A method and system for managing infusion based on artificial intelligence

Abstract

The application belongs to the field of intelligent medical care, and discloses a method and system for managing infusion based on artificial intelligence. The application constructs a multi-modal infusion management model by using infusion tube data, infusion tube monitoring image data, patient information feature data, and infusion reminder levels. The constructed multi-modal infusion management model generates a current infusion reminder level for medical staff during the infusion process of a patient according to real-time infusion tube data, real-time infusion tube monitoring image data, and real-time patient information feature data collected by each medical device received by the medical station. The current infusion reminder level is sent to the nurse station screen and the handheld terminal device to remind the medical staff to go to the corresponding ward to pull the needle in time, and the transmission of the infusion reminder level information is closed to achieve precise infusion care. The application meets the real-time monitoring of infusion of patients in each ward by medical staff, realizes intelligent infusion monitoring effect, reduces the work intensity of medical staff on ward rounds, and reduces the doctor-patient contradiction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent medical care, and particularly relates to a method and system for managing infusion based on artificial intelligence. BACKGROUND

[0002] According to the investigation of the National Health Commission in 2023, 90% of inpatients need infusion treatment, 50% of patients have no accompanying personnel, children account for 22% of infusion patients, medical disputes account for 26% during infusion, the average infusion time is 220 minutes per person, and the time spent by clinical nurses in solving infusion links accounts for 30%. Reducing medical disputes during infusion, improving infusion drug safety, and improving patient satisfaction have become urgent problems that need to be solved.

[0003] The existing infusion is that after the liquid is infused, the nurse receives the message and then pulls the needle. At the same time, the patient cannot rest during the infusion, and the medical staff needs to observe the liquid at all times to prevent the liquid from being infused and not being able to notify the medical staff in time, causing backflow. Under the condition that the medical staff observes the infusion for a long time, visual fatigue of the medical staff is easy to occur, and mental stress is increased. In addition, the medical staff will increase the frequency of visiting the bed during the infusion. According to the calculation of 50 patients in a department, one patient is visited for one minute, and 50 minutes are needed for one time. Until the infusion of all patients in the ward is completed, the nurse visits the bed not less than 3 times. In addition, the notification of termination of infusion and needle pulling for each patient, the workload is very large and time-consuming and labor-consuming. Such high-intensity work brings great mental and physical loss.

[0004] For some patients with difficulty in movement or the elderly, nursing personnel need to be hired to observe the infusion at all times during the infusion. Once the infusion is completed, the nurse is notified in the first time, which greatly increases the labor cost. Due to too many infusion patients, the bell of the nurse station rings one after another after the liquid is infused, which also affects the rest of the patients.

[0005] The existing technology has the following infusion management methods. The first infusion monitoring is a single machine version and cannot be connected to a network. After the liquid is infused, a drop sound is emitted, but it is limited to use in the ward, that is, it affects the rest of the patients in the room. The nurse cannot receive the signal that the infusion is completed, and still needs to manually notify the nurse to replace the liquid or terminate the infusion. The second traditional weighing infusion alarm device cannot accurately alarm immediately after the liquid is infused, and the operation is complex. A large number of parameters need to be input before use. Since the weight of the infusion bottle or infusion bag generated by each manufacturer is different, false reporting and omission may occur. In addition, the terminal screen of the interconnection cannot display whether the patient is allergic to the drug and the power of the device.

[0006] Therefore, an artificial intelligence-based infusion management method and system are urgently needed, which can reduce the frequency of nurse bed rounds, reduce the work intensity of medical staff, accurately identify infusion conditions, interconnect with medical staff terminals, correctly alarm infusion conditions in the ward, timely remind medical staff to remove the needle, and reduce the occurrence of medical staff conflicts. SUMMARY

[0007] To solve the above technical problems, the present application provides an artificial intelligence-based infusion management method and system.

[0008] In the first aspect of the present application, an artificial intelligence-based infusion management method is provided, characterized in that the method comprises:

[0009] obtaining first infusion tube data and first infusion tube monitoring image data of an infusion tube, first patient information feature data of a patient, and a first infusion reminder level;

[0010] forming a first infusion management vector feature according to the first infusion tube data and the first infusion tube monitoring image data, and the first patient information feature data;

[0011] constructing a multi-modal infusion management model using the first infusion management vector feature and the first infusion reminder level;

[0012] receiving second infusion tube data and second infusion tube monitoring image data of an infusion tube, and second patient information feature data of a patient, generating a second infusion management vector feature according to the second infusion tube data, the second infusion tube monitoring image data, and the second patient information feature data, processing the second infusion management vector feature using the multi-modal infusion management model to generate a second infusion reminder level, and reminding medical staff to perform infusion care according to the second infusion reminder level.

[0013] Further, the first infusion tube data or the second infusion tube data uses an infrared probe of an infusion monitor clamped on the upper end of an infusion tube drip bottle to check the liquid in the infusion tube, and the first infusion tube data or the second infusion tube data is specifically the first or second to-be-infused liquid height probed by the infrared probe with the upper end of the infusion tube drip bottle as the origin.

[0014] Further, the first infusion tube monitoring image data or the second infusion tube monitoring image data includes infusion bag images and drip bottle images.

[0015] The first patient information feature data of the patient or the second patient information feature data of the patient includes the severity of the patient's illness and the age of the patient.

[0016] Further, before constructing or utilizing the multi-modal infusion management model, target feature processing is performed on the infusion bag image and the drip bottle image of the first infusion tube monitoring image data or the second infusion tube monitoring image data, and the target feature part is separated from the non-target feature part. The threshold variance R between the target feature part and the non-target feature part is defined.

[0017] Further, the mean of the sum of the image feature gray scale values greater than the threshold variance R in the first infusion tube monitoring image data or the second infusion tube monitoring image data is taken as the first target feature value or the second target feature value.

[0018] Further, the first target feature value and the first patient information feature data are combined to obtain a first infusion management vector feature, and the second target feature value and the second patient information feature data are combined to obtain a second infusion management vector feature.

[0019] Further, the multi-modal infusion management model adopts an improved Fisher criterion classifier.

[0020] Also provided is an artificial intelligence-based infusion management system, which comprises an infusion monitor module, a ward infusion tube monitoring module, a nurse station patient information storage module, a nurse station image processing module, a multi-modal infusion management model construction module, and an infusion reminder module, characterized in that:

[0021] The infusion monitor module is configured to acquire first infusion tube data of an infusion tube and to acquire second infusion tube data of the infusion tube.

[0022] The ward infusion tube monitoring module is configured to acquire first infusion tube monitoring image data of an infusion tube and to acquire second infusion tube monitoring image data of the infusion tube.

[0023] The nurse station patient information storage module is configured to receive first patient information features of a patient, to receive second patient information features of the patient, to receive a first infusion reminder level, and to receive and store second infusion management vector features and a second infusion reminder level.

[0024] The nurse station image processing module is configured to form a first infusion management vector feature based on the first infusion tube data, the first infusion tube monitoring image data, and the first patient information feature data, and to form a second infusion management vector feature based on the second infusion tube data, the second infusion tube monitoring image data, and the second patient information feature data.

[0025] The multi-modal infusion management model construction module receives the first infusion management vector feature from the nurse station image processing module, and constructs a multi-modal infusion management model by using the first infusion management vector feature and the first infusion reminding level.

[0026] The infusion reminding module receives the second infusion management vector feature, and calls the multi-modal infusion management model in the multi-modal infusion management model construction module to process the second infusion reminding level, which is sent to the nurse station large screen and the display screen of the medical staff handheld device for reminding.

[0027] Further, the first infusion tube monitoring image data or the second infusion tube monitoring image data comprises infusion bag images and drip bottle images.

[0028] The first patient information feature data of the patient or the second patient information feature data of the patient comprises the severity of the disease of the patient and the age of the patient.

[0029] Further, before the multi-modal infusion management model is constructed or utilized, the infusion bag images and the drip bottle images of the first infusion tube monitoring image data or the second infusion tube monitoring image data are subjected to target feature processing, the target feature parts are separated from the non-target feature parts, and the threshold variance R between the target feature parts and the non-target feature parts is defined.

[0030] The method and system for managing infusion based on artificial intelligence of the present application are different from the conventional method of processing by combining the weight of the infusion tube with the image. The present application adopts the infusion liquid level height combined with the infusion liquid image feature value obtained after processing and combines the patient feature information to determine the infusion reminding level. The improved machine learning algorithm is more accurate for the infusion reminding of the infusion tube in a specific application scenario, and the weight distribution is performed considering the image feature difference between the drip bottle and the infusion bag. The inaccurate infusion monitoring caused by the different sizes of infusion bags due to different manufacturers is ignored, the real-time infusion monitoring of the infusion patients in each ward of the medical staff is satisfied, the intelligent infusion monitoring effect is realized, the work intensity of the medical staff is reduced, and the doctor-patient contradiction is reduced, so that the frequency of nurse bed rounds can be reduced, the work intensity of the medical staff can be reduced, the infusion situation can be accurately identified, the medical staff terminal can be interconnected, the ward infusion situation can be correctly alarmed, the medical staff can be timely reminded to pull out the needle, and the occurrence of medical staff contradiction can be reduced. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a method flowchart for managing infusion based on artificial intelligence of the present application;

[0032] Figure 2It is a structure schematic diagram of an infusion system based on artificial intelligence management according to the present application.

[0033] Figure 3 It is a classification principle diagram of a classifier based on Fisher criterion in the present application.

[0034] Figure 4 It is a target feature part diagram of an infusion tube in an embodiment of the present application.

[0035] Among them, 1 represents an infusion bag, and 2 represents a drip bottle.

[0036] Figure 5 It is an infusion monitoring schematic diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0037] In the following, the application is further described in combination with the drawings and the specific embodiments.

[0038] The first embodiment of the present application is as follows:

[0039] In the first aspect of the present application, in order to solve the above technical problems, a method for managing infusion based on artificial intelligence is provided, characterized in that the method comprises:

[0040] obtaining first infusion tube data and first infusion tube monitoring image data of an infusion tube, first patient information feature data of a patient, and a first infusion reminding level;

[0041] forming a first infusion management vector feature according to the first infusion tube data and the first infusion tube monitoring image data, and the first patient information feature data;

[0042] constructing a multi-modal infusion management model by using the first infusion management vector feature and the first infusion reminding level;

[0043] receiving second infusion tube data and second infusion tube monitoring image data of an infusion tube, and second patient information feature data of a patient, generating a second infusion management vector feature according to the second infusion tube data, the second infusion tube monitoring image data, and the second patient information feature data, processing the second infusion management vector feature by using the multi-modal infusion management model to generate a second infusion reminding level, and reminding medical staff to perform infusion nursing according to the second infusion reminding level.

[0044] The first infusion reminding level and the second infusion reminding level in the embodiment include a vector representation of the infusion remaining time and the reminding signal composition, such as the infusion reminding level vector L(0, 600, 0) representing a primary infusion reminding level, the infusion remaining time of 600 seconds to the infusion tube, no countdown, and sending a 0-level signal green signal to the nurse station large screen and the mobile terminal device. Such as the infusion reminding level vector L(-1, 60, 1) representing a medium infusion reminding level, the infusion remaining time of 60 seconds to the infusion tube, 60 seconds countdown, and sending a 1-level signal yellow signal to the nurse station large screen and the mobile terminal device. The infusion reminding level vector L can be set to the level required by the technical personnel. Such as the infusion reminding level vector L(1, 30, 2) representing a high infusion reminding level, the infusion remaining time of 30 seconds to the infusion tube, 30 seconds countdown, and sending a 2-level signal red signal to the nurse station large screen and the mobile terminal device. The infusion reminding level vector L can be set to the level required by the technical personnel.

[0045] Further, the first infusion tube data or the second infusion tube data checks the liquid in the infusion tube by using the infrared probe of the infusion monitor card on the upper end of the infusion tube, and the first infusion tube data or the second infusion tube data is specifically the first to-be-infused liquid height or the second to-be-infused liquid height probed by the infrared probe with the upper end of the infusion tube drip bottle as the origin.

[0046] Further, the first infusion tube monitoring image data or the second infusion tube monitoring image data includes infusion bag images and drip bottle images.

[0047] The first patient information feature data of the patient or the second patient information feature data of the patient includes the severity of the disease of the patient and the age of the patient.

[0048] Further, before the multi-modal infusion management model is constructed or utilized, the infusion bag images and the drip bottle images of the first infusion tube monitoring image data or the second infusion tube monitoring image data are subjected to target feature processing, and the target feature parts are separated from the non-target feature parts. The threshold variance R between the target feature parts and the non-target feature parts is defined as:

[0049] ;

[0050] In the formula, is the proportion of the gray value belonging to the target feature part of the infusion bag image to the overall gray value of the infusion bag image, is the proportion of the gray value belonging to the non-target feature part of the infusion bag image to the overall gray value of the infusion bag image, is the proportion of the gray value belonging to the target feature part of the drip bottle image to the overall gray value of the drip bottle image, a proportion of the gray value belonging to the non-target feature part of the drip bottle image in the overall gray value of the drip bottle image, G is the total average gray value of the infusion bag image, a target feature part of the infusion bag image, a non-target feature part of the infusion bag image, A is the total average gray value of the drip bottle image, a target feature part of the drip bottle image, a non-target feature part of the drip bottle image, 、 is a weight coefficient.

[0051] Further, the mean value of the sum of the image feature gray values greater than the threshold variance R in the first infusion tube monitoring image data or the second infusion tube monitoring image data is taken as the first target feature value or the second target feature value.

[0052] Further, the first to-be-infused liquid height, the first target feature value, and the first patient information feature data are adhered to obtain a first infusion management vector feature, and the second to-be-infused liquid height, the second target feature value, and the second patient information feature data are adhered to obtain a second infusion management vector feature.

[0053] Further, the multi-modal infusion management model adopts an improved Fisher criterion classifier, and the calculation formula is as follows:

[0054] ;

[0055] In the formula, X is the input infusion management vector feature, g(X) is the output infusion reminder level, W T is a normal vector perpendicular to the hyperplane, a proportion of the gray value belonging to the target feature part of the infusion bag image in the overall gray value of the infusion bag image, a proportion of the gray value belonging to the non-target feature part of the infusion bag image in the overall gray value of the infusion bag image, a proportion of the gray value belonging to the target feature part of the drip bottle image in the overall gray value of the drip bottle image, a proportion of the gray value belonging to the non-target feature part of the drip bottle image in the overall gray value of the drip bottle image, 、 is a weight coefficient, .

[0056] In this embodiment, the value of g(X) is used to output the infusion reminder level L. If the value of g(X) is less than 0, the medium infusion reminder level is output, if it is equal to 0, the primary infusion reminder level is output, and if it is greater than 0, the high infusion reminder level is output. The value of g(X) is the product of each feature vector in the output infusion reminder level vector L.

[0057] Also provided is an artificial intelligence-based infusion system management system, comprising an infusion monitor module, a ward infusion tube monitoring module, a nurse station patient information storage module, a nurse station image processing module, a multi-modal infusion management model construction module, and an infusion reminder module, characterized in that:

[0058] The infusion monitor module is configured to collect first infusion tube data of an infusion tube and to collect second infusion tube data of the infusion tube.

[0059] The ward infusion tube monitoring module is configured to collect first infusion tube monitoring image data of an infusion tube and to collect second infusion tube monitoring image data of the infusion tube.

[0060] The nurse station patient information storage module is configured to receive first patient information features of a patient and to receive second patient information features of the patient, to receive a first infusion reminder level, and to receive and store second infusion management vector features and a second infusion reminder level.

[0061] The nurse station image processing module is configured to form a first infusion management vector feature based on the first infusion tube data, the first infusion tube monitoring image data, and the first patient information feature data, and to form a second infusion management vector feature based on the second infusion tube data, the second infusion tube monitoring image data, and the second patient information feature data.

[0062] The multi-modal infusion management model construction module receives the first infusion management vector feature from the nurse station image processing module and uses the first infusion management vector feature and the first infusion reminder level to construct a multi-modal infusion management model.

[0063] The infusion reminder module receives the second infusion management vector feature and calls the multi-modal infusion management model in the multi-modal infusion management model construction module to process the second infusion reminder level, which is sent to a nurse station large screen and a medical staff handheld device display screen for reminding.

[0064] Further, the first infusion tube monitoring image data or the second infusion tube monitoring image data comprises an infusion bag image and a drip bottle image.

[0065] The first patient information feature data of the patient or the second patient information feature data of the patient comprises the disease severity of the patient and the age of the patient.

[0066] Further, before the multi-modal infusion management model is constructed or utilized, target feature processing is performed on the infusion bag image and the drip bottle image of the first infusion tube monitoring image data or the second infusion tube monitoring image data, target feature parts are separated from non-target feature parts, and the threshold variance R between the target feature parts and the non-target feature parts is defined as:

[0067] ;

[0068] In the formula, is the proportion of the gray value belonging to the target feature part of the infusion bag image to the overall gray value of the infusion bag image, is the proportion of the gray value belonging to the non-target feature part of the infusion bag image to the overall gray value of the infusion bag image, is the proportion of the gray value belonging to the target feature part of the drip bottle image to the overall gray value of the drip bottle image, is the proportion of the gray value belonging to the non-target feature part of the drip bottle image to the overall gray value of the drip bottle image, G is the total average gray of the infusion bag image, is the average gray of the target feature part of the infusion bag image, is the average gray of the non-target feature part of the infusion bag image, A is the total average gray of the drip bottle image, is the average gray of the target feature part of the drip bottle image, is the average gray of the non-target feature part of the drip bottle image, 、 is a weight coefficient.

[0069] The method and system for managing infusion based on artificial intelligence of the present application are different from the conventional method of combining the weight of the infusion tube with the image for processing. The present application combines the infusion liquid level height with the infusion liquid image feature value obtained after processing and combines the patient feature information to determine the infusion reminding level. The improved machine learning algorithm is more accurate for the infusion reminding of the infusion tube in a specific application scene, and the weight distribution is performed considering the image feature difference between the drip bottle and the infusion bag, so as to meet the real-time monitoring of the infusion of the patients in each ward by the medical staff, realize the intelligent infusion monitoring effect, reduce the work intensity of the medical staff, and reduce the medical disputes, so as to reduce the frequency of nurse bed rounds, reduce the work intensity of the medical staff, accurately identify the infusion situation, interconnect with the medical station terminal, correctly alarm the infusion situation of the ward, timely remind the medical staff to pull out the needle, and reduce the occurrence of medical disputes.

[0070] The combination of the embodiments of the present application can realize all the effects described above, but it is not required that each embodiment of the present application realizes all the advantages and effects described above, because each embodiment of the present application can constitute a separate technical solution and make one or more contributions to the prior art.

[0071] The module structure of the part of the present application which is not particularly clear is subject to the content described in the prior art. The prior art mentioned in the foregoing background section and the specific embodiment section of the present application can be used as a part of the present application to understand the meaning of some technical features or parameters. The scope of protection of the present application is subject to the content actually described in the claims.

Claims

1. A method for managing intravenous infusions based on artificial intelligence, characterized in that, The method includes: Acquire the first infusion tube data and the first infusion tube monitoring image data, the patient's first patient information feature data, and the first infusion reminder level; The first infusion management vector feature is composed of the first infusion tube data, the first infusion tube monitoring image data, and the first patient information feature data; A multimodal infusion management model is constructed by utilizing the features of the first infusion management vector and the first infusion reminder level. The system receives second infusion tube data, second infusion tube monitoring image data, and second patient information feature data. Based on the second infusion tube data, second infusion tube monitoring image data, and second patient information feature data, a second infusion management vector feature is generated. The second infusion management vector feature is processed using the multimodal infusion management model to generate a second infusion reminder level. Based on the second infusion reminder level, medical staff are reminded to perform infusion care. The first infusion tube data or the second infusion tube data is used to check the liquid in the infusion tube by using an infrared probe of an infusion monitor that is attached to the upper end of the infusion tube drip chamber. Specifically, the first infusion tube data or the second infusion tube data is the height of the first infusion liquid or the height of the second infusion liquid detected by the infrared probe with the upper end of the infusion tube drip chamber as the origin. The monitoring image data of the first infusion tube or the monitoring image data of the second infusion tube includes images of the infusion bag and the drip chamber; The patient's first or second patient information feature data includes the severity of the patient's illness and the patient's age. Before constructing or utilizing the multimodal infusion management model, target feature processing is performed on the infusion bag image and the drip chamber image of the first infusion tube monitoring image data or the second infusion tube monitoring image data to separate the target feature parts from the non-target feature parts, and the threshold variance R between the target feature parts and the non-target feature parts is defined. ； In the formula, This refers to the proportion of the grayscale value of the target feature region in the infusion bag image to the total grayscale value of the infusion bag image. The grayscale value is the proportion of the grayscale value of the non-target feature area in the image of the infusion bag to the total grayscale value of the image of the infusion bag. This refers to the proportion of grayscale values ​​belonging to the target feature region of the dripper image to the total grayscale values ​​of the dripper image. G represents the proportion of gray values ​​belonging to non-target feature areas of the drip chamber image to the total gray value of the drip chamber image, and G is the total average gray value of the infusion bag image. The average gray level of the target feature region in the image of the infusion bag. A is the average gray level of the non-target feature area in the infusion bag image, and A is the total average gray level of the drip chamber image. The average gray level of the target feature region in the dripping pot image. The average gray level of the non-target feature area in the dripper image. , These are the weighting coefficients; The mean of the sum of image feature gray values ​​greater than the threshold variance R in the first infusion tube monitoring image data or the second infusion tube monitoring image data is used as the first target feature value or the second target feature value. The first infusion management vector feature is obtained by combining the first infusion fluid height, the first target feature value, and the first patient information feature data; the second infusion management vector feature is obtained by combining the second infusion fluid height, the second target feature value, and the second patient information feature data. The multimodal infusion management model employs an improved Fisher criterion-based classifier, and the calculation formula is shown below: ； In the formula, X represents the input infusion management vector feature, g(X) represents the output infusion reminder level, and W... T Let be the normal vector perpendicular to the hyperplane. This refers to the proportion of the grayscale value of the target feature region in the infusion bag image to the total grayscale value of the infusion bag image. The grayscale value is the proportion of the grayscale value of the non-target feature area in the image of the infusion bag to the total grayscale value of the image of the infusion bag. This refers to the proportion of grayscale values ​​belonging to the target feature region of the dripper image to the total grayscale values ​​of the dripper image. This refers to the proportion of gray values ​​belonging to non-target feature areas of the dripper image to the total gray values ​​of the dripper image. , These are the weighting coefficients. .

2. An artificial intelligence-based infusion management system, the system implementing the method as described in claim 1, the system comprising an infusion monitoring module, a ward infusion tube monitoring module, a nurse station patient information storage module, a nurse station image processing module, a multimodal infusion management model construction module, and an infusion reminder module, characterized in that: The infusion monitoring module is used to collect data from the first infusion tube and also to collect data from the second infusion tube. The ward infusion tube monitoring module is used to acquire monitoring image data of the first infusion tube and also to acquire monitoring image data of the second infusion tube. The nurse station patient information storage module is used to receive the patient's first patient information features, and also to receive the patient's second patient information features, receive the first infusion reminder level, and also to receive and store the second infusion management vector features and the second infusion reminder level; The nurse station image processing module is used to: define a first infusion management vector feature based on the first infusion tube data, the first infusion tube monitoring image data, and the first patient information feature data; and also to define a second infusion management vector feature based on the second infusion tube data, the second infusion tube monitoring image data, and the second patient information feature data. The multimodal infusion management model construction module: receives the first infusion management vector features from the nurse station image processing module, and uses the first infusion management vector features and the first infusion reminder level to jointly construct a multimodal infusion management model; The infusion reminder module receives the second infusion management vector feature and calls the multimodal infusion management model in the multimodal infusion management model construction module to process and obtain the second infusion reminder level. The second infusion reminder level is then sent to the large screen at the nurse station and the display screen of the handheld device of the medical staff for reminder.

Images