JP2023106967A - Program, method for processing information, information processor, and method for generating model - Google Patents

Abstract

To provide a program which can present a recovery treatment to do on a patient who has become sicker after taking an administered medicine.SOLUTION: A computer acquires the vital data of a patient who has become in bad shape after an injection was administered or being administered to the patient and the type of the injection and the amount and the rate of the dosage according to a program. When the type of the injection, the amount and the rate of the dosage, the vital data are input, the computer inputs the acquired type of the injection and the acquired amount and rate of the dosage and the acquired vital data into a learning model which has learned to output advice information on a recovery treatment and outputs the advice information on the recovery treatment on the patient with the vital data.SELECTED DRAWING: Figure 1

Description

The present invention relates to a program, an information processing method, an information processing apparatus, and a model generation method.

2. Description of the Related Art A system has been proposed that assists a doctor in diagnosing a patient's condition in a medical institution or the like. For example, Patent Document 1 discloses a system for estimating a possible disease diagnosis or a patient's condition based on clinical information including patient examination records, nursing records, test results, medication records, medical images, and the like. there is The system disclosed in Patent Document 1 presumes and presents a patient's diagnosis based on the contents of a drug prescription and an injection order given to the patient, and also compares the diagnosis and the drug prescription. and the content of the injection order, outputting a warning when the ordered drug is contraindicated for the disease condition of the diagnosis name.

Japanese Unexamined Patent Application Publication No. 2007-18460

A wide variety of drugs are handled at medical institutions, etc., and the dosage and administration rate differ depending on the drug. Great care must be taken to prevent this from occurring. If the type, dosage, or rate of administration of the drug is incorrect, the patient may be in a serious condition, so accurate work is required of the medical staff, which places a heavy burden on the medical staff. In addition, if the wrong type, dosage, or administration rate of the drug is administered, and the patient's condition worsens afterward, it is necessary to take measures to recover, but medical professionals such as doctors and nurses It is difficult to make appropriate judgments when the experience and knowledge of a person are poor. The system disclosed in Patent Document 1 estimates the diagnosis from the drug ordered for the patient by associating the diagnosis name or the patient's condition with the name of the drug that can be used or the name of the contraindicated drug. In addition, it merely determines whether or not the ordered drug is contraindicated, and cannot suggest the treatment to be performed for the patient whose condition has deteriorated after the drug is administered.

An object of one aspect of the present invention is to provide a program or the like capable of suggesting recovery measures to be taken to a patient whose condition has deteriorated after administration.

A program according to one aspect acquires vital data of a patient who is in an abnormal state during or after administration of an injection, and the type, dosage, and administration rate of the injection, and determines the type of injection, administration The acquired type of injection, dosage and administration rate, and vital data are input to a learning model trained to output advice information regarding recovery treatment when the amount and administration rate, and vital data are input. The computer is caused to execute a process of inputting and outputting advice information regarding recovery treatment for the patient of the vital data.

In one aspect, remedial actions to be taken can be suggested to a patient whose condition has deteriorated after administration.

It is a block diagram which shows the structural example of an information processing apparatus. It is a schematic diagram which shows the structural example of electronic medical record DB. FIG. 4 is a schematic diagram showing a configuration example of a learning model; FIG. 11 is a flowchart showing an example of a learning model generation processing procedure; FIG. FIG. 11 is a flowchart illustrating an example of a procedure for presenting a recovery action; FIG. FIG. 5 is an explanatory diagram showing an example of a screen; FIG. 10 is a schematic diagram showing a configuration example of a learning model according to Embodiment 2; FIG. 11 is a flow chart showing an example of a procedure for presenting a recovery action according to the second embodiment; FIG. FIG. 5 is an explanatory diagram showing an example of a screen; FIG. 16 is a flowchart showing an example of a procedure for presenting a recovery action according to the third embodiment; FIG. FIG. 5 is an explanatory diagram showing an example of a screen; FIG. 4 is an explanatory diagram showing a configuration example of a learning model used for patient condition determination; FIG. 16 is a flow chart showing an example of a procedure for presenting a recovery action according to the fourth embodiment; FIG. FIG. 5 is an explanatory diagram showing an example of a screen;

Hereinafter, the program, the information processing method, the information processing apparatus, and the model generation method of the present disclosure will be described in detail based on the drawings showing the embodiments.

(Embodiment 1)
Based on the vital data of the patient whose vital data became abnormal during or after administration of the injection, and medication information including the type of injection administered (drug name), dosage, and administration rate, An information processing apparatus that presents advice on measures for recovery (recovery measures) will be described. In this embodiment, for an injection used in a drug administration method in which a drug is administered directly into a blood vessel using, for example, an infusion pump or a syringe pump, recovery measures to be taken for a patient whose condition has deteriorated during or after drug administration are described. Present. Injections are not limited to drugs that are directly administered into blood vessels, and may be drugs that are administered intracutaneously, subcutaneously, or into body tissues such as muscles using an injection needle. Further, the deterioration of the condition means that the blood pressure, pulse rate, heart rate, respiratory rate, electrocardiogram, or body temperature included in the vital data deviate from the normal range, for example. It should be noted that the normal range is set in advance according to, for example, age, sex, weight, height, medical condition, and the like.

FIG. 1 is a block diagram showing a configuration example of an information processing apparatus. The information processing device 10 is a device capable of various information processing and transmission/reception of information, and is, for example, a server computer, a personal computer, or the like. A plurality of information processing apparatuses 10 may be provided to perform distributed processing, or may be realized by a plurality of virtual machines provided in one apparatus. The information processing apparatus 10 is installed and used in, for example, a medical institution. When the information processing device 10 is configured by a server computer, the information processing device 10 may be a local server installed in a medical institution, or may be a cloud server connected for communication via a network such as the Internet. good.

The information processing apparatus 10 performs recovery for recovering the post-medication state of the patient, based on the medication information regarding the drug (injection) being administered to the patient or after administration, and the patient's vital data before and after the drug administration. A process of outputting advice information for advising a treatment is performed. Specifically, as will be described later, the information processing apparatus 10 performs machine learning to learn predetermined training data, inputs medication information and vital data, and generates a learning model 12M (see FIG. 3) that outputs advice information. Prepared in advance. The information processing device 10 inputs medication information and vital data of a patient whose condition has deteriorated during or after medication to the learning model 12M, and acquires advice information presenting treatment for recovering the condition from the learning model 12M.

The information processing apparatus 10 includes a control section 11, a storage section 12, a communication section 13, an input section 14, a display section 15, a reading section 16, etc. These sections are interconnected via a bus. The control unit 11 has one or more processors such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit), or an AI chip (AI semiconductor). The control unit 11 appropriately executes the control program 12P stored in the storage unit 12 to perform various information processing, control processing, etc. that the information processing apparatus 10 should perform.

The storage unit 12 includes a RAM (Random Access Memory), flash memory, hard disk, SSD (Solid State Drive), and the like. The storage unit 12 stores in advance a control program 12P (program product) executed by the control unit 11 and various data necessary for executing the control program 12P. The storage unit 12 also temporarily stores data and the like generated when the control unit 11 executes the control program 12P. The storage unit 12 also stores a learning model 12M that has learned training data by, for example, machine learning. The learning model 12M is a model that has been learned to output advice information including recovery measures for recovering the post-medication state when the patient's medication information and vital data are input. The learning model 12M is assumed to be used as a program module that constitutes artificial intelligence software. The learning model 12M performs a predetermined calculation on an input value and outputs the calculation result. In the storage unit 12, data such as coefficients of functions and threshold values that define this calculation are stored as the learning model 12M. be done. The storage unit 12 also stores an electronic medical chart DB 12a, which will be described later. The electronic medical record DB 12a may be stored in another storage device connected to the information processing device 10, or may be stored in another storage device with which the information processing device 10 can communicate.

The communication unit 13 is a communication module for connecting to a network such as the Internet or a LAN (Local Area Network) by wired communication or wireless communication, and transmits and receives information to and from another device via the network. The input unit 14 receives an operation input by the user and sends a control signal corresponding to the content of the operation to the control unit 11 . The display unit 15 is a liquid crystal display, an organic EL display, or the like, and displays various information according to instructions from the control unit 11 . The input unit 14 and the display unit 15 may be a touch panel integrally configured.

The reading unit 16 reads information stored in a portable storage medium 1a including CD (Compact Disc)-ROM, DVD (Digital Versatile Disc)-ROM, USB (Universal Serial Bus) memory, SD (Secure Digital) card, etc. read. The control program 12</b>P (program product) and various data stored in the storage unit 12 may be read from the portable storage medium 1 a by the control unit 11 via the reading unit 16 and stored in the storage unit 12 . The control program 12P and various data stored in the storage unit 12 may be downloaded from another device by the control unit 11 via the communication unit 13 and stored in the storage unit 12 .

FIG. 2 is a schematic diagram showing a configuration example of the electronic medical chart DB 12a. The electronic medical chart DB 12a is a database that stores electronic medical chart data (medical records) of patients who use medical institutions. The electronic medical chart DB 12a shown in FIG. 2 includes a patient ID column, a patient information column, a vital data column, a medication information column, and the like. The patient ID column stores identification information (patient ID, patient code) for identifying each patient. The patient ID may be, for example, a patient registration card number issued by a medical institution. The patient information column stores patient information about the patient in association with the patient ID. The patient information includes, for example, attribute information including patient age and sex, diagnosis name, medical history, height and weight, and the like. The vital data column stores vital data measured from the patient in association with the patient ID. Vital data includes, for example, blood pressure, pulse rate, heart rate, respiration rate, electrocardiogram, body temperature, etc. These data are stored in respective columns in association with, for example, measurement date and time. In addition, each data of the vital data may be represented by a graph showing a change over time, and the graph data is, for example, a series of data recorded each time measurement is performed, including a pair of measurement date and time and measured value. be. The medication information column stores medication information related to an injection that has already been administered to a patient or an injection that is scheduled to be administered (medicine), in association with the patient ID. Medication information includes, for example, date and time of administration, name of drug (type of drug), dosage, administration rate, progress during or after administration, recovery measures taken when the progress has deteriorated, and the like. include. Progressive status includes normal status (good) and non-normal status. An abnormal state is, for example, a state in which vital data (blood pressure, pulse rate, heart rate, respiratory rate, electrocardiogram, body temperature, etc.) deviates from the normal range after the start of medication, such as an increase in blood pressure above the normal range. or below the normal range, heart rate above the normal range or below the normal range, respiratory rate above the normal range or below the normal range, abnormal electrocardiogram waveform, body temperature It includes an increase above the normal range or a decrease below the normal range. Recovery measures include, for example, discontinuing (interrupting) an ongoing medication, increasing or decreasing the dose of an ongoing medication, increasing or decreasing the rate of administration of an ongoing medication, administering an antagonist to an ongoing medication. , administering another drug, etc.

The contents stored in the electronic medical chart DB 12a are not limited to the example shown in FIG. Various information about the patient may be stored, such as comments entered by the patient. The examination information includes X-ray images, ultrasound images, CT (Computed Tomography) images, MRI (Magnetic Resonance Imaging) images, PET (Positron Emission Tomography) ) images.

FIG. 3 is a schematic diagram showing a configuration example of the learning model 12M. A learning model 12M shown in FIG. Based on this, calculations are performed to determine recovery measures to be taken for the patient, and the results of the calculations (advice information indicating recovery measures) are learned to be output. The learning model 12M may be configured such that patient medication information stored in the electronic medical record DB 12a and data included in the electronic medical record data are input. The learning model 12M is composed of a CNN (Convolutional Neural Network), which is a neural network model generated by deep learning, for example. In addition to CNN, the learning model 12M is configured using algorithms such as RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), Transformer, Decision Tree, Random Forest, SVM (Support Vector Machine), and the like. It may be configured by combining a plurality of algorithms. The information processing apparatus 10 performs machine learning for learning predetermined training data to generate a learning model 12M in advance. Then, the information processing apparatus 10 inputs the patient's medication information registered in the electronic medical record DB 12a and the vital data before and after the medication to the learning model 12M for the patient whose condition deteriorated during or after the medication. Predict remedial actions to be performed on the patient based on the output information. The learning model 12M has a plurality of input nodes, each input node is associated with information to be input, and an input associated with each information of medication information and vital data. It is input to the learning model 12M via the node.

The learning model 12M shown in FIG. 3 has a plurality of output nodes, and each output node is associated with a preset recovery action. It outputs the probability that it should be determined as the action to be taken. Recovery measures include discontinuing administration of the drug being administered, increasing the dose of the drug being administered, increasing or decreasing the rate of administration of the drug being administered, administering an antagonist to the drug being administered (antagonist A1 , dose B1, administration rate C1), administering another drug (drug A3, administration at dose B3), and the like. In the example shown in FIG. 3, output node 0 outputs the probability to be determined that the medication is stopped, output node 1 outputs the probability that it is determined that the dosage is increased, and output node 2 outputs the medication rate. When it is raised, it outputs the probability to be discriminated. Recovery measures associated with each output node are not limited to the example shown in FIG. 3, and include recovery measures to be taken when any vital data deviates from the normal range. The output value of each output node is, for example, a value between 0 and 1.0, and the sum of the discrimination probabilities output from each output node is 1.0 (100%). With the above-described configuration, the learning model 12M of this embodiment, when the patient's medication information and vital data are input, is related to recovery measures for recovering the state (deteriorated state) indicated by the input vital data. Output information (discrimination probability for each recovery action).

In the learning model 12M described above, the information processing apparatus 10 selects, for example, a predetermined number (for example, three) of output values (discrimination probabilities) in descending order of the output values from each output node, and selects the selected output value. The recovery action associated with the outputted output node is specified as the action to be implemented. As a result, a predetermined number of recovery measures can be presented as candidates for recovery measures to be implemented. Note that the information processing apparatus 10 assigns the recovery action associated with the output node that outputs the maximum output value (discrimination probability) among the output values from the output nodes of the learning model 12M to the action to be executed. may be specified. Further, the learning model 12M may have one output node that outputs the recovery action with the highest discrimination probability instead of having a plurality of output nodes that output the discrimination probability for each recovery action.

The learning model 12M machine-learns an unlearned learning model using training data including medication information and vital data for training, and information (correct label) indicating a recovery treatment that has been performed and has an effect. generated by The training data includes medication information and vital data before and after administration of patients whose conditions deteriorated during or after administration of various injections, and information indicating effective recovery measures. is given and generated. The medication information, vital data, and effective recovery treatment can be obtained from the medication information, vital data, and recovery treatment of the patient who has been medicated stored in the electronic medical chart DB 12a. Specifically, among the medication information stored in the electronic medical record DB 12a, the medication information of a patient whose progress has deteriorated and whose vital data has returned to the normal range after implementation of recovery treatment, and vital data before and after the medication. , the performed recovery actions are retrieved to generate training data.

When training medication information and vital data are input to the learning model 12M, the output value from the output node corresponding to the recovery action indicated by the correct label approaches 1.0, and the output value from the other output nodes is learned to approach 0.0. In the learning process, the learning model 12M performs calculations based on the input medication information and vital data, and calculates output values from each output node. Then, the learning model 12M calculates a value corresponding to the calculated output value of each output node and the correct label (1 for the output node corresponding to the recovery action indicated by the correct label, and 0 for other output nodes). are compared, and the parameters used in the arithmetic processing are optimized so that each output value approximates a value corresponding to the correct label. The parameters are weights between neurons in the learning model 12M, and the like. The parameter optimization method is not particularly limited, but error backpropagation, steepest descent method, or the like can be used. As a result, the learning model 12M that has been learned to predict the recovery treatment to be performed on the patient when the patient's medication information and vital data is input and to output the prediction result is obtained.

Learning of the learning model 12M may be performed by another learning device. A trained learning model 12M generated by learning in another learning device is downloaded from the learning device to the information processing device 10 via, for example, a network or via the portable storage medium 1a, and stored in the storage unit 12. . The learning model 12M is not limited to the configuration shown in FIG. The learning model 12M can be configured to receive various types of information used when predicting recovery actions to be taken. For example, any one or more of each information such as height and weight, medical history, examination information, treatment information, surgery information, medication history, comments entered by doctors etc. contained in electronic medical record data may be input. good. In this case, the learning model 12M discriminates the recovery action to be performed based on not only the medication information and vital data but also various information included in the electronic medical record data, and outputs the discrimination probability for each recovery action. Configured. The electronic medical record data input to the learning model 12M may be time-series data, for example, time-series data (test information) of test results of regularly performed blood tests or urine tests. It may be time-series data (treatment information) on various treatments performed. In this case, for example, the patient's electronic medical record data for the most recent day, electronic medical record data for the most recent week, or the like can be configured to be input to the learning model 12M.

Processing for learning the training data and generating the learning model 12M will be described below. FIG. 4 is a flow chart showing an example of a processing procedure for generating the learning model 12M. The following processing is performed by the control unit 11 of the information processing device 10 according to the control program 12P stored in the storage unit 12, but may be performed by another learning device.

First, the control unit 11 of the information processing device 10 acquires patient information to be used for training data from the electronic medical record DB 12a, generates training data, and uses the generated training data to learn the learning model 12M. For example, the control unit 11 determines whether there is a patient whose condition has deteriorated after starting medication (during or after medication) based on the progress of medication information for each patient stored in the electronic medical record DB 12a. (S11). Specifically, the control unit 11 controls vital data (blood pressure, pulse rate, heart rate, respiration rate, electrocardiogram , body temperature, etc.) is outside the normal range. If it is determined that there is a patient whose condition deteriorated after the start of medication (S11: YES), the control unit 11 stores the patient's medication information, vital data before and after medication, and recovery measures taken after the condition worsened. The indicated information is obtained from the electronic medical chart DB 12a (S12). Note that the control unit 11 acquires vital data before and after drug administration, for example, vital data from a predetermined time (several minutes to several tens of minutes) before the start of drug administration until the patient recovers from aggravated condition. The control unit 11 generates training data by adding a correct label indicating the acquired recovery treatment to the acquired medication information and vital data, and stores the training data in the storage unit 12 (S13).

The control unit 11 determines whether or not there is a patient (unprocessed patient) for whom training data generation processing has not been performed among the patients whose condition has deteriorated after the start of medication (S14). If it is determined that there is an unprocessed patient (S14: YES), the control unit 11 returns to the process of step S12, and electronically stores the unprocessed patient's medication information, pre- and post-medication vital data, and information indicating recovery measures. It is acquired from the chart DB 12a (S12) and stored in the storage unit 12 as training data (S13). The control unit 11 repeats the processing of steps S12 to S14 until it determines that there is no untreated patient. As a result, training data used for learning of the learning model 12M can be generated and accumulated based on the information of the patient whose condition has deteriorated after the start of drug administration.

If it is determined that there is no patient whose condition has deteriorated after the start of drug administration (S11: NO), the control unit 11 terminates the process without performing the above-described process. For a patient whose condition has deteriorated after the start of medication, medication information, vital data before and after medication, and information on recovery measures taken are collected in advance, and based on the collected data, the control unit 11 performs learning. A configuration for generating training data used for learning the model 12M may also be used.

When it is determined that there is no unprocessed patient (S14: NO), the control unit 11 performs learning processing of the learning model 12M using each accumulated training data (S15). Here, the control unit 11 inputs medication information and vital data included in the training data to the learning model 12M, and acquires output values from each output node. The control unit 11 sets the output value of each output node and the value corresponding to the correct label (1 for the output node corresponding to the correct recovery action and 0 for the output node corresponding to the incorrect recovery action). ) are compared, and parameters such as weights between neurons in the learning model 12M are optimized using, for example, error backpropagation so that the two approximate each other.

The control unit 11 determines whether or not there is training data (unprocessed data) for which learning processing has not been performed among the training data accumulated in step S13 (S16). If it is determined that there is unprocessed data (S16: YES), the control unit 11 returns to the process of step S15 and repeats the learning process using the unprocessed training data. If it is determined that there is no unprocessed data (S16: NO), the control section 11 terminates the series of processes.

Through the above-described processing, the learning model 12M that has been learned to output information indicating recovery treatment to be performed on the patient when the patient's medication information and vital data is input is generated. The learning model 12M can be further optimized by repeating the learning process using the training data as described above. Also, the learning model 12M that has already been trained can be re-learned by performing the above-described processing, and in this case, the learning model 12M with higher discrimination accuracy can be generated. Further, for example, by learning training data for each medical institution, the learning model 12M corresponding to each medical institution may be generated. Since the recovery treatment that can be performed on the patient differs somewhat depending on the medical equipment or drugs that can be used at the medical institution, by generating the learning model 12M for each medical institution, the recovery treatment that is frequently performed at each medical institution can be performed. Treatment considerations allow for the determination of remedial actions to be performed on the patient.

The following describes the process of determining recovery measures to be taken for a patient and outputting advice when, for example, the condition of a patient who is taking medication worsens, using the learning model 12M generated by the above-described process. . FIG. 5 is a flow chart showing an example of a procedure for presenting recovery measures, and FIG. 6 is an explanatory diagram showing an example of a screen. The following processing is performed by the control unit 11 of the information processing device 10 according to the control program 12P stored in the storage unit 12. FIG.

When the patient's condition deteriorates during or after administering the drug to the patient in the medical field, the doctor or nurse can enter the drug during or after administration via the input screen as shown in FIG. By inputting the medication information (drug name, dosage, administration rate) about the medication, advice on recovery measures to be taken can be received. Medication to a patient is performed in, for example, an operating room, an ICU (intensive care unit), a hospital room, a treatment room, or the like. For the sake of simplicity of explanation, a configuration in which the information processing device 10 is provided in a room where medication is administered will be described below as an example. A terminal having an input unit and a display unit may be provided in the room where medication is administered, and the terminal and the information processing apparatus 10 in the medical institution may be connected for communication. In this case, the medication information input via the terminal is transmitted to the information processing device 10, the recovery treatment predicted by the information processing device 10 based on the medication information is transmitted to the terminal, and the doctor Or presented to the nurse.

The control unit 11 of the information processing device 10 displays an input screen as shown in FIG. 6A on the display unit 15, for example, according to an operation by a doctor or a nurse via the input unit 14 (S21). The screen shown in FIG. 6A includes an input field for a patient ID, input fields for a drug name, a dose, and an administration rate as medication information, and an OK button for specifying recovery measures based on the entered medication information. and are provided. Each entry field may be configured to allow entry of arbitrary information or numerical values, or may be provided with a pull-down menu for selecting an arbitrary one from a plurality of options. In addition to the configuration shown in FIG. 6A, the input screen may be provided with an input field for inputting the manufacturer of the drug. , may be configured to display a list of drug names starting with the entered letter or number.

The control unit 11 acquires the patient ID input via the input unit 14, for example, and displays it in the patient ID input field (S22). For example, when a code reader such as a barcode reader or a QR code (registered trademark) reader is connected to the information processing apparatus 10, the control unit 11 may read the patient ID by the code reader and display it in the input field. good. For example, when a patient wears a wristband printed with a patient ID code, the patient ID may be read from the wristband code by a code reader. The patient ID may be read from the

Also, the control unit 11 (acquisition unit) acquires the medication information input via the input unit 14, for example, and displays it in each input field (S23). Based on the input patient ID, the control unit 11 reads the medication information about the currently administered medication or the most recently administered medication from the patient's electronic medical record data stored in the electronic medical record DB 12a. It may be displayed in each input field. Also, when administering medication using an infusion pump or syringe pump, the drug name, dosage, and administration rate are set in the pump. Therefore, when the information processing device 10 is configured to be able to communicate with the pump, the control unit 11 transmits and receives information to and from the pump, thereby controlling the drug name, dosage, and The administration rate may be obtained from the pump and displayed in the input field. When acquiring each information from the pump, input errors of each information do not occur, and the operation burden on the doctor or nurse can be reduced. Furthermore, when a sticker printed with a code in which drug information including a drug name is coded is attached to the drug package, the control unit 11 reads the drug name from the code of the drug package by the code reader. good. Further, when an RFID (Radio Frequency Identifier) tag storing drug information including a drug name is attached to the drug, and a reader for reading the stored information of the RFID tag is connected to the information processing apparatus 10, the control unit 11 may read the drug name from the RFID tag with a reader.

The control unit 11 determines whether or not the OK button on the input screen has been operated (S24), and if it determines that the OK button has not been operated (S24: NO), returns to the process of step S23, and administers medication. Repeat acquisition and display of information. When determining that the OK button has been operated (S24: YES), the control unit 11 (acquisition unit) acquires the patient's vital data here from the electronic medical record DB 12a (S25). Here, based on the patient ID (patient identification code) input on the input screen, the control unit 11 selects the vital data measured in the most recent predetermined time from the vital data stored in association with the patient ID. to get Note that the control unit 11 acquires vital data measured from a predetermined time (several minutes, several tens of minutes before) to immediately before the start of drug administration, for example. Vital data includes blood pressure, pulse rate, heart rate, respiration rate, electrocardiogram, and the like.

The control unit 11 inputs the medication information input via the input screen and the vital data acquired in step S25 to the learning model 12M, and based on the output values from the learning model 12M, the recovery to be performed for the patient. A treatment is specified (predicted) (S26). For example, in the learning model 12M, the control unit 11 selects a predetermined number (for example, three) of output values (discrimination probabilities) in descending order of the output values from each output node, and outputs the selected output value to the output node. is identified as a candidate for the recovery action to be implemented. Then, the control unit 11 generates advice information for presenting the identified recovery treatment candidates to the doctor or nurse (S27). For example, the control unit 11 generates advice information listing messages explaining the recovery action specified in step S26. Specifically, the control unit 11 controls "stop administration", "increase the dose of the drug being administered", "increase (or decrease) the administration rate of the drug being administered", "antagonist A1, Advisory information including messages such as "administer dose B1 at dose rate C1", "administer drug A3 and dose B3", and "wait and see" is generated.

Then, the control unit 11 (output unit) displays an advice screen as shown in FIG. 6B on the display unit 15 based on the generated advice information (S28). The advice screen shown in FIG. 6B presents three recovery measures in descending order of discrimination probability as candidates for recovery measures to be performed on the patient. Note that the control unit 11 may generate advice information that proposes implementation of a plurality of recovery actions based on the discrimination probability for each recovery action. For example, when the highest discrimination probability and the second discrimination probability are approximately the same value, the control unit 11 may generate advice information suggesting implementation of these two recovery measures. In the example shown in FIG. 6B, Recommendation 2 suggests two recovery actions: "discontinue medication" and "administer antagonist A1, dose B1, administration rate C1." Also, the advice screen may display the determination probability output from the output node corresponding to each recovery action in association with each recovery action. In this case, the degree of recommendation for each recovery action can be presented as a discrimination probability. The doctor confirms the recovery action recommended by the advice screen, selects an appropriate recovery action, and implements the recovery action on the patient. If the information processing apparatus 10 has an audio output unit such as a speaker, the control unit 11 may present the advice information to the doctor or the nurse by outputting the advice information by audio in addition to the display of the advice information. In addition, when the doctor is in a separate room or the like, by providing another terminal capable of communicating with the information processing device 10 in the separate room or the like, the nurse who administered the medication can use the information processing device 10 to display an advice screen. While browsing, it is possible to receive an instruction to perform recovery processing from a doctor via a terminal provided in a separate room or the like. Specifically, the information processing device 10 transmits the advice screen to the terminal in the separate room, and the doctor in the separate room decides on the appropriate treatment from the recovery treatment presented on the advice screen. Nurses can be instructed to perform appropriate recovery procedures.

With the above-described processing, if the patient's condition deteriorates while the drug is being administered to the patient, it is possible to present advice on recovery measures that should be taken to the patient in order to recover from the deteriorated condition. Therefore, the physician can quickly determine the appropriate recovery treatment considering the presented recovery treatment options. When a patient's condition worsens due to medication, it is difficult to make an appropriate judgment because the situation requires urgency and the degree of urgency is high. In addition, when a small number of doctors are available, such as at night, there is a possibility that appropriate support cannot be obtained from experienced or knowledgeable doctors. Even in such a situation, by using the information processing apparatus 10 of the present embodiment, the doctor can appropriately determine the recovery treatment to be performed on the patient, and the nurse can perform appropriate recovery according to the instructions from the doctor. Since treatment can be performed, it is possible to recover the patient's condition earlier. Therefore, it is possible to reduce the burden on medical personnel such as doctors and nurses who determine appropriate recovery measures in an urgent situation. In addition, in the present embodiment, the learning model 12M is used to specify the recovery treatment to be performed on the patient from the patient's medication information and vital data. can identify more appropriate recovery actions.

In this embodiment, appropriate recovery measures can be determined early, so that the worsened patient's condition can be calmed early and the impact on the patient can be minimized. Minimizing the impact on the patient eliminates the need for further recovery procedures, and the medical equipment, drugs, and human resources (medical workers such as doctors and nurses) used for recovery procedures that are no longer required. Reduced use of medical resources, including

In the present embodiment, the information processing apparatus 10 uses the learning model 12M to locally perform processing for specifying recovery measures for recovering from the worsened patient's condition, but the present invention is not limited to this configuration. For example, the information processing apparatus 10 may be configured as a server computer, and the server computer may be configured to perform recovery action specifying processing using the learning model 12M. In this case, the server computer acquires the patient's medication information and vital data from a terminal device used by a doctor or nurse connected via a network, for example, and outputs and displays the specified recovery treatment information to the terminal device. may be configured to Note that the server computer may store the electronic medical record DB 12a, and in this case, the server computer may acquire only patient medication information from the terminal device. Alternatively, a configuration may be adopted in which a first server that performs recovery treatment identification processing using the learning model 12M and a second server that stores the electronic medical record DB 12a are separately provided. In this case, the first server can be configured to acquire the patient's medication information and vital data from the second server, and perform recovery treatment identification processing using the learning model 12M. Even with such a configuration, the same processing as in the present embodiment described above is possible, and the same effects can be obtained.

(Embodiment 2)
When predicting a recovery treatment for recovery from the patient's medication information and vital data, the degree of contribution to the obtained prediction result (discrimination result) among the medication information and vital data input to the learning model 12M is An information processing apparatus that specifies large input data and presents it together with a prediction result will be described. Since the information processing apparatus of this embodiment has the same configuration as the information processing apparatus 10 of Embodiment 1, detailed description of the configuration will be omitted. In addition, in the information processing apparatus 10 of the present embodiment, the learning model stored in the storage unit 12 is slightly different from the learning model 12M of the first embodiment shown in FIG.

FIG. 7 is a schematic diagram showing a configuration example of a learning model according to the second embodiment. The learning model 12Ma of the present embodiment is a so-called explainable AI (XAI), which uses the patient's medication information and vital data as inputs to obtain information on recovery treatment to be performed on the patient (discrimination probability of each recovery treatment ) and explanatory data indicating the grounds for the prediction. Also in this embodiment, the learning model 12Ma may be configured to receive medication information and electronic medical record data. The learning model 12Ma uses the SHAP or Attention mechanism or the like to calculate the degree of contribution (SHAP value, Attention, etc.) that should be used as the basis for obtaining the prediction result (model output) for each input data. is configured to The degree of contribution is a value indicating the degree to which each input data of the learning model 12Ma contributed to the output of the predicted recovery action (prediction result). It is a value that indicates how much attention is given to the element). Therefore, input data with a high degree of contribution can be regarded as the basis for obtaining the prediction result. A learning model 12Ma shown in FIG. 7 has, in addition to the same configuration as the learning model 12M shown in FIG. . Therefore, the learning model 12Ma is configured to output the degree of contribution (description data) to each input data from each output node for description data. For example, output node 101 outputs the degree of contribution to input data 1 (eg, input data of input node 0), and output node 102 outputs the degree of contribution to input data 2 (eg, input data of input node 1).

The information processing apparatus 10 selects a predetermined number (for example, three) of the output values (contributions) from the output nodes for explanation data in descending order, and sends the selected output values to the output node that outputs the output values. Identify the associated input data as a candidate for the basis of prediction. In this case, a predetermined number of pieces of input data can be specified as candidates for the basis of prediction. Note that the information processing apparatus 10 may specify input data associated with an output node that outputs a contribution (output value) equal to or greater than a predetermined value as candidates for grounds for prediction. The learning model 12Ma of the present embodiment has a plurality of output nodes that output the contribution to each input data, instead of outputting information indicating the input data with the highest contribution (for example, an input node number). A configuration having an output node may also be used.

FIG. 8 is a flow chart showing an example of the procedure for presenting recovery measures according to the second embodiment, and FIG. 9 is an explanatory diagram showing an example of a screen. The process shown in FIG. 8 is obtained by adding step S31 between steps S27 and S28 in the process shown in FIG. Description of the same steps as in FIG. 5 will be omitted. The control unit 11 of the information processing apparatus 10 of this embodiment performs the same processes as steps S21 to S27 in FIG. As a result, even in this embodiment, when the patient ID and medication information are input via the input screen, the recovery treatment to be performed on the patient is predicted based on the medication information and vital data, and the predicted recovery treatment is performed. Advice information for notifying is generated. In this embodiment, in step S26, the control unit 11 predicts the recovery action based on the output value from the learning model 12Ma, and specifies the input data that serves as the basis for the prediction result. For example, the control unit 11 selects a predetermined number (for example, three) of contributions from the explanation data (contributions of each input data) output from the learning model 12Ma in descending order, and outputs the selected contributions. The input data associated with the output node is specified as a basis candidate for the prediction result.

After the process of step S27, the control unit 11 generates prediction basis information for notifying the input data that is the basis of the prediction result specified in step S26 (S31). For example, if a drug name is specified as the input data that is the basis of the prediction result, the control unit 11 generates a message (for example, "Recommended action basis: drug A”). Further, when blood pressure is specified as the input data that serves as the basis for the prediction result, the control unit 11 generates a message that notifies the blood pressure as the basis for the prediction result (recommended recovery action). At this time, based on the vital data (blood pressure data) acquired in step S25, the control unit 11 determines whether the blood pressure has risen above the normal range or fallen below the normal range, and predicts changes in blood pressure. may be notified as a basis for

Then, the control unit 11 displays an advice screen as shown in FIG. 9 on the display unit 15 based on the advice information generated in step S27 and the ground information for prediction generated in step S31 (S28). The advice screen shown in FIG. 9 presents, in addition to the content of the advice screen shown in FIG. 6B, input data that has greatly contributed to the prediction of the recommended recovery action. In the example shown in FIG. 9, three pieces of vital data, namely blood pressure, pulse rate, heart rate, and respiration rate, are presented as input data that contributed to the prediction result in descending order of contribution. Note that the contribution of each input data may be displayed in association with each input data. In this case, the degree of contribution of the input data that contributed to the prediction result can also be grasped. The doctor selects an appropriate recovery action by confirming the recovery action recommended by the advice screen and the input data that contributed to the prediction of this recovery action, and implements the recovery action on the patient. In this embodiment, which of the medication information and vital data contributes to the prediction result (prediction of recovery treatment) is presented, so that the doctor can select a more appropriate recovery treatment. can.

In this embodiment, the learning model 12Ma is not limited to the configuration shown in FIG. can be configured to input various information such as comments input by the . In this case, the learning model 12Ma discriminates the recovery action to be performed based on not only the medication information and vital data but also various information included in the electronic medical record data, and at that time, each input for the discrimination result (prediction result) It can be configured to output the contribution of the data. Also in the learning model 12Ma, the electronic medical record data to be input may be time-series data, for example, the patient's electronic medical record data for the most recent day, or the electronic medical record data for the most recent week, etc. are learned. It can be configured to be input to the model 12Ma.

In this embodiment, the same effects as those of the first embodiment described above can be obtained. Moreover, in this embodiment, it is possible not only to predict the recovery treatment to be performed on the patient, but also to present which input data is the basis for the prediction result. Therefore, doctors, nurses, etc. judge whether the presented input data and recovery measures are appropriate or not, and if appropriate recovery measures are adopted and implemented on the patient, the patient's condition can be improved. can be recovered at an early stage. Also in the present embodiment, it is possible to apply the modified examples appropriately described in the first embodiment.

(Embodiment 3)
Embodiments 1 and 2 described above are configured to present recovery measures to be taken to the patient when the patient's condition deteriorates during or after administration of the drug. In this embodiment, based on the patient's vital data during or after administration of the drug, it is automatically determined whether the vital data has become abnormal (whether the condition has deteriorated), and the abnormal condition is determined. A description will be given of an information processing apparatus that presents a patient with recovery measures to be taken when it is determined that the condition has become abnormal. Since the information processing apparatus of this embodiment has the same configuration as the information processing apparatus 10 of Embodiment 1, detailed description of the configuration will be omitted.

FIG. 10 is a flow chart showing an example of a procedure for presenting recovery measures according to the third embodiment, and FIG. 11 is an explanatory diagram showing an example of a screen. The process shown in FIG. 10 is obtained by adding steps S41 to S42 instead of steps S21 to S22 in the process shown in FIG. Description of the same steps as in FIG. 5 will be omitted. The information processing apparatus 10 of the present embodiment measures, for example, patient's vital data (blood pressure, pulse rate, heart rate, respiratory rate, electrocardiogram, body temperature, etc.), and checks whether the measured data is within a normal range. A monitoring device (not shown) is connected to determine and output an alert if the normal range is deviated. In this case, the control unit 11 of the information processing device 10 can be configured to acquire an alert (vital data deviating from the normal range) detected by the monitor device from the monitor device. Note that the monitor device may be configured to output the measured vital data of the patient to the information processing device 10. In this case, the control unit 11 of the information processing device 10 detects that the vital data acquired from the monitor device is It is determined whether or not it is within the normal range.

The control unit 11 of the information processing apparatus 10 of the present embodiment determines whether the patient's vital data is normal depending on whether an alert has been obtained from the monitoring device monitoring the patient's vital data during or after the drug administration. It is determined whether or not a state has been established (S41). If it is determined that the vital data is not normal (S41: NO), that is, if it is determined that the vital data is normal, the controller 11 waits. If it is determined that the vital data is not normal (S41: YES), the control unit 11 generates an alert screen notifying of the abnormal vital data and displays it on the display unit 15 (S42). FIG. 11A shows an example of an alert screen, which displays the patient ID of the patient being monitored and the vital data determined to be abnormal. In the example shown in FIG. 11A, it is notified that the blood pressure has risen above the normal range. The alert screen also has input fields for medication information (drug name, dosage and administration rate) and an OK button, similar to the input screen shown in FIG. 6A.

After displaying the alert screen as shown in FIG. 11A, the control unit 11 executes the processes after step S23. That is, the control unit 11 obtains the medication information and displays it in each input field in the same manner as the processing shown in FIG. Obtained from the electronic medical record DB 12a. Then, based on the obtained medication information and vital data, the control unit 11 identifies recovery measures using the learning model 12M, and displays an advice screen presenting the specified recovery measures on the display unit 15 . In this embodiment, an advice screen as shown in FIG. 11B is displayed. The advice screen shown in FIG. 11B notifies vital data that has become abnormal in addition to the display contents of the advice screen shown in FIG. 6B. With such an advice screen (display information), the patient's vital data that is not within the normal range can be presented, and recovery measures can be proposed to bring the vital data back within the normal range.

In the above-described process, the process of determining whether the patient's vital data is in a normal state may be performed using a learning model learned by machine learning. For example, using an algorithm such as CNN, RNN, LSTM, etc., when medication information about a drug being administered or administered to a patient and vital data from before medication to during medication or after medication are input A learned model can be used that is trained to output information indicative of the patient's condition. In this case, the control unit 11 inputs the medication information about the drug being administered or administered and the vital data acquired from the monitoring device to the learning model, and based on the output information from the learning model, the current state of the patient. It is possible to determine the state or predict the future state.

FIG. 12 is an explanatory diagram showing a configuration example of a learning model used for patient condition determination. The learning model shown in FIG. 12 includes medication information (drug name, dosage and administration rate) of the drug being administered or administered to the patient, and patient vital data (blood pressure, pulse rate, heart rate, Respiratory rate, electrocardiogram, body temperature, etc.) are input, and based on the input data, calculations are performed to discriminate the patient's condition, and the result of calculation (discrimination probability for each condition) is learned to be output. . The vital data input to the learning model may be chronological vital data from before to during or after drug administration. In addition to medication information and vital data, the learning model shown in FIG. It may be configured such that a plurality of pieces of information are input. Further, the electronic medical record data input to the learning model may be time-series data. It may be time-series data (treatment information) on various treatments that have been obtained.

The learning model shown in FIG. 12 has a plurality of output nodes, and each output node outputs a discrimination probability for each associated state. For example, output node 0 outputs the probability that it should be determined that the patient's condition is good, output node 1 outputs the probability that it should be determined that the blood pressure rises above the normal range, and output node 2 outputs the probability that the blood pressure rises above the normal range. It outputs the probability to be determined when it falls below the normal range. The output value of each output node is, for example, a value between 0 and 1.0, and the sum of the discrimination probabilities output from each output node is 1.0. It should be noted that the state assigned to each output node can use a state change that can occur due to medication.

The learning model shown in FIG. 12 uses training data including training medication information and vital data, and information (correct label) indicating the patient's condition that may occur during or after medication. It is generated by machine learning the learning model. In the learning model, when medication information and vital data for training are input, the output value from the output node corresponding to the state indicated by the correct label approaches 1.0, and the output value from the other output nodes approaches 0. .0 is learned. In the learning process, the learning model performs calculations based on the input medication information and vital data, and calculates output values from each output node. Then, the learning model compares the calculated output value of each output node with the value (0 or 1) corresponding to the correct label, and performs arithmetic processing so that each output value approximates the value corresponding to the respective correct label. Optimize the parameters used for Again, parameters to be optimized are weights between neurons in the learning model, and the optimization method is not particularly limited. This results in a learning model that is trained to determine the patient's current or potential future state given the patient's medication information and vital data. Training of the learning model may be performed on other learning devices. A learned learning model generated by learning in another learning device is downloaded from the learning device to the information processing device 10 via, for example, a network or via the portable storage medium 1a, and stored in the storage unit 12.

When using the learning model as shown in FIG. 12, the control unit 11 of the information processing apparatus 10 associates the output node that outputs the maximum output value (discrimination probability) among the output values from the output nodes. Identify the patient's condition as the patient's condition. Note that the control unit 11 selects a predetermined number (for example, three) of the output values from the output nodes in descending order of the output values, and changes the state in which the selected output values are associated with the output nodes that output them. , may be identified as a candidate for the patient's condition. When the patient's condition is discriminated or predicted using the learning model shown in FIG. determine whether or not Then, when it is determined that the state is not normal, the control unit 11 displays an alert screen that notifies the state of the vital data determined to be not normal (S42). Even in such a configuration, the same processing as in the present embodiment described above is possible, and the same effects can be obtained.

In this embodiment, the same effects as those of the above-described embodiments can be obtained. Further, in this embodiment, it is possible to automatically determine whether the patient's condition is normal. Therefore, when an abnormal state is automatically determined, it is possible to notify the abnormal state and provide advice on recovery measures for recovering from the state. Therefore, in this embodiment, deterioration of the patient's condition can be grasped at an early stage, and recovery measures to be taken can be advised. It can assist doctors who need to make good decisions.

The configuration of the present embodiment can be applied to the information processing apparatuses 10 of the first and second embodiments, and similar effects can be obtained even when applied to the information processing apparatuses 10 of the first and second embodiments. Also in this embodiment, it is possible to apply the modified examples appropriately described in each of the above-described embodiments.

(Embodiment 4)
A description will be given of an information processing apparatus that performs a process of receiving an instruction from a doctor to change the content of recovery treatment presented when the patient's condition worsens during or after medication. Since the information processing apparatus of this embodiment has the same configuration as the information processing apparatus 10 of Embodiment 1, detailed description of the configuration will be omitted.

FIG. 13 is a flow chart showing an example of a procedure for presenting recovery measures according to the fourth embodiment, and FIG. 14 is an explanatory diagram showing an example of a screen. The process shown in FIG. 13 is obtained by adding steps S51 to S56 after step S28 in the process shown in FIG. Description of the same steps as in FIG. 5 will be omitted. In FIG. 13, steps S21 to S27 in FIG. 5 are omitted.

The control unit 11 of the information processing apparatus 10 of this embodiment performs the same processes as steps S21 to S28 in FIG. As a result, even in this embodiment, when the patient ID and medication information are input via the input screen, the recovery treatment to be performed on the patient is predicted based on the medication information and vital data, and the predicted recovery treatment is performed. is displayed on the display unit 15 . The control unit 11 of this embodiment displays an advice screen as shown in FIG. 14A. The screen shown in FIG. 14A has the same configuration as the advice screen shown in FIG. 6B, and the recovery measures to be presented are displayed so as to be selectable. Therefore, the control unit 11 accepts selection of one of the recovery measures by operating the input unit 14 on the advice screen. The example shown in FIG. 14A shows a state in which the "recommended 2" recovery action is selected.

The control unit 11 determines whether or not a selection for any of the recovery actions displayed on the advice screen has been received (S51). If it is determined that the selection of the recovery action has not been accepted (S51: NO), the control unit 11 waits until the selection is accepted. Note that when the close button on the advice screen is operated, the control unit 11 terminates the series of processes. When the control unit 11 determines that selection of any recovery action has been accepted (S51: YES), the control unit 11 displays a change acceptance screen for accepting an instruction to change the selected recovery action on the display unit 15 (S52). FIG. 14B shows an example of a change acceptance screen. The screen shown in FIG. 14B displays the recovery action selected on the screen shown in FIG. there is In the example shown in FIG. 14B, input fields are provided for "antagonist A1", dose "B1", and administration rate "C1", and corresponding values are displayed in each input field. The screen shown in FIG. 14B also has a change button for giving an instruction to change the information entered in the input field, and a close button for giving an instruction to end the display of the change acceptance screen. The doctor refers to the recovery measures presented on the advice screen shown in FIG. 14A and selects the recovery measures to be performed on the patient. Modification contents for a part of the contents of the recovery action are input via the change acceptance screen shown in FIG. 14B.

The control unit 11 accepts changes to the information being displayed in the input fields on the change acceptance screen by operating the input unit 14 (S53). When receiving changes to any of the input fields, the control unit 11 displays and updates the received changes in the input fields. The control unit 11 determines whether or not the change button in the change acceptance screen has been operated (S54), and if it is determined that it has not been operated (S54: NO), returns to the process of step S53, and changes the change acceptance screen. We will continue to accept changes to the input fields in .

When the control unit 11 determines that the change button has been operated (S54: YES), the post-change recovery action including the medication information acquired in step S23, the vital data acquired in step S25, and the change received in step S53 (S55). Then, the control unit 11 assigns a correct label indicating the post-change recovery treatment to the acquired medication information and vital data, generates training data, and stores the training data in the storage unit 12 (S56). For example, the control unit 11 stores the generated training data in a DB for training data provided in the storage unit 12 .

By the above-described processing, when the advised recovery action when the patient's condition worsens due to medication is partially changed and adopted, training data with the changed content recovery action as the correct label is generated. can be generated. Therefore, by re-learning the learning model 12M using such training data, the recovery treatment corrected by the doctor can be fed back to the learning model 12M. Therefore, it is possible to generate the learning model 12M that reflects the content of the recovery treatment actually adopted by the doctor. In the above-described processing, it is determined whether or not the training data using the corrected recovery treatment is used as training data depending on whether the patient's condition has recovered to a normal state after the recovery treatment. You may The doctor determines whether or not the vital data of the patient who has undergone recovery treatment has recovered within the normal range, and registers the determination result in the training data stored in the training data DB. In this case, the control unit 11 of the information processing device 10 uses only the training data including the recovery treatment by which the patient has recovered when performing the learning process of the learning model 12M using the training data accumulated in the training data DB. It is possible to re-learn by In this case, it is possible to feed back to the learning model 12M the remedial treatment that has been effective for the patient.

In this embodiment, the same effects as those of the above-described embodiments can be obtained. In addition, in this embodiment, it is possible to generate a learning model 12M that reflects the contents of recovery measures actually adopted by a doctor, and by using such a learning model 12M, it is possible to give advice on more appropriate recovery measures. becomes.

The configuration of the present embodiment can be applied to the information processing apparatuses 10 of Embodiments 1 to 3, and similar effects can be obtained even when it is applied to the information processing apparatuses 10 of Embodiments 1 to 3. Also in this embodiment, it is possible to apply the modified examples appropriately described in each of the above-described embodiments.

The embodiments disclosed this time are illustrative in all respects and should not be considered restrictive. The scope of the present invention is indicated by the scope of the claims rather than the meaning described above, and is intended to include all modifications within the scope and meaning equivalent to the scope of the claims.

REFERENCE SIGNS LIST 10 information processing device 11 control unit 12 storage unit 13 communication unit 14 input unit 15 display unit 12a electronic chart DB
12M learning model 12Ma learning model

Claims (10)

Obtaining vital data of a patient who became abnormal during or after administration of an injection, and the type, dosage and administration rate of the injection,
When inputting the type, dosage, and administration rate of the injection and vital data, the learned model is trained to output advice information on recovery treatment, and the obtained type, dosage, and administration rate of the injection. and vital data, and outputting advice information on recovery treatment for the patient based on the vital data. obtaining electronic medical record data of the patient;
When the type of injection, dosage and administration rate, vital data, and electronic medical record data are input, the learned model is trained to output advice information on recovery measures, and the acquired type of injection. , dosage and rate of administration, vital data, and electronic medical chart data, and outputting advice information regarding said recovery treatment. 3. The program according to claim 1 or 2, which causes the computer to execute a process of outputting display information that displays vital data indicating an abnormal state and advice information regarding the recovery action. acquiring from the learning model information according to the degree of contribution of each piece of data input to the learning model to the output of the advice information;
4. The computer according to any one of claims 1 to 3, wherein, based on the acquired information, the data input to the learning model and having a high degree of contribution to the output of the advice information are output. program described in . 5. The program according to any one of claims 1 to 4, wherein the advice information includes information on the type of injection, dosage and administration rate. 6. The program according to any one of claims 1 to 5, which causes the computer to execute a process of accepting a change instruction for the outputted advice information on the recovery action. generating training data including the type, dosage, and administration rate of the injection input to the learning model, vital data, and advice information regarding recovery measures taken when the patient recovers;
The program according to any one of claims 1 to 6, which causes the computer to execute a process of re-learning the learning model using the generated training data. Obtaining vital data of a patient who became abnormal during or after administration of an injection, and the type, dosage and administration rate of the injection,
When inputting the type, dosage, and administration rate of the injection and vital data, the learned model is trained to output advice information on recovery treatment, and the obtained type, dosage, and administration rate of the injection. and vital data, and outputting advice information regarding recovery treatment for the patient based on the vital data. an acquisition unit that acquires vital data of a patient who is in an abnormal state during or after administration of an injection, and the type, dosage, and administration rate of the injection;
When inputting the type, dosage, and administration rate of the injection and vital data, the learned model is trained to output advice information on recovery treatment, and the obtained type, dosage, and administration rate of the injection. and an output unit for inputting vital data and outputting advice information regarding recovery treatment for the patient based on the vital data. Acquiring training data including the type, dosage and administration rate of the injection administered to the patient, vital data after administration of the injection of the patient, and advice information regarding recovery treatment performed on the patient;
Using the acquired training data, the computer executes a process of generating a learning model that outputs advice information regarding recovery measures when the type, dosage and administration rate of the injection and vital data are input. Model generation method.

Images