KR20220038968A - Artificial intelligence-based sleep anesthesia safety prediction device and method - Google Patents

Abstract

The present invention relates to an artificial intelligence-based sleep anesthesia safety prediction device and method. The device includes: a triggering factor defining unit defining a plurality of triggering factors related to side effects of sleep anesthesia based on user information of the patient; a bio-vitality collection unit for collecting bio-vitality indicators for the patient during the sleep anesthesia; a learning model building unit for performing machine learning on correlations between variables when the triggering factors and the vitality indicators are set as independent variables, and the presence or absence of the side effects during the sleep anesthesia are set as dependent variables; a side effect prediction unit for predicting the possibility of occurrence of side effects during sleep anesthesia of a specific patient by using the learning model generated as a result of the machine learning; and a safety level calculation unit for calculating a safety level of the sleep anesthesia of the specific patient based on the possibility of occurrence of the side effects. Therefore, the risk of respiratory depression is predicted by integrating and analyzing various triggering factors of side effects of sleep anesthesia and vitality indicators measured during the sleep anesthesia.

Description

Artificial intelligence-based sleep anesthesia safety prediction device and method

The present invention relates to a technology for predicting the safety of sleep anesthesia, and more specifically, an artificial intelligence-based technology that can predict the risk of respiratory depression by integrating various causative factors of sleep anesthesia side effects and bio vitality indicators measured during the sleep anesthesia process. It relates to a device and method for predicting the safety of sleep anesthesia.

In the case of an automobile accident, many situations with a high probability of an accident are known empirically. For example, the more frequently you use your cell phone while driving, the more you mainly use the road with many illegally parked vehicles, the higher the risk of breakdowns and malfunctions, the more the vehicle has no safety-related functions, and the time of day when visibility is poor, such as at night. The higher the driving speed, the slower the driver's reaction time, and the higher the average driving speed, the higher the probability of an accident.

However, if the above situations are individually judged, the occurrence of an accident can be predicted intuitively and easily, whereas it is impossible for an actual driver to numerically accurately predict the risk of an accident by synthesizing all the factors listed above for an actual driver. will fit the area.

Similarly, even in the case of sedation anesthesia, which is widely used in the medical field, it is well known under what conditions the probability of occurrence of respiratory depression, which is a representative side effect of sedation, is high. For example, the higher the dose of anesthetic agent administered, the more obese the subject, the shorter the neck length of the subject, the more smokers, the more a history of respiratory disease, the thicker the tongue, the more prone to airway pressure. The risk of respiratory depression may increase as (cervical vertebrae), a history of sleep apnea, a longer duration of sleep anesthesia, a male subject, an older subject, and a person with respiratory allergies.

Therefore, if the occurrence of side effects of sleep anesthesia, which is not easy to judge by human cognitive ability, can be predicted more accurately, medical accidents can be prevented and patient safety can be ensured.

Korean Patent Publication No. 10-2012-0049337 (2012.05.16)

An embodiment of the present invention provides an artificial intelligence-based sleep anesthesia safety prediction device and method that can predict the risk of respiratory depression by integrating and analyzing various triggering factors of side effects of sleep anesthesia and bio-vitality indicators measured in the course of sleep anesthesia want to

An embodiment of the present invention establishes a statistical probability-based learning model for predicting the occurrence of side effects based on artificial intelligence, and predicts the probability of occurrence of side effects of a subject under sleep anesthesia to notify the occurrence of risks by artificial intelligence-based sleep anesthesia An object of the present invention is to provide an apparatus and method for predicting safety.

In embodiments, the artificial intelligence-based sleep anesthesia safety prediction apparatus includes: a trigger factor definition unit defining a plurality of trigger factors related to the side effects of sleep anesthesia based on user information of a patient; a biological vitality collecting unit for collecting biological vitality indicators for the patient in the sleep anesthesia process of the patient; a learning model construction unit for performing machine learning on the correlation between the plurality of inducing factors and the bio-vitality index as independent variables and using the presence or absence of side effects as a dependent variable during the sleep anesthesia process; a side effect prediction unit for predicting the possibility of side effects occurring in a specific patient's sleep anesthesia process using the learning model generated as a result of the machine learning; and a safety level calculation unit for calculating the safety level of anesthesia for the specific patient based on the possibility of occurrence of the side effect.

The trigger factor definition unit may define the plurality of trigger factors, including the type and dose of the anesthetic agent, anesthesia time, personal information of the patient, body information, and life information.

The biological vitality collecting unit may collect at least one of blood pressure, respiration rate, heart rate, body temperature, electrocardiogram, and oxygen saturation of the patient as the biological vitality index.

The learning model building unit may include: a feature vector generating module that generates a feature vector related to the independent variable; a predictive vector obtaining module for obtaining a predictive vector including a probability of occurrence of a plurality of predefined side effects as a result of inputting the feature vector into a pre-established learning model; an error calculation module for calculating an error between an actual value of a dependent variable corresponding to the independent variable and the prediction vector; and a weight correction module for correcting the weight of the previously built learning model in a direction to reduce the error.

The error calculation module calculates the error between the prediction vectors after applying a genetic weight to each of the plurality of side effects constituting the dependent variable based on the patient's genetic information, and the weight correction module is configured to calculate each of the plurality of side effects The correction may be performed by differentially applying an error reduction rate based on the priority according to the dielectric weight.

The side effect prediction unit obtains the probability of occurrence for the plurality of side effects from the learning model and predicts the occurrence probability and expected time for respiratory depression by integrating the occurrence probability, and the safety calculation unit relates to the respiratory depression Based on the probability of occurrence and the expected time, it is possible to calculate the degree of safety under sedation at the present time.

The safety level calculator may provide an alarm regarding the occurrence of a risk when the safety level of sleep anesthesia is less than a preset value.

Among the embodiments, the method for predicting the safety level of sleep anesthesia based on artificial intelligence includes: defining a plurality of triggering factors related to the side effects of sleep anesthesia based on user information of a patient; collecting bio-vitality indicators for the patient during the sleep anesthesia process of the patient; performing machine learning on the correlation between the variables using the plurality of triggering factors and the bio-vitality index as independent variables and the occurrence of side effects in the sleep anesthesia process as a dependent variable; predicting the possibility of side effects occurring in the sleep anesthesia process of a specific patient by using the learning model generated as a result of the machine learning; and calculating the safety level of anesthesia for the specific patient based on the possibility of occurrence of the side effect.

The disclosed technology may have the following effects. However, this does not mean that a specific embodiment should include all of the following effects or only the following effects, so the scope of the disclosed technology should not be understood as being limited thereby.

The artificial intelligence-based sleep anesthesia safety prediction apparatus and method according to an embodiment of the present invention can predict the risk of respiratory depression by integrating various triggering factors of sleep anesthesia side effects and bio-vitality indicators measured during the sleep anesthesia process. .

Artificial intelligence-based sleep anesthesia safety prediction apparatus and method according to an embodiment of the present invention builds a statistical probability-based learning model to perform artificial intelligence-based side effect occurrence prediction, It can predict and inform the occurrence of danger.

1 is a view for explaining a safety prediction system according to the present invention.
FIG. 2 is a diagram for explaining the system configuration of the safety level prediction device of FIG. 1 .
FIG. 3 is a view for explaining the functional configuration of the safety level prediction device of FIG. 1 .
4 is a flowchart illustrating a method for predicting the safety level of sleep anesthesia based on artificial intelligence according to the present invention.
FIG. 5 is a view for explaining a functional configuration of the learning model building unit of FIG. 3 .
6 is a flowchart illustrating a learning model building process performed by the learning model building unit of FIG. 3 .
FIG. 7 is a view for explaining an embodiment of a feature vector generation process performed by the learning model building unit of FIG. 3 .

Since the description of the present invention is merely an embodiment for structural or functional description, the scope of the present invention should not be construed as being limited by the embodiment described in the text. That is, since the embodiment may have various changes and may have various forms, it should be understood that the scope of the present invention includes equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only such effects, it should not be understood that the scope of the present invention is limited thereby.

On the other hand, the meaning of the terms described in the present application should be understood as follows.

Terms such as “first” and “second” are for distinguishing one component from another, and the scope of rights should not be limited by these terms. For example, a first component may be termed a second component, and similarly, a second component may also be termed a first component.

When a component is referred to as being “connected to” another component, it may be directly connected to the other component, but it should be understood that other components may exist in between. On the other hand, when it is mentioned that a certain element is "directly connected" to another element, it should be understood that the other element does not exist in the middle. Meanwhile, other expressions describing the relationship between elements, that is, "between" and "between" or "neighboring to" and "directly adjacent to", etc., should be interpreted similarly.

The singular expression is to be understood as including the plural expression unless the context clearly dictates otherwise, and terms such as "comprises" or "have" refer to the embodied feature, number, step, action, component, part or these It is intended to indicate that a combination exists, and it should be understood that it does not preclude the possibility of the existence or addition of one or more other features or numbers, steps, operations, components, parts, or combinations thereof.

In each step, identification numbers (eg, a, b, c, etc.) are used for convenience of description, and identification numbers do not describe the order of each step, and each step clearly indicates a specific order in context. Unless otherwise specified, it may occur in a different order from the specified order. That is, each step may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

The present invention can be embodied as computer-readable codes on a computer-readable recording medium, and the computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored. . Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like. In addition, the computer-readable recording medium is distributed in a computer system connected to a network, so that the computer-readable code can be stored and executed in a distributed manner.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs, unless otherwise defined. Terms defined in the dictionary should be interpreted as being consistent with the meaning of the context of the related art, and cannot be interpreted as having an ideal or excessively formal meaning unless explicitly defined in the present application.

1 is a view for explaining a safety prediction system according to the present invention.

Referring to FIG. 1 , an artificial intelligence-based sleep anesthesia safety prediction system (hereinafter, safety level prediction system) 100 may include a user terminal 110 , a safety level prediction device 130 , and a database 150 . .

The user terminal 110 may correspond to a computing device capable of confirming the prediction result of sleep anesthesia safety level, and may be implemented as a smartphone, a notebook computer, or a computer, but is not necessarily limited thereto, and may be implemented in various devices such as a tablet PC can be The user terminal 110 may be connected to the safety level prediction apparatus 130 through a network, and a plurality of user terminals 110 may be simultaneously connected to the safety level prediction apparatus 130 to exchange data. Meanwhile, the user terminal 110 may function as a display by interworking with the safety level prediction device 130 if necessary. In this case, the user terminal 110 may interwork with the safety prediction device 130 through a dedicated program.

The safety prediction device 130 works with the database 150 to monitor the patient's status information in the course of anesthesia of the patient and predict the occurrence of side effects of sleep anesthesia, which corresponds to a computer or program that can accurately predict the safety of sleep anesthesia. It can be implemented as a server. Here, sedation anesthesia may correspond to anesthesia for maintaining the consciousness of the patient at the level of a sleep state. The safety prediction device 130 may be connected to the user terminal 110 through a wired network or a wireless network such as Bluetooth or WiFi, and may communicate with the user terminal 110 through a wired or wireless network.

In one embodiment, the safety level prediction device 130 may store information necessary for predicting the safety level of anesthesia by interworking with the database 150 . Meanwhile, unlike FIG. 1 , the safety level prediction device 130 may be implemented by including the database 150 therein. In addition, the safety prediction device 130 may be implemented including a processor, a memory, a user input/output unit, and a network input/output unit, which will be described in more detail with reference to FIG. 2 .

In one embodiment, the safety prediction device 130 may be implemented by including various sensors for identifying the state of the patient performing sedation. For example, the safety prediction device 130 interlocks with sensors for measuring the patient's blood pressure, body temperature, heart rate, respiration rate, oxygen saturation and electrocardiogram to obtain various biometric information during the sleep anesthesia process of the patient. there is.

In addition, the safety level prediction device 130 may obtain user information of the patient by interworking with an external system (not shown in the drawing). For example, the safety prediction device 130 may operate in conjunction with a health insurance system, an administrative system, or a hospital system, and may obtain personal information, body information, disease history, and accident history of a patient as necessary. there is.

The database 150 may store various pieces of information for the safety level prediction device 130 to predict the level of safety under sleep anesthesia in a patient's sleep anesthesia process. For example, the database 150 may store data about the patient's biometric information, and may store data about a learning dataset and a learning model for predicting sleep anesthesia safety, but is not necessarily limited thereto, and artificial intelligence Information collected or processed in various forms can be stored in the process of predicting the safety level of sleep anesthesia based on sleep anesthesia.

FIG. 2 is a diagram for explaining the system configuration of the safety level prediction device of FIG. 1 .

Referring to FIG. 2 , the safety prediction device 130 may be implemented including a processor 210 , a memory 230 , a user input/output unit 250 , and a network input/output unit 270 .

The processor 210 may execute a procedure for processing each step in the process of the safety prediction device 130 operating, and manage the memory 230 that is read or written throughout the process, and the memory ( 230) may schedule a synchronization time between the volatile memory and the non-volatile memory. The processor 210 may control the overall operation of the safety prediction device 130 , and is electrically connected to the memory 230 , the user input/output unit 250 , and the network input/output unit 270 to control data flow therebetween. can do. The processor 210 may be implemented as a central processing unit (CPU) of the safety prediction device 130 .

The memory 230 is implemented as a non-volatile memory, such as a solid state drive (SSD) or a hard disk drive (HDD), and may include an auxiliary storage device used to store overall data required for the safety prediction device 130, and , and may include a main memory implemented as a volatile memory such as random access memory (RAM).

The user input/output unit 250 may include an environment for receiving a user input and an environment for outputting specific information to the user. For example, the user input/output unit 250 may include an input device including an adapter such as a touch pad, a touch screen, an on-screen keyboard, or a pointing device, and an output device including an adapter such as a monitor or a touch screen. In an embodiment, the user input/output unit 250 may correspond to a computing device accessed through remote access, and in such a case, the safety prediction device 130 may be implemented as a server.

The network input/output unit 270 includes an environment for connecting with an external device or system through a network, for example, a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), and a VAN (Wide Area Network) (VAN). It may include an adapter for communication such as Value Added Network).

FIG. 3 is a view for explaining a functional configuration of the safety level prediction device of FIG. 1 .

Referring to FIG. 3 , the safety predicting device 130 includes a trigger factor definition unit 310 , a biological vitality collection unit 320 , a learning model construction unit 330 , a side effect prediction unit 340 , and a safety level calculation unit ( 350) and a control unit 360.

The trigger factor definition unit 310 may define a plurality of trigger factors related to the side effects of anesthesia based on the user information of the patient. Here, the triggering factor may correspond to the causative factors that cause side effects in the sleep anesthesia process. Inducing factors may be defined in various ways, including the patient's health status, conditions of sleep anesthesia, and the like, and representative factors known to have a major influence on the occurrence of side effects of sleep anesthesia are described as examples, but are not necessarily limited thereto.

In an embodiment, the trigger factor definition unit 310 may define a plurality of trigger factors including the type and dose of the anesthetic agent, anesthesia time, personal information of the patient, body information, and living information. More specifically, the triggering factors related to the patient's personal information may include gender, age, etc., and the triggering factors related to the patient's physical information include the patient's weight, neck length, tongue thickness, and possible airway security during surgery. It may include position, etc., and may include factors related to the patient's life information, such as whether smoking, a history of respiratory diseases such as tuberculosis, presence of sleep apnea or snoring, presence of respiratory allergies, and the like. The plurality of triggers defined by the trigger factor definition unit 310 may be measured and collected in various ways in tens of thousands of sleep anesthesia processes, and may be used as learning data for machine learning.

The biological vitality collecting unit 320 may collect the biological vitality index for the patient during the sleep anesthesia of the patient. Here, the bio-vitality index may correspond to a bio-signal showing a state of a body function that maintains life. In an embodiment, the bio-vitality collection unit 320 may collect at least one of blood pressure, respiration rate, heart rate, body temperature, electrocardiogram, and oxygen saturation of the patient as a bio-vitality index of the patient under sedation. The biological vitality collecting unit 320 may perform a pre-processing operation on the collected data in order to collect signals with higher accuracy. For example, the biological vitality collection unit 320 may convert data into a form more suitable for subsequent operations through preprocessing operations such as noise filtering, signal amplification, normalization, and resampling. there is.

The learning model building unit 330 may perform machine learning on the correlation between the variables by using a plurality of triggering factors and bio-vitality indicators as independent variables, and the occurrence of side effects in the sleep anesthesia process as a dependent variable. That is, the learning model building unit 330 learns various cases related to the occurrence of side effects in various combinations of a plurality of triggering factors and biological vitality indicators to predict the occurrence of respiratory depression side effects in a patient under sedation relatively accurately. A learning model can be built. On the other hand, a learning algorithm for machine learning can be used in various ways for each situation or purpose, and representatively, a support vector machine (SVM), a long short-term memory (LSTM), an autoencoder, etc. may be used.

In one embodiment, the learning model building unit 330 calculates the probability of occurrence of a plurality of side effects defined as a result of inputting the feature vector generating module and the feature vector to the pre-built learning model for generating a feature vector related to the independent variable. A prediction vector acquisition module for acquiring a predictive vector including an error calculation module for calculating an error between an actual value of a dependent variable corresponding to an independent variable and a predictive vector, and a weight for correcting the weight of a pre-established learning model in the direction of reducing the error A calibration module may be included. This will be described in more detail with reference to FIGS. 5 and 6 .

The side effect prediction unit 340 may predict the possibility of side effects occurring in the sleep anesthesia process of a specific patient by using a learning model generated as a result of machine learning. The side effect prediction unit 340 may predict the possibility of side effects by using the results of inputting the biosignals collected from the patient under sedation, the patient's personal information, and life information into the learning model. That is, the safety prediction device 130 provides the effect of increasing the safety of sedation anesthesia and ensuring the safety of the patient by predicting the occurrence of side effects in advance through the side effect prediction unit 340 and notifying them to prepare a response therefor. can do.

In one embodiment, the side effect prediction unit 340 obtains the probability of occurrence for a plurality of side effects from the learning model and integrates the probability of occurrence to predict the occurrence probability and expected timing of respiratory depression, respectively. Side effects of sedation anesthesia may include respiratory depression, cardiac arrest, headache, vomiting, etc., and the learning model may generate the probability of occurrence for each side effect under input conditions and provide it as an output. The side effect prediction unit 340 may integrate each occurrence probability of a plurality of side effects into one using various methods.

For example, the side effect prediction unit 340 may determine the occurrence probability of respiratory depression through the average of the occurrence probabilities, or determine the occurrence probability of respiratory depression through the average after excluding the highest or lowest occurrence probability. . In addition, the side effect prediction unit 340 may determine the occurrence probability related to respiratory depression by applying a weight differentially applied to each occurrence probability for a plurality of side effects based on the correlation with the respiratory depression side effect and integrating them into one. .

The safety level calculation unit 350 may calculate the safety level of sleep anesthesia of a specific patient based on the possibility of occurrence of side effects. The degree of safety under sedation may correspond to a numerical value related to safety that is calculated by considering the probability of side effects occurring in the course of sedation. The safety level calculation unit 350 may calculate the degree of safety under sleep anesthesia by reflecting the bio-vitality index of the patient at the current time based on the possibility of occurrence of side effects derived through the learning model. For example, the safety level calculator 350 may quantify the safety level of anesthesia by applying a weight w according to the probability of occurrence of side effects p and the patient's bio-vitality index. At this time, the weight w according to the biological vitality index is a weighted score for each grade derived after grading the measured values of various biological information constituting the biological vitality index, for example, body temperature, heart rate, respiration rate, blood pressure, etc. for each biological information. It can be calculated as a result of integrating

In one embodiment, the safety level calculation unit 350 may calculate the safety level of sleep anesthesia at the current time based on the probability of occurrence and the expected time for respiratory depression. The safety level calculation unit 350 may calculate the sedation anesthesia safety level using the result predicted by the side effect prediction unit 340, and the sedation anesthesia safety level increases as the probability of occurrence of respiratory depression increases and the predicted time increases can be highly calculated.

In one embodiment, the safety level calculator 350 may provide an alarm regarding the occurrence of a risk when the safety level of anesthesia is less than a preset value. The safety level calculator 350 may predict the occurrence of a risk of side effects for the patient based on the safety level of sedation, and may provide an alarm when the risk to the patient is high compared to a preset value. In this case, the alarm may be implemented and delivered as visual, auditory, or tactile information.

For example, the alarm may be provided in a manner that reproduces a predetermined sound, outputs a predetermined light source, or generates a predetermined vibration. On the other hand, the safety level prediction device 130 may be implemented to include a display for outputting related information, and the safety level calculation unit 350 may output information about the alarm through various implemented interfaces in conjunction with the display. may be

The control unit 360 controls the overall operation of the safety prediction device 130 , and the trigger factor definition unit 310 , the biological vitality collection unit 320 , the learning model construction unit 330 , the side effect prediction unit 340 and A control flow or data flow between the safety level calculators 350 may be managed.

4 is a flowchart illustrating a method for predicting the safety level of sleep anesthesia based on artificial intelligence according to the present invention.

Referring to FIG. 4 , the safety prediction device 130 may define a plurality of trigger factors related to the side effects of sedation based on the user information of the patient through the trigger factor definition unit 310 (step S410 ). The safety prediction device 130 may collect the bio-vitality index for the patient during the sleep anesthesia process of the patient through the bio-vitality collection unit 320 (step S430). The safety prediction device 130 uses a plurality of triggering factors and bio-vitality indicators as independent variables through the learning model building unit 330, and the occurrence or non-occurrence of side effects in the sleep anesthesia process as a dependent variable to determine the correlation between the variables. machine learning may be performed (step S450).

Also, the safety prediction device 130 may predict the possibility of side effects occurring in the sleep anesthesia process of a specific patient by using a learning model generated as a result of machine learning through the side effect prediction unit 340 (step S470). The safety level prediction device 130 may calculate the sleep anesthesia safety level of a specific patient based on the possibility of occurrence of side effects through the safety level calculation unit 350 (step S490).

FIG. 5 is a view for explaining a functional configuration of the learning model building unit of FIG. 3 .

Referring to FIG. 5 , the learning model building unit 330 includes a feature vector generation module 510 , a prediction vector acquisition module 530 , an error calculation module 550 , a weight correction module 570 , and a control module 590 . may include

The feature vector generation module 510 may generate a feature vector for an independent variable. The independent variable may include a plurality of inducing factors and bio-vitality indicators, and the feature vector may be generated based on parameters derived from the independent variables. The plurality of triggering factors may be expressed numerically as values within a specific range, and the bio-vitality index may be expressed as a value converted based on a bio-signal. That is, the feature vector may be defined as a set of values corresponding to each of the independent variables.

On the other hand, some of the plurality of triggering factors and bio-vitality indicators may be converted into numerical values through image data transformation. In FIG. 7 , a biosignal may be measured as a time series value that is continuous according to time through a measurement sensor, and may be expressed as imaged data such as a two-dimensional graph according to a reference time (a). A two-dimensional graph can be converted into image information by a predetermined procedure (b). For example, the two-dimensional graph may be modified to a size that minimizes operation while having image features before learning. The converted image information may be converted into a numeric vector through encoding (c). The numeric vector may be normalized to a value within a predetermined range as necessary to generate a final feature vector (d).

The predictive vector obtaining module 530 may obtain a predictive vector including the occurrence probability of a plurality of predefined side effects as a result of inputting the feature vector into the pre-constructed learning model. The learning model may use a feature vector as input data and generate a predictive vector as output data, and in the process, dimensionality reduction between the feature vector and the predictive vector may occur. That is, the number of dimensions of the prediction vector may correspond to the number of side effects to be predicted, and may be preset through the safety prediction device 130 .

The error calculation module 550 may calculate an error between the actual value of the dependent variable corresponding to the independent variable and the prediction vector. The error calculation module 550 may calculate the error by comparing the information predicted by the learning model with the occurrence of side effects collected during the sleep anesthesia process of the actual patient. Errors can be calculated for each side effect corresponding to the dependent variable, and the errors for each side effect can be integrated into one and converted into information about the overall error rate.

In an embodiment, the error calculation module 550 may calculate an error between prediction vectors after applying a genetic weight to each of a plurality of side effects constituting the dependent variable based on the patient's genetic information. That is, when the occurrence probability of a specific side effect is high according to the characteristics of the genetic information, the error calculation module 550 may calculate an error for the prediction result after applying the genetic weight of the specific side effect to be relatively higher than that of other side effects. For example, in the case of a patient having genetic information with a high probability of occurrence in the order of the first side effect, the third side effect, and the second side effect, the error calculation module 550 sets the first side effect, the third side effect, and the second side effect in the order of the highest. Genetic weighting can be applied. Through this, the learned learning model can generate a more accurate prediction result by reflecting the difference in the occurrence probability according to the patient's genetic information.

In another embodiment, the error calculation module 550 may determine a user group having similar genetic information based on the patient's genetic information, and assign a genetic weight for the user group to each of a plurality of side effects constituting the dependent variable. After application, the error between the prediction vectors can be calculated.

The weight correction module 570 may correct the weight of the previously built learning model in a direction to reduce an error. In an embodiment, the weight correction module 570 may perform weight correction by differentially applying an error reduction rate to each of the plurality of side effects based on the priority according to the genetic weight. That is, the weight correction module 570 may reduce the prediction uncertainty due to the genetic factor by applying a higher error reduction rate centering on the high-priority dielectric weight.

The control module 590 controls the overall operation of the learning model building unit 330, and the feature vector generation module 510, the prediction vector acquisition module 530, the error calculation module 550, and the weight correction module 570) It can manage the control flow or data flow between them.

6 is a flowchart illustrating a learning model building process performed by the learning model building unit of FIG. 3 .

Referring to FIG. 6 , the learning model building unit 330 may generate a feature vector related to an independent variable through the feature vector generating module 510 (step S610). The learning model building unit 330 may obtain a predictive vector including the occurrence probability of a plurality of predefined side effects as a result of inputting the feature vector into the pre-built learning model through the predictive vector obtaining module 530 ( step S630).

Also, the learning model building unit 330 may calculate an error between the actual value of the dependent variable corresponding to the independent variable and the prediction vector through the error calculation module 530 (step S650 ). The learning model building unit 330 may correct the weight of the previously built learning model in a direction to reduce an error through the weight correction module 570 (step S670).

Although the above has been described with reference to the preferred embodiment of the present invention, those skilled in the art can variously modify and change the present invention within the scope without departing from the spirit and scope of the present invention as described in the claims below. You will understand that it can be done.

100: safety prediction system
110: user terminal 130: safety prediction device
150: database
210: processor 230: memory
250: user input/output unit 270: network input/output unit
310: trigger factor definition unit 320: vitality collection unit
330: learning model building unit 340: side effect prediction unit
350: safety level calculation unit 360: control unit
510: feature vector generation module 530: prediction vector acquisition module
550: error calculation module 570: weight correction module
590: control module

Claims (8)

a trigger factor definition unit defining a plurality of trigger factors related to the side effects of anesthesia based on the user information of the patient;
a biological vitality collection unit for collecting biological vitality indicators for the patient in the sleep anesthesia process of the patient;
a learning model construction unit for performing machine learning on the correlation between the plurality of inducing factors and the biological vitality index as independent variables and using the occurrence of side effects in the sleep anesthesia process as a dependent variable;
a side effect prediction unit for predicting the possibility of side effects occurring in a specific patient's sleep anesthesia process using the learning model generated as a result of the machine learning; and
An artificial intelligence-based sleep anesthesia safety prediction device including a safety level calculation unit for calculating the sleep anesthesia safety level of the specific patient based on the possibility of occurrence of side effects.
The method of claim 1, wherein the trigger factor definition unit
Artificial intelligence-based sleep anesthesia safety prediction device, characterized in that the plurality of triggering factors are defined, including type and dose of anesthetic agent, sleep anesthesia time, patient's personal information, body information, and life information.
According to claim 1, wherein the biological vitality collecting unit
Artificial intelligence-based sleep anesthesia safety prediction device, characterized in that collecting at least one of blood pressure, respiration rate, heart rate, body temperature, electrocardiogram, and oxygen saturation of the patient as the bio-vitality index.
According to claim 1, wherein the learning model building unit
a feature vector generation module that generates a feature vector for the independent variable;
a predictive vector obtaining module for obtaining a predictive vector including a probability of occurrence of a plurality of predefined side effects as a result of inputting the feature vector into a pre-established learning model;
an error calculation module for calculating an error between an actual value of a dependent variable corresponding to the independent variable and the prediction vector; and
Artificial intelligence-based sleep anesthesia safety prediction device, characterized in that it comprises a weight correction module for correcting the weight of the pre-established learning model in a direction to reduce the error.
5. The method of claim 4,
The error calculation module calculates the error between the prediction vectors after applying a genetic weight to each of a plurality of side effects constituting the dependent variable based on the patient's genetic information,
The weight correction module performs the correction by differentially applying an error reduction rate based on the priority according to the genetic weight for each of the plurality of side effects. .
The method of claim 1, wherein the side effect prediction unit
Obtaining the occurrence probability for the plurality of side effects from the learning model and integrating the occurrence probability to predict the occurrence probability and expected timing of respiratory depression, respectively,
The safety level calculation unit artificial intelligence-based sleep anesthesia safety prediction device, characterized in that for calculating the degree of safety of sleep anesthesia at the current time based on the probability of occurrence and the expected time of the respiratory depression.
According to claim 1, wherein the safety calculation unit
Artificial intelligence-based sleep anesthesia safety prediction device, characterized in that it provides an alarm regarding the occurrence of a risk when the degree of safety under sleep anesthesia is less than a preset value.
Defining a plurality of trigger factors related to the side effects of anesthesia based on the user information of the patient;
collecting bio-vitality indicators for the patient during the sleep anesthesia of the patient;
performing machine learning on the correlation between the variables using the plurality of triggering factors and the bio-vitality index as independent variables and the occurrence of side effects in the sleep anesthesia process as a dependent variable;
predicting the possibility of side effects occurring in the sleep anesthesia process of a specific patient by using the learning model generated as a result of the machine learning; and
Artificial intelligence-based sedation anesthesia safety prediction method comprising the step of calculating the sedation anesthesia safety level of the specific patient based on the possibility of occurrence of the side effects.

Images