JP7820161B2 - Information processing method, program, information processing device, and model generation method - Google Patents

Description

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

It is known that aspiration pneumonia can occur when food in the mouth enters the trachea along with saliva, or when stomach contents reflux up the esophagus and enter the trachea. Additionally, elderly people and others may experience silent aspiration, which does not result in coughing or choking. Because silent aspiration is not recognized by the patient, there is a need for an objective method of detecting aspiration. Patent Document 1 discloses a testing device that can be used to check for the occurrence of gastroesophageal reflux with simple operation.

Japanese Patent Application Laid-Open No. 2019-141511

When performing an examination using a dedicated examination device, such as the examination device disclosed in Patent Document 1, it is difficult to reduce examination costs. Therefore, there is a demand for a system that can easily determine the possibility of aspiration occurring based on the patient's condition before performing an examination using a dedicated examination device.

One aspect of this is to provide an information processing method that can present information regarding the onset of aspiration based on the patient's condition.

In one aspect of the information processing method, a computer executes a process of acquiring status information regarding a patient's condition and outputting onset information regarding the onset of aspiration based on the acquired status information.

On one hand, the patient's condition can provide information about the onset of aspiration.

FIG. 1 is a block diagram illustrating an example of the configuration of an information processing device. FIG. 2 is a schematic diagram showing an example of the configuration of an electronic medical record DB. FIG. 1 is a schematic diagram illustrating an example of the configuration of a learning model. 10 is a flowchart illustrating an example of a learning model generation processing procedure. 10 is a flowchart illustrating an example of a procedure for predicting the occurrence of aspiration due to gastroesophageal reflux. FIG. 10 is an explanatory diagram showing an example of a screen. FIG. 10 is an explanatory diagram showing an example of a screen. FIG. 2 is a schematic diagram showing an example of the appearance of an inspection device. FIG. 2 is a block diagram showing an example of the configuration of an inspection device. FIG. 10 is a schematic diagram illustrating an example of the configuration of a learning model according to a second embodiment. FIG. 10 is a schematic diagram showing an example of a screen of an inspection device. 13 is a flowchart illustrating an example of a procedure for predicting the occurrence of aspiration according to the third embodiment. FIG. 10 is a schematic diagram showing an example of a screen according to a third embodiment.

The information processing method, program, information processing device, and model generation method disclosed herein are described in detail below with reference to drawings illustrating embodiments thereof.

(Embodiment 1)
This paper describes an information processing device that predicts whether a patient is experiencing aspiration based on their condition and presents the results to medical professionals, such as doctors, nurses, and caregivers. In this embodiment, the occurrence of aspiration is predicted for patients receiving enteral nutrition, in which nutrients are administered by natural drip or directly to the stomach using an enteral nutrition pump. Therefore, this embodiment predicts the occurrence of aspiration due to gastroesophageal reflux. Enteral nutrition includes nasogastric tube feeding, in which nutrients are administered using a tube inserted through the nose, and gastrostomy tube feeding, in which nutrients are administered using a tube inserted through a fistula created in the abdomen.

Figure 1 is a block diagram showing an example configuration of an information processing device. The information processing device 10 is a device capable of various information processing and sending and receiving information, such as a server computer or personal computer. The information processing device 10 may be configured to have multiple devices for distributed processing, or may be realized by multiple virtual machines installed in a single device. The information processing device 10 is installed and used, for example, in a medical institution. When the information processing device 10 is configured as a server computer, the information processing device 10 may be a local server installed in the medical institution, or a cloud server connected to the medical institution via a network such as the Internet.

The information processing device 10 performs a process of predicting the possibility of aspiration occurring in a patient at multiple stages based on status information indicating the patient's condition, and notifying the patient by alert when the predicted possibility is high. Specifically, as described below, the information processing device 10 performs machine learning to learn predetermined training data, and has prepared in advance a learning model 12M (see Figure 3) that inputs the patient's status information and outputs the possibility of aspiration occurring due to gastroesophageal reflux. The information processing device 10 inputs the patient's status information into the learning model 12M and obtains the possibility of aspiration occurring due to gastroesophageal reflux from the learning model 12M. The patient's status information includes any one or more of body temperature, blood oxygen saturation, amount of phlegm, presence or absence of wheezing, information about facial expression, presence or absence of oral odor, oral acidity level, presence or absence of sugar in the oral cavity, etc.

The information processing device 10 includes a control unit 11, a memory unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, etc., and these units 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 a program 12P stored in the memory unit 12 to perform various information processing, control processing, etc. that should be performed by the information processing device 10.

The memory unit 12 includes RAM (Random Access Memory), flash memory, a hard disk, an SSD (Solid State Drive), etc. The memory unit 12 pre-stores the program 12P (program product) executed by the control unit 11 and various data necessary for executing the program 12P. The memory unit 12 also temporarily stores data generated when the control unit 11 executes the program 12P. The memory unit 12 also stores a learning model 12M that has learned training data, for example, through machine learning. The learning model 12M is a trained model that has been trained to output the possibility of aspiration due to gastroesophageal reflux (onset information related to the onset of aspiration of gastroesophageal reflux material) when patient condition information is input. The learning model 12M is intended to be used as a program module constituting artificial intelligence software. The learning model 12M performs a predetermined calculation on input values and outputs the calculation results. Data such as function coefficients and thresholds that define this calculation are stored in the memory unit 12 as the learning model 12M. The storage unit 12 also stores an electronic medical record 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 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) via wired or wireless communication, and transmits and receives information to other devices via the network. The input unit 14 accepts operation input by the user and sends control signals corresponding to the operation content to the control unit 11. The display unit 15 is a liquid crystal display, organic EL display, or the like, and displays various information in accordance with instructions from the control unit 11. The input unit 14 and display unit 15 may be integrated into a touch panel. Note that the input unit 14 and display unit 15 are not required, and the information processing device 10 may be configured to accept operations via a connected terminal device and output information to be displayed to an external display device.

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

Figure 2 is a schematic diagram showing an example configuration of the electronic medical record DB 12a. The electronic medical record DB 12a is a database that stores electronic medical record data (medical records) of patients who use medical institutions. The electronic medical record DB 12a shown in Figure 2 includes a patient ID column, a patient information column, a condition information column, and so on. The patient ID column stores identification information (patient ID, patient code) for identifying each patient. The patient ID may be, for example, the patient card number on a patient card issued by a medical institution. The patient information column stores patient information related to the patient in association with the patient ID. The patient information includes, for example, attribute information such as the patient's age and gender, medical history, diagnosis, height, and weight. The patient information also includes information about symptoms experienced by the patient, such as information indicating that the patient has experienced aspiration. Aspiration includes aspiration due to gastroesophageal reflux, in which stomach contents reflux up the esophagus and enter the trachea, and aspiration in which oral contents enter the trachea.

The status information string stores status information related to the patient's condition, associated with the patient ID. The status information includes vital data measured from the patient, such as body temperature, blood pressure, heart rate, and respiratory rate. The status information also includes blood oxygen saturation, amount of phlegm, presence or absence of wheezing, presence or absence of oral odor, facial expression level, oral acidity level, and presence or absence of sugar in the oral cavity. Blood oxygen saturation is measured using a pulse oximeter. The amount of phlegm is the amount of secretions (phlegm, saliva, etc.) aspirated from the patient's oral cavity, pharynx, and larynx using a suction device. Phlegm suction is performed, for example, after administering nutrients via tube feeding. The presence or absence of wheezing indicates whether the patient makes sounds such as throat rasping, whistling, or wheezing when breathing. The presence or absence of wheezing can be determined, for example, by a medical professional listening to the sounds around the patient's mouth, or by collecting sounds around the patient's mouth with a microphone and determining whether the collected sounds match those produced by wheezing. The presence or absence of an odor in the oral cavity indicates whether the odor of the nutritional supplement administered to the patient (e.g., a sweet odor) can be detected in the patient's oral cavity. The presence or absence of an odor can be determined, for example, by a medical professional smelling the patient's mouth, or by using an odor-detecting device to detect the odor of the patient's mouth and determining whether the detected odor matches the odor of the nutritional supplement being administered. The facial expression level can be determined, for example, by a medical professional who observes the patient daily by checking the patient's face. For example, a facial photograph of each patient can be taken at the time of admission and compared with the facial photograph taken at the time of admission and the patient's facial photograph taken at the time of examination. The facial expression level indicates, for example, whether the patient appears distressed or able to communicate compared to their usual facial expression. For example, a normal, normal expression is assigned a level 1, and the level increases as the patient appears distressed or unable to communicate. It is expressed on a scale of, for example, five levels (levels 1 to 5). These data are stored in each column, for example, associated with the measurement date and time, and each data measurement is performed a predetermined number of times per day (e.g., approximately three to five times). The acidity level in the oral cavity is the pH value in the esophagus measured by an esophageal pH monitoring test. The presence or absence of sugar in the oral cavity is measured, for example, using a urine glucose test strip. Note that if stomach contents (gastric juice) are refluxing into the esophagus, the pH value in the esophagus will be low, and the occurrence of gastroesophageal reflux can be determined based on the pH value in the esophagus. Furthermore, if stomach contents (food, nutrients) are refluxing into the oral cavity, sugar in the stomach contents will be detected in the oral cavity, and the occurrence or absence of gastroesophageal reflux can be determined based on the presence or absence of sugar in the oral cavity. Each piece of data may be represented as a graph showing changes over time, and the graph data is a series of data recorded each time a measurement is taken, for example, consisting of a pair of measurement date and time and the measured value.

The contents stored in electronic medical record DB12a are not limited to the example shown in Figure 2, and may include various types of information related to the patient, such as test information related to various tests performed on the patient, such as blood tests and urine tests, treatment information related to treatment, surgery information related to surgery, medication history, and comments entered by doctors, etc. The test information may also include medical images such as X-ray images, ultrasound images, CT (Computed Tomography) images, MRI (Magnetic Resonance Imaging) images, and PET (Positron Emission Tomography) images.

Figure 3 is a schematic diagram showing an example configuration of learning model 12M. Learning model 12M shown in Figure 3 is trained to input patient condition information (body temperature, blood oxygen saturation, amount of phlegm, presence or absence of wheezing, facial expression level, and presence or absence of oral odor), perform a calculation to determine the possibility of aspiration due to gastroesophageal reflux in the patient based on the input data, and output the calculation result (the possibility of aspiration due to gastroesophageal reflux). Learning model 12M is configured, for example, by a convolutional neural network (CNN), a neural network model generated by deep learning. Learning model 12M may be configured using algorithms other than CNN, such as recurrent neural networks (RNNs), long short-term memory (LSTMs), decision trees, random forests, and support vector machines (SVMs), or may be configured by combining multiple algorithms. Information processing device 10 generates learning model 12M in advance by performing machine learning to learn predetermined training data. The information processing device 10 then inputs the patient's condition information registered in the electronic medical record DB 12a into the learning model 12M, and predicts the possibility that the patient is experiencing aspiration due to gastroesophageal reflux based on the output information from the learning model 12M. Each piece of data input into the learning model 12M may be time-series data measured periodically.

The learning model 12M shown in FIG. 3 has three output nodes. Each output node is associated with a state in which there is a high, medium, or low possibility of aspiration due to gastroesophageal reflux, and outputs the probability that the associated state should be determined. A low probability state refers to a normal state in which there is almost no possibility of aspiration. In the example shown in FIG. 3, output node 0 outputs the probability that there is a high possibility of aspiration due to gastroesophageal reflux, output node 1 outputs the probability that there is a medium possibility of aspiration due to gastroesophageal reflux, and output node 2 outputs the probability that there is a low possibility of aspiration due to gastroesophageal reflux. The output value of each output node is, for example, a value between 0 and 1.0, and the sum of the determination probabilities output from each output node is 1.0 (100%). With the above-described configuration, when patient condition information is input, learning model 12M outputs the possibility that the patient has aspiration due to gastroesophageal reflux (the determination probabilities for each of high, medium, and low possibilities).

In the above-mentioned learning model 12M, the information processing device 10 identifies, for example, the possibility associated with the output node that outputs the largest output value (discrimination probability) among the output values from each output node as the possibility to be predicted. Note that instead of having multiple output nodes that output discrimination probabilities for each possibility, the learning model 12M may be configured to have a single output node that outputs the possibility with the highest discrimination probability.

The learning model 12M is generated by machine learning an untrained learning model using training data including training condition information and information (correct answer labels) indicating the possibility of aspiration due to gastroesophageal reflux. The training data is generated by adding information indicating a high possibility of aspiration due to gastroesophageal reflux (onset information related to the onset of aspiration of gastroesophageal reflux material) to the condition information of a patient in whom gastroesophageal reflux has occurred. The training data may also be generated by adding information indicating a low possibility of aspiration due to gastroesophageal reflux to the condition information of a patient in whom gastroesophageal reflux has not occurred. The patient's condition information and information on whether gastroesophageal reflux has occurred can be obtained from information stored in the electronic medical record DB 12a.

When training status information is input, the learning model 12M learns so that the output value from the output node corresponding to the possibility indicated by the correct label approaches 1.0 and the output values from other output nodes approach 0.0. During the learning process, the learning model 12M performs calculations based on the input status information to calculate output values from each output node. The learning model 12M then compares the calculated output value of each output node with the value corresponding to the correct label (1 for the output node corresponding to the possibility indicated by the correct label, and 0 for other output nodes), and optimizes parameters used in the calculation process so that each output value approximates the value corresponding to the correct label. These parameters are, for example, weights between neurons in the learning model 12M. The parameter optimization method is not particularly limited, but examples that can be used include backpropagation and steepest descent. As a result, a learning model 12M is obtained that is trained to predict the possibility of aspiration due to gastroesophageal reflux and output the prediction result when patient status information is input.

The learning model 12M may be trained by another learning device. The trained learning model 12M generated by training on another learning device is downloaded from the learning device to the information processing device 10, for example, via a network or a portable storage medium 1a, and stored in the storage unit 12. The learning model 12M is not limited to the configuration shown in FIG. 3. The learning model 12M may be configured to receive information about the patient's condition used to predict the occurrence of aspiration due to gastroesophageal reflux and the occurrence of aspiration pneumonia. For example, in addition to the data shown in FIG. 3, the learning model 12M may be configured to receive information indicating the acidity level in the patient's oral cavity or the presence or absence of sugar in the oral cavity. Furthermore, the data input to the learning model 12M may be time-series data, such as time-series data of measurement results that are periodically measured and assessment results that are periodically assessed. In this case, the learning model 12M may be configured to receive, for example, data from the patient's most recent day or several days.

The process of learning training data and generating a learning model 12M is described below. Figure 4 is a flowchart showing an example of the processing steps for generating a learning model 12M. The following process is performed by the control unit 11 of the information processing device 10 in accordance with the program 12P stored in the memory unit 12, but may also be performed by another learning device.

The control unit 11 of the information processing device 10 first obtains 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 train the learning model 12M. For example, the control unit 11 determines whether any patient has experienced aspiration due to gastroesophageal reflux based on the information about each patient stored in the electronic medical record DB 12a (S11). If it determines that any patient has experienced aspiration due to gastroesophageal reflux (S11: YES), the control unit 11 obtains the patient's condition information from the electronic medical record DB 12a (S12). For example, the control unit 11 reads from the electronic medical record DB 12a each piece of data related to the patient's condition, including body temperature, blood oxygen saturation, amount of sputum, presence or absence of wheezing, facial expression level, and presence or absence of oral odor, measured or determined during a predetermined period before and after the occurrence of aspiration due to gastroesophageal reflux. The control unit 11 generates training data by attaching a correct answer label indicating a "high" likelihood of aspiration due to gastroesophageal reflux to the acquired condition information, and stores the data in the memory unit 12 (S13).

The control unit 11 returns to step S11 and determines whether there are any other patients who have experienced aspiration due to gastroesophageal reflux. Each time it determines that there are such patients (S11: YES), it performs steps S12 to S13 to generate training data. The control unit 11 repeats steps S12 to S13 until it determines that there are no patients who have experienced aspiration due to gastroesophageal reflux. This allows it to generate and accumulate training data to be used for learning the learning model 12M based on information about patients who have experienced aspiration due to gastroesophageal reflux. If the control unit 11 determines that there are no patients who have experienced aspiration due to gastroesophageal reflux (S11: NO), it determines whether there are any patients who have never experienced aspiration due to gastroesophageal reflux (patients who have not experienced gastroesophageal reflux) (S14). If it determines that there are patients who have not experienced aspiration due to gastroesophageal reflux (S14: YES), the control unit 11 obtains the patient's condition information from the electronic medical record DB 12a (S15). Here, the control unit 11 reads the same data as in step S12 from the electronic medical record DB 12a. Then, the control unit 11 generates training data by attaching a correct answer label indicating a "low" likelihood of aspiration due to gastroesophageal reflux to the acquired condition information, and stores the data in the memory unit 12 (S16).

The control unit 11 returns to the processing of step S14 and determines whether there are any other patients who have not experienced aspiration due to gastroesophageal reflux. Each time it determines that there are such patients (S14: YES), it performs the processing of steps S15 to S16 to generate and store training data. The control unit 11 repeats the processing of steps S15 to S16 until it determines that there are no patients who have not experienced aspiration due to gastroesophageal reflux. This allows training data to be used for learning the learning model 12M to be generated and stored based on information about patients who have not experienced aspiration due to gastroesophageal reflux. Note that in the above-mentioned processing, training data associated with a correct answer label indicating a "low" likelihood of occurrence is generated based on information about patients who have not experienced aspiration due to gastroesophageal reflux, but this configuration is not limited to this. For example, training data associated with a correct answer label indicating a "low" likelihood of occurrence may be generated based on status information about patients who have experienced aspiration due to gastroesophageal reflux, when aspiration due to gastroesophageal reflux has not occurred.

If the control unit 11 determines that there are no patients who have never experienced aspiration due to gastroesophageal reflux (S14: NO), it performs a learning process for the learning model 12M using the accumulated training data (S17). Here, the control unit 11 inputs the state information contained in the training data into the learning model 12M and obtains the output value from each output node. The control unit 11 compares the output value of each output node with the value corresponding to the correct answer label (1 for the output node corresponding to the possibility of the correct answer, 0 for other output nodes), and optimizes parameters such as the weights between neurons in the learning model 12M using, for example, backpropagation, so that the two values approximate each other.

The control unit 11 determines whether or not there is any training data (unprocessed data) that has not yet undergone learning processing among the training data accumulated in steps S13 and S16 (S18). If it determines that there is unprocessed data (S18: YES), the control unit 11 returns to the processing of step S17 and repeats the learning processing using the unprocessed training data. If it determines that there is no unprocessed data (S18: NO), the control unit 11 ends the series of processing steps.

The above-described processing generates a learning model 12M that has been trained to output the possibility that a patient is experiencing aspiration due to gastroesophageal reflux when patient condition information is input. It is possible to further optimize the learning model 12M by repeatedly performing the learning processing using the training data described above. Furthermore, a learning model 12M that has already been trained can also be retrained by performing the above-described processing, in which case a learning model 12M with higher discrimination accuracy can be generated.

The following describes a process that uses the learning model 12M generated by the above-described process to predict the possibility of aspiration due to gastroesophageal reflux in a patient and output an alert when a medical professional enters the results of their observation of the patient's condition into the electronic medical record. Figure 5 is a flowchart showing an example of the processing procedure for predicting the occurrence of aspiration due to gastroesophageal reflux, and Figures 6 and 7 are explanatory diagrams showing example screens. The following process is performed by the control unit 11 of the information processing device 10 in accordance with the program 12P stored in the memory unit 12.

Nursing staff in medical settings make rounds to observe hospitalized patients a set number of times a day. During these rounds, they measure the patient's temperature, blood pressure, heart rate, blood oxygen saturation, etc., and enter these values into the electronic medical record. They also determine whether the patient is wheezing, whether there is a specific odor in the mouth, and the level of facial expression, and enter the presence or absence of wheezing, the presence or absence of odor, and the level of facial expression into the electronic medical record. Furthermore, at a set time after administering nutrients via tube feeding, nursing staff or caregivers use a suction device to suction secretions from the patient's mouth, etc., and enter the amount of secretions suctioned (amount of phlegm) into the electronic medical record. The amount of phlegm entered into the electronic medical record may include the rinsing water used during suction, in addition to phlegm and saliva.

For simplicity's sake, the following description will use as an example a process in which a medical professional, such as a nursing staff member, inputs patient information via the input unit 14 of the information processing device 10 to update the electronic medical record. However, the medical professional may also update the electronic medical record using a staff terminal (not shown) that can communicate with the information processing device 10. In this case, the medical professional accesses the electronic medical record DB 12a of the information processing device 10 using the staff terminal, which has an input unit and a display unit. In this configuration, an electronic medical record screen displaying the electronic medical record is transmitted from the information processing device 10 to the staff terminal and displayed on the staff terminal, and each piece of data input via the electronic medical record screen is transmitted from the staff terminal to the information processing device 10 and registered in the electronic medical record DB 12a.

When the control unit 11 of the information processing device 10 is instructed to launch the electronic medical record via the input unit 14, it displays an electronic medical record screen on the display unit 15, displaying the electronic medical record data of the specified patient (S21). Figure 6 shows an example of an electronic medical record screen, and the electronic medical record screen shown in Figure 6 displays the patient's patient ID, name, age, and gender. This information is read from the electronic medical record DB 12a and displayed on the electronic medical record screen. The screen shown in Figure 6 displays the current time and has input fields for entering the patient's body temperature, blood oxygen saturation, amount of sputum, presence or absence of wheezing, presence or absence of oral odor, and facial expression level measured at the current time. The screen shown in Figure 6 also has an aspiration determination button for instructing the execution of a process to determine the possibility of the patient having aspiration due to gastroesophageal reflux based on the data entered in each input field. The input fields for body temperature, blood oxygen saturation, and sputum volume are configured to allow the input of any numerical value, the input fields for wheezing and odor are configured with radio buttons that allow the user to select either "yes" or "no," and the input field for facial expression level is configured to allow the user to select one of levels 1 to 5. In addition to the above-mentioned input fields, the screen shown in FIG. 6 may also have an input field for inputting the acidity level in the oral cavity, an input field for inputting the presence or absence of sugar in the oral cavity, etc.

The medical professional enters data indicating the patient's condition into each input field on the screen shown in Figure 6, and operates the aspiration determination button when wanting to determine the possibility of aspiration due to gastroesophageal reflux. In the information processing device 10, the control unit 11 determines whether input of patient condition information has been accepted via each input field on the screen shown in Figure 6 (S22), and if it determines that input has not been accepted (S22: NO), it waits. If the control unit 11 (acquisition unit) determines that input of patient condition information has been accepted (S22: YES), it displays the entered condition information in the respective input fields (S23).

The control unit 11 determines whether the aspiration determination button on the screen shown in FIG. 6 has been operated (S24). If it determines that it has not been operated (S24: NO), the control unit 11 returns to the processing of step S22 and continues to accept data input into each input field. If it determines that the aspiration determination button has been operated (S24: YES), the control unit 11 predicts the possibility that the patient has aspiration due to gastroesophageal reflux based on the input condition information (S25). Specifically, the control unit 11 inputs the condition information input via the screen shown in FIG. 6 into the learning model 12M, and predicts the possibility of aspiration due to gastroesophageal reflux based on the output value from the learning model 12M. For example, the control unit 11 identifies the output node that output the largest output value (discrimination probability) among the output values from the learning model 12M, and determines the possibility associated with the identified output node as the possibility of the patient.

Based on the prediction results, the control unit 11 determines whether a high probability of occurrence has been predicted (S26). If the control unit 11 (output unit) determines that a high probability of occurrence has been predicted (S26: YES), it outputs an alert to notify the user of the high probability of occurrence (S27). For example, as shown in FIG. 7A, the control unit 11 outputs an alert by displaying a message on the electronic medical record screen notifying the user that aspiration due to gastroesophageal reflux is highly likely and that confirmation by a physician is required. This allows medical professionals who have entered patient information into the electronic medical record to understand the predicted high probability of aspiration due to gastroesophageal reflux based on the patient's condition information. In the above-described process, the likelihood of aspiration due to gastroesophageal reflux is determined as high, medium, or low. However, it may also be determined using a numerical value, such as 80% or more, 60% or more but less than 80%, 40% or more but less than 60%, 20% or more but less than 40%, or less than 20%. In such cases, an alert can be output if the predicted likelihood is equal to or greater than a predetermined value (e.g., 60% or more). If the information processing device 10 has an audio output unit such as a speaker, the control unit 11 may be configured to output the alert by outputting the alert message as audio in addition to displaying the alert message.

When an alert like the one shown in Figure 7A is output, for example, a nursing staff member will ask a physician to confirm whether or not aspiration due to gastroesophageal reflux has occurred. The physician will observe the patient's condition and, if necessary, perform chest X-rays, CT scans, blood tests, etc., and make a definitive diagnosis based on the test results. Furthermore, if the physician determines that treatment is necessary for the patient based on the results of the patient's observation or definitive diagnosis, he or she will instruct the nursing staff member to implement the treatment. Treatments here include, for example, slowing the rate of nutritional supplement administration, extending the time the patient is kept upright during and after nutritional supplement administration, changing the type of nutritional supplement, or switching from nasogastric tube feeding to gastrostomy feeding. Changing the type of nutritional supplement includes, for example, switching from liquid nutritional supplements to semi-solid nutritional supplements or variable viscosity nutritional supplements, or using digested or semi-digested nutritional supplements.

If the control unit 11 determines that a high likelihood of occurrence is not predicted (S26: NO), it determines whether a low likelihood of occurrence is predicted (S28). If the control unit 11 determines that a low likelihood of occurrence is not predicted (S28: NO), i.e., if the control unit 11 predicts a moderate likelihood of occurrence, the control unit 11 outputs advice to prevent aspiration due to gastroesophageal reflux (S29). For example, as shown in FIG. 7B, the control unit 11 outputs the advice by displaying a message on the electronic medical record screen notifying the patient that aspiration due to gastroesophageal reflux is moderate and recommending appropriate countermeasures. Recommended countermeasures include, for example, extending the wait time during and after administration of the nutritional supplement, slowing the administration rate of the nutritional supplement, etc. This allows the medical professional who entered the patient's information into the electronic medical record to understand that the predicted moderate likelihood of aspiration due to gastroesophageal reflux is predicted based on the patient's condition information, and to understand appropriate countermeasures to prevent aspiration due to gastroesophageal reflux. Here again, the control unit 11 may output the advice by audio output in addition to displaying the advice message.

When advice such as that shown in Figure 7B is output, for example, nursing staff will take action in accordance with the advice. If the control unit 11 determines that it has predicted a low probability of occurrence (S28: YES), that is, if it predicts that the condition is normal and the probability of occurrence is almost zero, it skips the processing of step S29 and ends the series of processes. The control unit 11 may also display a message on the display unit 15 notifying the user that the probability of aspiration due to gastroesophageal reflux occurring is low, or that the condition is normal and the probability of occurrence is almost zero. The control unit 11 stores the information (electronic medical record data) entered via the electronic medical record screen in the electronic medical record DB 12a.

Through the above-described processing, when medical professionals such as nursing staff and caregivers enter information about a patient's condition into an electronic medical record, the likelihood of aspiration due to gastroesophageal reflux is predicted and notified. Therefore, for example, when nursing staff perform rounds to observe the condition of a hospitalized patient, they can determine whether a doctor's consultation is necessary and whether any treatment is required for the patient. This allows even inexperienced nursing staff to determine whether a doctor's consultation is necessary during their normal nursing duties, thereby reducing the likelihood of aspiration due to gastroesophageal reflux being overlooked and enabling early detection. Furthermore, in this embodiment, a learning model 12M is used to predict the likelihood of aspiration due to gastroesophageal reflux from patient condition information. Therefore, when a learning model 12M that has been trained with a large amount of training data is used, more accurate predictions are possible.

In this embodiment, a medical professional inputs patient condition information via the electronic medical record screen shown in FIG. 6 , but this configuration is not limited thereto. For example, if the information processing device 10 is capable of transmitting and receiving information between a thermometer, a pulse oximeter, and a suction device, it may receive the patient's temperature from the thermometer, the blood oxygen saturation from the pulse oximeter, and the amount of sputum aspirated from the suction device. In this case, the control unit 11 of the information processing device 10 displays each received data in the respective input fields on the screen shown in FIG. 6 . Alternatively, a microphone may collect voice data from the patient's mouth and transmit it to the information processing device 10, which may then determine whether the patient is wheezing based on the received voice data. Alternatively, an odor sensor may detect odors from the patient's mouth and transmit them to the information processing device 10, which may then determine whether a specific odor is present based on the received odor data. Alternatively, a camera may capture an image of the patient's face and transmit it to the information processing device 10, which may then determine an expression level based on the received facial image data, depending on whether the patient appears distressed. When determining the facial expression level from a patient's facial image, for example, template images corresponding to each facial expression level may be prepared in advance, and the patient's facial expression level may be determined by pattern matching using the template images. Alternatively, for example, a facial image may be taken of the patient upon admission and the features of the facial image may be registered, and the patient's facial expression level may be determined based on the results of a comparison with the features of the facial image taken at the time of admission. The camera used to capture the patient's facial image may be, for example, a surveillance camera installed in the hospital room, a camera installed in AR (Augmented Reality) glasses worn by medical staff, or a camera installed in a staff terminal carried by medical staff. In such cases, the control unit 11 of the information processing device 10 displays each determined data in each input field on the screen shown in FIG. 6.

In this embodiment, a configuration has been described in which the information processing device 10 locally performs processing to predict the likelihood of aspiration due to gastroesophageal reflux using the learning model 12M, but this configuration is not limited to this. For example, the information processing device 10 may be configured as a server computer, and the processing to predict the likelihood of aspiration due to gastroesophageal reflux using the learning model 12M may be performed on the server computer. In this case, the server computer may be configured to obtain patient condition information from a staff terminal used by a medical professional connected via a network, for example, and to output and display an alert or advice including the predicted likelihood of occurrence to the staff device. 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)
A configuration for predicting and presenting the possibility of aspiration due to gastroesophageal reflux based on patient condition information using a dedicated testing device will be described. FIG. 8 is a schematic diagram showing an example of the appearance of the testing device. The testing device 20 has a plate-like outer shape. FIG. 8A shows an example of one side (e.g., the front side) of the testing device 20, and FIG. 8B shows an example of the other side (e.g., the back side) of the testing device 20. As shown in FIG. 8 , a touch panel 24 is provided on one side of the testing device 20, and a microphone 25 and a camera 26 are provided on the other side. FIG. 9 is a block diagram showing an example of the configuration of the testing device 20. The testing device 20 includes a control unit 21, a memory unit 22, a communication unit 23, the touch panel 24, the microphone 25, the camera 26, etc., and these units are interconnected via a bus. The control unit 21, the memory unit 22, and the communication unit 23 of the testing device 20 have the same configurations as the control unit 11, the memory unit 12, and the communication unit 13 of the information processing device 10 of embodiment 1, and therefore detailed description thereof will be omitted. The memory unit 22 of the testing equipment 20 stores a program 22P executed by the control unit 21, a learning model 22M, a wheezing recognition model 22M1, and a facial expression recognition model 22M2.

The touch panel 24 is an integrated unit consisting of an input unit that accepts user input and a display unit such as a liquid crystal display or organic EL display. The microphone 25 is a sound collection unit that collects surrounding sounds in accordance with instructions from the control unit 21, and sends the acquired sound data to the memory unit 22 for storage. The camera 26 is a photographing unit that takes pictures in accordance with instructions from the control unit 21, and sends the acquired photographed images (image data) to the memory unit 22 for storage.

Figure 10 is a schematic diagram showing an example of the configuration of a learning model in embodiment 2. As shown in Figure 10A, this embodiment includes a learning model 22M having a configuration similar to that of learning model 12M in embodiment 1 shown in Figure 3, as well as a wheezing recognition model 22M1 and a facial expression recognition model 22M2. Learning model 22M in this embodiment is also configured to output the possibility that a patient is experiencing aspiration due to gastroesophageal reflux when patient condition information (body temperature, blood oxygen saturation, amount of sputum, presence or absence of wheezing, facial expression level, presence or absence of oral odor) is input. Learning model 22M in this embodiment receives input of the presence or absence of wheezing determined by wheezing recognition model 22M1 based on the voice at the patient's mouth, and the facial expression level determined by the facial expression recognition model based on an image of the patient's face.

Figure 10B shows an example configuration of wheezing recognition model 22M1. The wheezing recognition model 22M1 shown in Figure 10B is trained to input voice data collected from the patient's mouth, perform a calculation to determine whether the patient is wheezing based on the input voice data, and output the calculation result (wheezing presence or absence). The wheezing recognition model 22M1 is configured using an algorithm such as CNN or RNN, and may also be configured by combining multiple algorithms. The wheezing recognition model 22M1 shown in Figure 10B has two output nodes. Output node 0 outputs the probability that wheezing should be determined to be occurring, and output node 1 outputs the probability that wheezing should be determined to be absent. The output value of each output node is, for example, a value between 0 and 1.0, and the sum of the determination probabilities output from each output node is 1.0 (100%). With the above configuration, when voice data from the patient's mouth is input, the wheezing recognition model 22M1 outputs information indicating whether the patient is wheezing (the determination probability of wheezing or not). The information processing device 10 identifies the wheezing state associated with the output node that outputs the maximum (largest) output value (discrimination probability) among the output values from each output node in the above-mentioned wheezing recognition model 22M1 as the wheezing state to be predicted. Note that instead of having two output nodes that output discrimination probabilities for the presence or absence of wheezing, the wheezing recognition model 22M1 may be configured to have a single output node that outputs the wheezing state with the highest discrimination probability (wheezing present or absent).

Wheezing recognition model 22M1 is generated by machine learning an untrained learning model using training data containing training audio data and information indicating whether wheezing is occurring (correct answer label). Training data is generated by adding information indicating the presence of wheezing to audio data of a patient's mouth when wheezing occurs. Training data is also generated by adding information indicating the absence of wheezing to audio data of a patient's mouth when wheezing does not occur. Patient audio data may be collected using microphone 25 of testing device 20, or may be obtained from electronic medical record DB 12a if stored in electronic medical record DB 12a. Information on the presence or absence of wheezing can be obtained from information stored in electronic medical record DB 12a. When training audio data is input, wheezing recognition model 22M1 learns so that the output value from the output node corresponding to the wheezing state (presence or absence of wheezing) indicated by the correct answer label approaches 1.0, and the output values from other output nodes approach 0.0. Here too, wheezing recognition model 22M1 performs a learning process by optimizing the weights between neurons using backpropagation, steepest descent, etc. This results in wheezing recognition model 22M1 that has been trained to predict whether wheezing is occurring when patient voice data is input and output the prediction result.

Figure 10C shows an example configuration of facial expression recognition model 22M2. The facial expression recognition model 22M2 shown in Figure 10C is trained to input facial image data obtained by photographing a patient's face, perform a calculation to determine the patient's facial expression level based on the input facial image data, and output the calculation result (facial expression level). As in embodiment 1, facial expression levels are expressed on a five-level scale (levels 1 to 5), with level 1 representing a normal, normal expression and higher levels representing a state of distress or inability to communicate. The facial expression recognition model 22M2 is configured using algorithms such as CNN and RNN, and may also be configured by combining multiple algorithms. The facial expression recognition model 22M2 shown in Figure 10C has five output nodes, with output nodes 0 to 4 outputting the probability of determining the patient as an expression level 5 to 1, respectively. The output value of each output node is, for example, a value between 0 and 1.0, and the sum of the determination probabilities output from each output node is 1.0 (100%). With the above-described configuration, when facial image data of a patient is input, facial expression recognition model 22M2 outputs a discrimination probability for each facial expression level. In the above-described facial expression recognition model 22M2, information processing device 10 identifies the facial expression level associated with the output node that outputs the largest output value (discrimination probability) among the output values from each output node as the facial expression level to be predicted. Note that, instead of having multiple output nodes that output discrimination probabilities for each facial expression level, facial expression recognition model 22M2 may be configured to have a single output node that outputs the facial expression level with the highest discrimination probability.

The facial expression recognition model 22M2 is generated by machine learning an untrained learning model using training data including training facial image data and the facial expression level (correct answer label) of the patient whose facial image data is captured. The training data is generated by assigning the patient's facial expression level, determined by a medical professional or other professional based on the facial image data, to the patient's facial image data. The patient's facial image data may be captured using the camera 26 of the inspection device 20, or may be obtained from the electronic medical record DB 12a if stored in the electronic medical record DB 12a. The patient's facial expression level may be obtained from information stored in the electronic medical record DB 12a. When training facial image data is input, the facial expression recognition model 22M2 learns so that the output value from the output node corresponding to the facial expression level indicated by the correct answer label approaches 1.0 and the output values from other output nodes approach 0.0. Here, too, the facial expression recognition model 22M2 performs the learning process by optimizing the weights between neurons using backpropagation, steepest descent, or the like. This results in a facial expression recognition model 22M2 that has been trained to predict the facial expression level of the patient being photographed when facial image data of the patient is input and output the predicted result.

The testing device 20 of this embodiment, like the information processing device 10 of embodiment 1 described above, performs processing to predict the likelihood of aspiration due to gastroesophageal reflux from patient condition information. Specifically, the control unit 21 of the testing device 20 is capable of executing processing similar to that shown in FIG. 5. Note that in step S21 of FIG. 5, the control unit 21 of the testing device 20 displays a screen such as that shown in FIG. 11A on the touch panel 24 instead of the electronic medical record screen. FIG. 11 is a schematic diagram showing an example screen of the testing device 20. The screen shown in FIG. 11A is an input screen for inputting patient condition information, and has a configuration similar to that of the screen shown in FIG. 6.

The control unit 21 of the testing device 20 accepts input of patient condition information via the input screen shown in FIG. 11A and displays each piece of input information in the respective input fields (S22-S23). Note that each piece of information may be entered into the input screen by a nursing staff member or other person operating the touch panel 24, or may be received by the testing device 20 from another testing device. For example, if the testing device 20 is configured to send and receive information to and from a thermometer, pulse oximeter, and aspirator, the control unit 21 of the testing device 20 receives the patient's body temperature, blood oxygen saturation, and amount of sputum from the thermometer, pulse oximeter, and aspirator, respectively. In this case, the control unit 21 displays each piece of received data in the respective input fields on the screen shown in FIG. 11A.

The control unit 21 of the testing device 20 also collects the patient's voice using the microphone 25, and based on the obtained voice data, determines whether wheezing is occurring using the wheezing recognition model 22M1, and obtains the determination result (wheezing or not). In this case, the control unit 21 displays the obtained determination result in the wheezing input field on the screen shown in FIG. 11A. The control unit 21 also captures a facial image of the patient using the camera 26, and based on the obtained facial image data, determines the facial expression level using the facial expression recognition model 22M2, and obtains the determination result. In this case, the control unit 21 displays the obtained determination result in the facial expression level input field on the screen shown in FIG. 11A. In addition to using the facial expression recognition model 22M2, the process of determining the facial expression level from the patient's facial image may also be performed by preparing template images corresponding to each facial expression level in advance and performing pattern matching using the template images. For example, a facial image may be taken at the time of admission to the patient, and the features of the facial image may be registered, and the patient's facial expression level may be determined based on the results of a comparison with the features of the facial image at the time of admission.

The testing device 20 may also be configured to have an odor sensor that detects odor around the patient's mouth and determines whether or not a specific odor is present. In this case, the control unit 21 uses the odor sensor to detect odor around the patient's mouth and determines whether or not a specific odor (the odor of the nutritional supplement administered to the patient) is present. The control unit 21 then displays the determination result in the input field for intraoral odor on the screen shown in FIG. 11A.

When the aspiration determination button on the input screen shown in FIG. 11A is operated, the control unit 21 of the testing device 20 predicts the possibility of aspiration due to gastroesophageal reflux for the patient based on the patient information entered via the input screen (S24-S25). If the control unit 21 determines that aspiration due to gastroesophageal reflux is highly likely (S26: YES), it displays a message such as that shown in FIG. 11B on the touch panel 24 and outputs an alert (S27). The screen shown in FIG. 11B has the same configuration as the screen shown in FIG. 7A. This allows medical personnel, in this embodiment as well, to understand the high possibility of aspiration due to gastroesophageal reflux, as predicted based on the patient's condition information. The alert message shown in FIG. 11B also has a reset button. When the reset button is operated, the control unit 21 terminates the display of the alert message and erases and resets the data entered in each input field on the screen.

Also in this embodiment, if the control unit 21 determines that it has not predicted that the likelihood of aspiration due to gastroesophageal reflux is low (S28: NO), it outputs advice to prevent aspiration due to gastroesophageal reflux (S29). Although not shown, here, instead of the alert message shown in FIG. 11B, an advice message such as that shown in FIG. 7B is displayed on the touch panel 24. This allows medical personnel to understand the appropriate measures to take to prevent aspiration due to gastroesophageal reflux.

This embodiment achieves the same effects as the first embodiment described above. Furthermore, this embodiment uses a dedicated testing device 20 to determine the possibility of aspiration due to gastroesophageal reflux, allowing nursing staff, caregivers, and others to easily perform the test.

(Embodiment 3)
An information processing device that predicts and displays the possibility of aspiration of oral contents, in addition to the possibility of aspiration due to gastroesophageal reflux, based on patient condition information will be described. In this embodiment, the occurrence of aspiration is predicted for patients who are able to take food orally, as well as for patients receiving tube feeding. The information processing device of this embodiment has the same configuration as the information processing device 10 of embodiment 1, and therefore a description of the configuration will be omitted.

Figure 12 is a flowchart showing an example of the aspiration occurrence prediction process procedure in embodiment 3, and Figure 13 is a schematic diagram showing an example screen in embodiment 3. The process shown in Figure 12 is the process shown in Figure 5 with steps S31 to S33 added instead of steps S28 to S29. Explanations of the same steps as in Figure 5 will be omitted.

The control unit 11 of the information processing device 10 of this embodiment performs processing similar to steps S21 to S27 in Figure 5. As a result, when patient condition information is input via the electronic medical record screen, the possibility of aspiration due to gastroesophageal reflux is predicted based on the condition information, and if a high possibility of occurrence is predicted, an alert is output to notify medical personnel.

When the control unit 11 of the information processing device 10 of this embodiment determines that aspiration due to gastroesophageal reflux is not highly likely (S26: NO), i.e., when the likelihood of aspiration due to gastroesophageal reflux is not high, it predicts the likelihood of aspiration of oral contents (S31). For example, the control unit 11 reads information such as the amount and time of food orally ingested by the patient, the state of coughing and phlegm production, and oral hygiene from the electronic medical record data stored in the electronic medical record DB 12a, and predicts the likelihood of aspiration of oral contents based on the read information. For example, if the amount of food orally ingested is equal to or greater than a predetermined amount, if the elapsed time since the time of ingestion is within a predetermined time, if the amount of coughing and phlegm production is increasing, if oral care is insufficient, etc., the control unit 11 predicts that aspiration of oral contents is highly likely. For example, the control unit 11 predicts whether aspiration of oral contents is highly likely by calculating a score indicating the likelihood of aspiration of oral contents based on each piece of data read from the electronic medical record DB 12a.

Based on the prediction result, the control unit 11 determines whether or not a high probability of oral content aspiration has been predicted (S32). For example, if a score indicating the probability of oral content aspiration has been calculated, the control unit 11 determines whether the calculated score is equal to or greater than a predetermined value. If it determines that a high probability of oral content aspiration has been predicted (S32: YES), the control unit 11 outputs an alert to notify the patient of the high probability of occurrence (S33). For example, as shown in FIG. 13, the control unit 11 outputs the alert by displaying a message on the electronic medical record screen notifying the patient that aspiration of oral content is highly likely and that confirmation by a doctor is required. This allows a medical professional who has entered patient information into the electronic medical record to understand the high probability of oral content aspiration predicted based on the patient's condition information. The alert message shown in FIG. 13 may also include advice such as providing oral care or swallowing rehabilitation or training.

When an alert such as that shown in FIG. 13 is output, for example, nursing staff will ask a doctor to confirm whether or not aspiration of oral contents has occurred. After confirming with the doctor, the nursing staff will provide the patient with oral care and swallowing rehabilitation or training. If the control unit 11 determines that it has not predicted a high possibility of aspiration of oral contents (S32: NO), that is, if the possibility of aspiration of oral contents is low, it skips the processing of step S33 and ends the series of processes. Here, the control unit 11 may display a message on the display unit 15 notifying that the possibility of aspiration is low or that the patient is in a normal state with almost no possibility of aspiration.

This embodiment achieves the same effects as the first embodiment described above. Furthermore, this embodiment can predict the possibility of aspiration of oral contents, in addition to the possibility of aspiration due to gastroesophageal reflux, and present this to medical professionals. Therefore, based on the presented prediction results, doctors can make a definitive diagnosis by taking chest X-rays or CT images and conducting detailed tests such as blood tests, as necessary.

The configuration of this embodiment is applicable to the information processing device 10 of embodiment 1 and also to the inspection device 20 of embodiment 2, and similar effects can be obtained when applied to the information processing device 10 of embodiment 1 or the inspection device 20 of embodiment 2. Furthermore, the modified examples described in each of the above embodiments can also be applied to this embodiment as appropriate.

The embodiments disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the claims, not by the meaning described above, and is intended to include all modifications within the meaning and scope of the claims.

REFERENCE SIGNS LIST 10 Information processing device 11 Control unit 12 Memory unit 13 Communication unit 14 Input unit 15 Display unit 20 Inspection equipment 21 Control unit 22 Memory unit 25 Microphone 26 Camera 12a Electronic medical record DB
12M Learning Model 22M Learning Model

Claims (10)

obtaining status information regarding the patient's condition;
An information processing method in which a computer executes a process of inputting acquired condition information into a learning model that has been trained to output onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient when condition information of the patient is input , and outputting onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient . the learning model is trained to output onset information indicating a possibility in multiple stages that the patient has developed aspiration of gastroesophageal reflux material;
The information processing method according to claim 1, wherein the computer executes a process of outputting onset information indicating that the patient has developed aspiration of gastroesophageal reflux material if the possibility output from the learning model is equal to or greater than a predetermined value. The information processing method according to claim 2 , wherein the computer executes a process of outputting an alert notifying a doctor that confirmation is required when the possibility output from the learning model is equal to or greater than a predetermined value. obtaining status information regarding the patient's condition;
outputting onset information regarding the onset of aspiration based on the acquired condition information;
An information processing method in which a computer executes a process of outputting information regarding countermeasures to prevent aspiration of gastroesophageal reflux material when onset information indicating a high possibility that the patient has developed aspiration of gastroesophageal reflux material is output, and outputting information regarding countermeasures for oral care when onset information indicating a high possibility that the patient has developed aspiration of oral contents is output. 5. The information processing method according to claim 1, wherein the condition information includes any one of the patient's body temperature, blood oxygen saturation, amount of phlegm, acidity level in the patient's oral cavity, presence or absence of sugar in the oral cavity, presence or absence of a specific odor in the oral cavity, information about the patient's facial expression, and presence or absence of wheezing. obtaining status information regarding the patient's condition;
A program that causes a computer to execute a process of inputting acquired condition information into a learning model that has been trained to output onset information regarding the onset of aspiration of gastroesophageal reflux material in a patient when condition information of the patient is input , and outputting onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient . an acquisition unit that acquires status information regarding the status of a patient;
an output unit that inputs the acquired condition information into a learning model that has been trained to output onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient when condition information of the patient is input , and outputs onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient . obtaining status information regarding the patient's condition;
outputting onset information regarding the onset of aspiration based on the acquired condition information ;
A program that causes a computer to execute a process of outputting information regarding countermeasures to suppress aspiration of gastroesophageal reflux material when onset information indicating a high possibility that the patient has developed aspiration of gastroesophageal reflux material is output, and outputting information regarding countermeasures for oral care when onset information indicating a high possibility that the patient has developed aspiration of oral contents is output . an acquisition unit that acquires status information regarding the status of a patient;
an output unit that outputs onset information regarding the onset of aspiration based on the acquired condition information ;
a treatment information output unit that outputs information regarding countermeasures for preventing aspiration of gastroesophageal reflux matter when onset information indicating that the patient is likely to have developed aspiration of gastroesophageal reflux matter is output, and that outputs information regarding countermeasures for oral care when onset information indicating that the patient is likely to have developed aspiration of oral contents is output;
An information processing device comprising: obtaining training data including patient condition information and onset information relating to onset of gastroesophageal reflux aspiration in the patient;
A model generation method in which a computer executes a process of generating a learning model that uses the acquired training data to output onset information regarding the onset of aspiration of gastroesophageal reflux material in the patient when condition information of the patient is input.