , specifically:

[0137] S' x,y = sort (S x,y ) = {s1, s2, …, s N};

[0138] Wherein, sort (S x,y ) represents a sorting function, that is, the gray values in S x,y are sorted in ascending order, s N represents the Nth gray value at pixel point (x, y) in S' x,y , and s1≤s2≤…≤s N ;

[0139] A background modeling algorithm is constructed, specifically as follows:

[0140]

[0141] Wherein, B(x, y) represents the gray value of the constructed background template at pixel point (x, y), and s k represents the kth element in S' x,y .

[0142] When no infusion is being administered, N consecutive frames of images of the drip chamber are acquired as the initial image sequence, and a two-dimensional image coordinate system is constructed with the midpoint pixel of the image as the origin. Subsequently, the gray values ​​of pixels at the same position in each frame are extracted to form a gray value sequence, which is then sorted. Based on the sorting result, the k-th smallest gray value is selected as the background gray value of the corresponding pixel in the background template, thus completing the construction of the static background template. By acquiring multiple frames of images of the drip chamber area in the non-infusion state and constructing a gray-scale background template, background noise and environmental interference factors are effectively eliminated, improving the accuracy and robustness of subsequent foreground difference images. At the same time, the establishment of a standardized coordinate system provides a unified spatial reference for droplet trajectory recognition and connected component centroid positioning, which helps to realize the fusion calculation of image analysis and physical monitoring data, thereby improving the reliability and accuracy of the entire infusion monitoring system.

[0143] The step of acquiring image data during the infusion process and constructing a binary foreground difference image includes:

[0144] Once the infusion begins, M frames of images of the infusion drip chamber are continuously acquired and recorded as infusion images;

[0145] The pixel located at the midpoint of the infusion image is taken as the origin of the coordinate system of the infusion image. A vertical ray is drawn through the origin as the vertical axis, i.e., the y-axis.

[0146] The ray perpendicular to the vertical axis and pointing horizontally to the right in the infusion image is used as the horizontal axis, i.e., the x-axis, to form the coordinate system of the infusion image;

[0147] The foreground difference extraction algorithm is constructed as follows:

[0148]

[0149] Among them, D s_i (x, y) represents the binary foreground mask image of the s_i-th frame of the infusion image, where 1 and 0 represent the values ​​of the foreground mask, respectively. s_i (x,y) represents the gray value at pixel (x,y) of the s_i-th frame of the infusion image, δ s_i (x,y) represents the adaptive difference threshold at pixel (x,y) of the s_i-th frame of the infusion image, as detailed below:

[0150] δ s_i (x,y)=μ s_i +2σ s_i ;

[0151] Where, μ s_i σ represents the average grayscale value of the infusion image in frame s_i. s_i This represents the standard deviation of the grayscale values ​​in the s_i-th frame of the infusion image;

[0152] the average gray value, specifically:

[0153]

[0154] wherein H represents the vertical pixel number of the infusion image, and W represents the horizontal pixel number of the infusion image;

[0155] the gray standard deviation, specifically:

[0156]

[0157] After the infusion starts, M frames of infusion drip chamber images are continuously collected, and a coordinate system is established with the center pixel point of the image; an adaptive differential threshold is constructed through the gray value, the average gray value and the gray standard deviation of each frame of image; the current frame of image is subjected to differential processing by using the threshold to generate a corresponding binary foreground mask image. This method can effectively extract the foreground region of the liquid drop movement, and lay a foundation for subsequent connected domain recognition and liquid drop trajectory tracking. By introducing the adaptive differential threshold, this method overcomes the problem of low recognition accuracy of the traditional fixed threshold under different illuminations or backgrounds, and can more accurately extract the foreground of the liquid drop image; at the same time, the foreground extraction is based on the statistical characteristics of the image, which improves the robustness and real-time performance of the liquid drop recognition, and helps to realize the accurate detection and intelligent monitoring of the liquid drop trajectory in the infusion process.

[0158] the connected domain is extracted from the binary foreground difference image, and the barycenter coordinates of each connected domain are calculated, including:

[0159] marking the connected domain of the binary foreground mask image corresponding to the s_i-th frame of infusion image;

[0160] The connected domain refers to a region composed of a specified pixel point as a center point and a pixel point with a surrounding adjacent foreground mask of 1;

[0161] extracting a pixel set corresponding to each connected domain, denoted as R l_k ;

[0162] wherein R l_k represents the pixel point set of the l_k-th connected domain;

[0163] The barycenter coordinates of each connected domain are calculated in the order of acquisition of the infusion image where the connected domain is located, specifically:

[0164]

[0165] wherein represents the barycenter coordinates of the l_k-th connected domain in the c-th frame of infusion image, c is less than or equal to M, and |R l_k | represents the total number of pixel points in the l_k-th connected domain.

[0166] The distinguishing judgment of the key point coordinates between adjacent connected domains includes:

[0167] According to the acquisition order of the infusion image, the distance of the key point coordinates between adjacent connected domains is calculated in sequence, specifically:

[0168] If The (l_k+1)th connected domain is determined to be the same target continuation track of the (l_k)th connected domain, that is, the same connected domain, τ d represents the maximum reachable distance threshold;

[0169] Otherwise, it is determined that the (l_k+1)th connected domain is not the same target continuation track as the (l_k)th connected domain, that is, different connected domains;

[0170] The above-mentioned maximum reachable distance threshold τ d = v p × Δt is calculated;

[0171] Where Δt represents the frame interval time of the infusion image acquisition, and v p represents the pixel speed of the center of gravity coordinates of the same connected domain, specifically:

[0172]

[0173] Where c_n is the pixel point base, and since the pixels of the photographed image are different, the value is also different, and the value range can be [5, 10].

[0174] The application provides a target distinguishing judgment method based on adjacent connected domain center of gravity coordinates, which realizes the continuity recognition of the droplet motion track by sequentially calculating the distance of the connected domain center of gravity coordinates between adjacent frames in the infusion image, combined with the preset maximum reachable distance threshold. The threshold is dynamically calculated according to the frame interval time and the pixel speed, ensuring that the judgment basis has time consistency and physical rationality, thereby effectively distinguishing the continuous track of the same droplet and the independent track of different droplets. The method significantly improves the accuracy and robustness of droplet recognition by constructing a connected domain tracking mechanism based on coordinate changes, avoiding the misidentification problem caused by droplet speed changes or image jitter in traditional image processing. At the same time, the method has good time matching ability and dynamic adaptability, and can be applied to infusion scenes with different droplet speeds and sampling frequencies, providing a reliable data basis for subsequent flow rate calculation and infusion risk identification.

[0175] The number of different connected domains is counted and combined with the droplet ratio to calculate the liquid flow rate, including:

[0176] The number of connected domains is obtained as the number of droplets, denoted as new_k;

[0177] Obtain the drop ratio of the current drip chamber model, denoted as r drip ;

[0178] Calculate the actual flow rate F of the liquid at the current time t , as follows:

[0179]

[0180] Where T m represents the time at which M frames of infusion images are collected.

[0181] The surgical room nursing infusion intelligent monitoring method provided by the application extracts the number of connected domains in the infusion process through image processing means, considers it as the number of drops, and combines the preset drip chamber drop ratio parameter to calculate the actual liquid flow rate per unit time. This process acquires consecutive image frames and counts the number of recognized connected domains, and combines the drop ratio model to realize the quantitative expression of flow rate, which has strong real-time and accuracy. In this way, the accurate measurement of liquid flow rate can be realized without additional hardware devices, overcoming the disadvantages of large drop speed estimation error and dependence on manual observation in the traditional infusion monitoring method, especially suitable for scenes such as operating rooms that require high nursing safety and monitoring accuracy; at the same time, the method has good adaptability and scalability, and can be flexibly applied to various infusion systems combined with different model drip chamber parameters, improving the intelligent level and risk prevention and control ability of the infusion process.

[0182] The construction algorithm calculates the infusion indicators, including the time sequence consistency score and the pressure diagnostic factor, including:

[0183] The time sequence consistency recognition algorithm is designed, specifically:

[0184]

[0185] Where S PCL (t) is the time sequence consistency score at time t, and the value range of time_i is [1, T m -t];

[0186] represents the pressure change rate at t-time_i;

[0187] represents the flow rate change rate at t-time_i;

[0188] represents the liquid level change rate at t-time_i;

[0189] sign(*) is a sign function, which returns 1 when is greater than 0, and returns 0 when When equal to 0, the return value is 0, when less than 0, the return value is -1;

[0190] Design differential pressure tracking diagnosis algorithm, calculate pressure diagnosis factor D FTD (t), specifically:

[0191]

[0192] Where, P t+1 and F t+1 respectively represent the pressure and liquid flow rate at t+1 moment.

[0193] The risk vector is constructed by the patient physiological parameters and the infusion index, and the dynamic risk mapping is generated by normalization processing, comprising:

[0194] The risk vector is constructed, denoted as Risk t =[S PCL (t), D FTD (t), H t ];

[0195] Define each item in the risk vector as r Risk_i (t)∈Risk t ;

[0196] Get the maximum and minimum values of each item r Risk_i (t) in the risk vector between time [t, T m ], and denote them as and

[0197] Each item r Risk_i (t) in the risk vector is processed by dimensionless, specifically:

[0198]

[0199] Where, Risk_i represents the Risk_i risk index in the risk vector after dimensionless processing, and epsilon is the denominator protection coefficient, which can be 0.0001, to prevent the denominator from being zero;

[0200] The risk vector after dimensionless processing is denoted as

[0201] The application designs timing consistency recognition algorithm and differential pressure tracking diagnosis algorithm, constructs key indicators in the infusion process, timing consistency score and pressure diagnosis factor, which can dynamically reflect the change trend of pressure, flow rate and liquid level and their mutual relationship in the infusion process. Based on this, the patient physiological parameters and infusion indicators are further fused to construct a risk vector, and a dynamic risk mapping is generated through normalization processing, realizing multi-dimensional and real-time quantitative evaluation of infusion risk. The method can accurately capture the small changes of infusion state, combined with physiological data such as patient heart rate and blood pressure, improve the individualization and accuracy of risk assessment; through dimensionless processing, the interference caused by the dimensional difference of data is effectively avoided, and the robustness and stability of the algorithm are enhanced; overall, the intelligent level of infusion monitoring in the operating room is significantly improved, early warning and dynamic management of potential abnormalities are realized, patient fluid safety is guaranteed, and the work burden of nursing personnel is reduced.

[0202] The risk assessment algorithm calculates a risk score and sets a risk level, and corresponding measures are taken for different risk levels, including:

[0203] The risk score algorithm is designed, specifically:

[0204]

[0205] Among them, represents the risk comprehensive score, represents the maximum value in the vector at time t.

[0206] The risk score level level is set, specifically:

[0207]

[0208] When Level = 1, the risk level is set to normal and continues to be monitored.

[0209] When Level = 2, the risk level is set to medium risk, and the nursing staff is reminded to check.

[0210] When Level = 3, the risk level is set to high risk, and the infusion operation is immediately terminated and an alarm is given.

[0211] The application provides a surgery room nursing infusion intelligent monitoring method based on a constructed risk assessment algorithm, and scientific division of risk levels in an infusion process is realized by calculating a comprehensive score of a dynamic risk vector. The method sets three risk levels of normal, medium risk and high risk, and corresponding monitoring and intervention measures are taken for different risk levels, including continuous monitoring, reminding nursing staff to check, and emergency termination of infusion and alarm, so that infusion safety is ensured. The method can evaluate potential risks in the infusion process in real time and accurately, comprehensively judge in combination with multidimensional dynamic data, improve the intelligence and accuracy of infusion monitoring, effectively reduce human negligence and missed reporting, ensure patient medication safety, and improve the overall quality and efficiency of surgical nursing.

[0212] Referring to Figure 2 A monitoring system for implementing the surgery room nursing infusion intelligent monitoring method, comprising:

[0213] A data collection module collects a set of infusion parameters in a surgery room for subsequent judgment of the risk level of infusion;

[0214] An image acquisition module acquires images before and during infusion to monitor whether an abnormality occurs in the infusion process in real time;

[0215] An image data processing module identifies falling droplets by processing the acquired image data;

[0216] A data calculation module calculates the pixel movement speed of the connected domain key coordinates, and further calculates the flow speed of the liquid according to the movement of the connected domain key coordinates;

[0217] A risk judgment module calculates the comprehensive risk score of the infusion in real time through the risk vector, and judges the risk level of the infusion.

[0218] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0219] The above is only the preferred embodiment of the application, and it should be noted that for those skilled in the art, without departing from the technical principles of the application, a number of improvements and refinements can be made, and these improvements and refinements should also be considered as the protection scope of the application.

Claims

1. An operating room nursing infusion intelligent monitoring method, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. A set of infusion parameters D at a time t is collected t which includes: The pressure sensor is used to collect the pressure of the liquid in the infusion pipeline at time t, which is recorded as P t ; calculating the flow rate of the liquid in the infusion line, denoted F t ; Obtain the liquid level of the infusion bottle, denoted as L t ; constructing a set of patient physiological state parameters H t including: The physiological parameters of a patient, including heart rate (HR), systolic blood pressure (SBP) and diastolic blood pressure (DBP); Image data of the drip chamber when no infusion is performed is collected, and a coordinate system and a background template of the image data are constructed; Image data during infusion is obtained, and a binary foreground difference image is constructed; By means of the binary foreground difference image, connected domains are extracted, and the barycentric coordinates of each connected domain are calculated; Adjacent connected domains are distinguished and judged according to the barycentric coordinates; The number of different connected domains is counted, and the liquid flow rate is calculated in combination with the drop number ratio; An algorithm is constructed, and infusion indexes, including time sequence consistency scores and pressure diagnostic factors, are calculated; A risk vector is constructed by means of the physiological parameters of the patient and the infusion indexes, and is normalized to generate a dynamic risk mapping; A risk assessment algorithm is constructed to calculate a risk score, and risk levels are set, and corresponding measures are taken for different risk levels.

2. The operating room nursing infusion intelligent monitoring method according to claim 1, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. N frames of images of the infusion drip chamber when no infusion is performed are continuously collected, and are recorded as initial images; A pixel point located at a midpoint in the initial image is taken as a coordinate origin of the initial image, a vertical upward ray is made through the coordinate origin, and the vertical upward ray is taken as a y-axis; A ray in the initial image which is perpendicular to the y-axis and horizontally rightward is taken as an x-axis, and a coordinate system of the initial image is formed; Acquire the gray value of each initial image at pixel point (x, y) respectively, and construct a gray value sequence S x,y Specifically: S x,y = {f1(x,y), f2(x,y),..., f N (x,y)}; wherein f N (x,y) denotes the gray value of the Nth frame image at the pixel point (x, y); The gray value sequence S x,y In ascending order, the ordered sequence S′ is obtained x,y , specifically: S' x,y = sort(S x,y ) = {s1,s2,……,s N}; wherein sort(S x,y ) represents a sorting function, i.e., sorting the gray values in S x,y from small to large, s N represents the Nth gray value in S′ x,y located at the pixel point (x, y), and s1≤s2≤……≤s N ; A background modeling algorithm is constructed, and the algorithm is as follows: where B(x, y) represents the gray value of the constructed background template at pixel point (x, y), s k represents the kth element in S' x,y .

3. The operating room nursing infusion intelligent monitoring method according to claim 2, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. When infusion starts, M frames of images of the infusion drip chamber are continuously obtained, and are recorded as infusion images; A pixel point located at a midpoint in the infusion image is taken as a coordinate origin of the infusion image, a vertical upward ray is made through the coordinate origin, and the vertical upward ray is taken as a y-axis; A ray in the infusion image which is perpendicular to the y-axis and horizontally rightward is taken as an x-axis, and a coordinate system of the infusion image is formed; A foreground difference extraction algorithm is constructed, and the algorithm is as follows: where D s_i (x, y) represents the binary foreground mask image of the s_i-th infusion image, 1 and 0 represent the values of the foreground mask respectively, f s_i (x, y) represents the gray value of the s_i-th infusion image at the pixel point (x, y), δ s_i (x, y) represents the adaptive differential threshold value of the s_i-th infusion image at the pixel point (x, y), which is specifically as follows: delta s_i (x,y) = mu s_i + 2sigma s_i ; where μ s_i represents the average gray value of the s_i-th frame infusion image, σ s_i represents the gray standard deviation of the s_i-th frame infusion image; The average gray value is specifically as follows: Wherein, H represents the vertical pixel number of the infusion image, and W represents the horizontal pixel number of the infusion image; The gray standard deviation is specifically as follows:

4. The operating room nursing infusion intelligent monitoring method according to claim 3, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. extracting a pixel set corresponding to each connected domain, denoted as R l_k ; wherein R l_k represents the pixel point set of the l_kth connected domain; The barycentric coordinates of each connected domain are calculated in sequence according to the acquisition sequence of the infusion images in which the connected domains are located, and the calculation is specifically as follows: wherein, represents the barycentric coordinates of the l_kth connected domain in the cth frame infusion image, c is less than or equal to M, |R l_k | represents the total number of pixel points in the l_kth connected domain.

5. The operating room nursing infusion intelligent monitoring method according to claim 4, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. The distance between the barycentric coordinates of adjacent connected domains is calculated in sequence according to the acquisition sequence of the infusion images, and the calculation is specifically as follows: If determines that the (l_k+1)th connected domain is the same target continuation track as the l_kth connected domain, i.e., the same connected domain, τ d denotes the threshold of the farthest reachable distance; Otherwise, it is determined that the (l_k+1)th connected domain and the (l_k)th connected domain are not the same target continuation track, that is, the connected domains are different. calculating the furthest reachable distance threshold τ d = v p x Δt; Wherein, Δt represents the inter-frame interval time of infusion image acquisition, v p represents the pixel speed of the same connected domain barycentric coordinate movement, specifically: Wherein, c_n is a pixel point base, and the value range is [5, 10].

6. The operating room nursing infusion intelligent monitoring method according to claim 1, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. The number of connected domains is obtained, and is taken as the number of liquid drops, and is recorded as new_k. Obtain the drop ratio of the current gondola model, denoted as r drip ; calculating the actual flow rate F of the liquid at the current time t as follows: wherein T m represents the time at which the M infusion images are acquired.

7. The operating room nursing infusion intelligent monitoring method according to claim 1, characterized in that, The application relates to a risk assessment method for infusion, and belongs to the field of medical treatment. A time sequence consistency identification algorithm is designed, and the algorithm is as follows: S PCL (t) the time consistency score at time t, time_i∈[1, T m -t] P(t) represents the pressure change rate at the t-time_i moment; represents the flow rate change rate at the t-time_i moment; represents the rate of change of the liquid level at the t-time_i moment; sign(*) is a sign function that returns a value of 1 when greater than 0, a value of 0 when equal to 0, and a value of -1 when less than 0. A differential pressure tracking diagnosis algorithm is designed to calculate the pressure diagnosis factor D FTD (t), specifically: where P t+1 and F t+1 represent the pressure and liquid flow rate at time t+1, respectively.

8. The operating room nursing infusion intelligent monitoring method according to claim 7, characterized in that, The risk vector is constructed by the patient physiological parameters and the infusion index, normalized processing generates a dynamic risk mapping, comprising: Construct a risk vector, denoted as Risk t = [S PCL (t), D FTD (t), H t ] Define each item in the risk vector as r Risk_i (t) ∈ Risk t ; obtaining each item r in the risk vector Risk_i (t) the maximum and minimum values between time instants [t, T m ] and denoted by and For each term r in the risk vector Risk_i (t) Normalization, in particular: wherein, Risk_i represents the i-th risk indicator in the risk vector after non-dimensionalization processing, and ε is a denominator protection coefficient. The risk vector after the dimensionless treatment is denoted as 9. The operating room nursing infusion intelligent monitoring method according to claim 8, characterized in that, The risk assessment algorithm calculates a risk score and sets a risk level, and corresponding measures are taken for different risk levels, comprising: The risk score algorithm is designed, specifically: wherein, denotes the risk composite score, denotes the maximum value in the vector at time t; The risk score level level is set, specifically: When Level=1, the risk level is set to normal and continues to be monitored; When Level=2, the risk level is set to medium risk, and the nursing staff is reminded to check; When Level=3, the risk level is set to high risk, and the infusion operation is immediately terminated, and an alarm is given.

10. A monitoring system using the intelligent monitoring method for operating room nursing infusion according to claim 1, characterized in that, Comprising: The data collection module collects the infusion parameter set in the operating room for subsequent judgment of the risk level of infusion; The image acquisition module collects images before and during infusion to monitor the infusion process in real time for abnormalities; The image data processing module processes the collected image data to identify the falling droplets; The data calculation module calculates the pixel movement speed according to the movement of the connected domain key coordinates, and then calculates the flow speed of the liquid; The risk judgment module calculates the risk comprehensive score of the infusion in real time through the risk vector to judge the risk level of the infusion.

Images