Information processing system, information processing method, and program

Abstract

The purpose of this invention is to analyze the vibrations, including sound emitted by a subject, to detect respiratory diseases in the subject at an early stage and to recommend the subject to a medical institution. The information processing system (100) includes: a signal acquisition unit (201) that acquires a feature signal, including at least one of a heartbeat signal, a respiratory vibration signal, and a body movement signal, and a lung sound signal, extracted from a detection signal output by a sensor (11) that detects vibrations emitted by a subject; a first prediction unit (202) that outputs first prediction information based on the feature signal and the lung sound signal, the first prediction information including at least one of first information related to the possibility that the subject has a respiratory disease and second information related to the respiratory disease suffered by the subject; and an output unit (206) that outputs at least one of the first prediction information and a medical institution recommendation information generated based on the first prediction information.

Description

Technical Field

[0001] This invention relates to a system for detecting vibrations emitted by a subject and an information processing system for detecting respiratory diseases in a subject. Background Technology

[0002] Devices for detecting pneumonia and other conditions based on the sounds emitted by a subject have been developed. For example, Patent Document 1 discloses a method for detecting abnormal breathing sounds by analyzing measured breathing sounds. Additionally, Patent Document 2 discloses a device for detecting cardiac and / or respiratory diseases such as pneumonia, pulmonary edema, and / or heart failure in a subject.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Publication No. 2001-505085

[0006] Patent Document 2: Japanese Patent Publication No. 2018-516616 Summary of the Invention

[0007] The problem the invention aims to solve

[0008] For example, in the case of elderly people with weakened immune systems, respiratory illnesses such as pneumonia can sometimes progress even without symptoms like fever and cough. Therefore, if they do not seek medical attention before symptoms appear, treatment may be delayed, making it difficult to prevent the condition from worsening.

[0009] In order to initiate treatment for respiratory diseases at an early stage, it is hoped that the disease can be detected before the subject notices symptoms, and that recommendations for medical treatment can be made.

[0010] One objective of this invention is to analyze the sound-containing vibrations emitted by a subject to detect respiratory diseases at an early stage and recommend the subject to a medical institution for treatment.

[0011] Problem Solving Methods

[0012] The information processing system of Embodiment 1 of this disclosure comprises: a signal acquisition unit that acquires a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds, the feature signal including at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; a first prediction unit that outputs first prediction information based on the acquired feature signal and the lung sound signal, the first prediction information including at least one of first information related to the likelihood of the subject developing a respiratory disease and second information related to the respiratory disease the subject has; and an output unit that outputs at least one of the first prediction information and consultation recommendation information generated based on the first prediction information, displaying a recommendation to seek medical treatment at a medical institution.

[0013] The information processing system of Embodiment 2 of this disclosure comprises: a signal acquisition unit that acquires a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds, the feature signal including at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; a sleep state determination unit that outputs a determination result related to the subject's sleep state based on the acquired feature signal; a second prediction unit that outputs second prediction information based on the determination result and the lung sound signal, the second prediction information including at least one of third information related to the possibility that the subject has developed a respiratory disease, and fourth information related to the respiratory disease the subject has; and an output unit that outputs at least one of the second prediction information and consultation recommendation information generated based on the second prediction information, displaying a recommendation to visit a medical institution.

[0014] The information processing method of Embodiment 3 of this disclosure includes: a signal acquisition step, acquiring a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds, wherein the feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibration, and a body movement signal displaying the subject's body movement; a first prediction step, outputting first prediction information based on the acquired feature signal and the lung sound signal, wherein the first prediction information includes at least one of first information related to the likelihood of the subject developing a respiratory disease and second information related to the respiratory disease the subject has; and an output step, outputting at least one of the first prediction information and consultation recommendation information generated based on the first prediction information, displaying a recommendation to seek medical treatment at a medical institution.

[0015] The information processing method of Embodiment 4 of this disclosure includes: a signal acquisition step, acquiring a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds, wherein the feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; a sleep state determination step, outputting a determination result related to the subject's sleep state based on the acquired feature signal; a second prediction step, outputting second prediction information based on the determination result and the lung sound signal, wherein the second prediction information includes at least one of a third piece of information related to the likelihood of the subject developing a respiratory disease, and a fourth piece of information related to the respiratory disease the subject has; and an output step, outputting at least one of the second prediction information and a medical recommendation information generated based on the second prediction information, displaying a recommendation to seek medical treatment at a medical institution.

[0016] The information processing systems of various embodiments disclosed herein can also be implemented by a computer. In this case, the control program of the information processing system implemented by the computer as the various parts (software elements) of the information processing system, and the computer-readable recording medium on which the program is recorded, also fall within the scope of this disclosure.

[0017] The effects of the invention

[0018] According to one method of this disclosure, it is possible to analyze the vibrations containing sound emitted by a subject, detect respiratory diseases in the subject at an early stage, and recommend the subject to medical institutions for treatment. Attached Figure Description

[0019] Figure 1This is a conceptual diagram illustrating an example of the configuration of the information processing system in Implementation 1.

[0020] Figure 2 This is a functional block diagram illustrating an example of the structure of the aforementioned information processing system.

[0021] Figure 3 This is a diagram illustrating a simplified configuration of the detection device.

[0022] Figure 4 This is a diagram illustrating the various signals contained in the detected signal.

[0023] Figure 5 This diagram illustrates the various sounds that may be included in lung sounds and the representative cases corresponding to each sound.

[0024] Figure 6 This is a schematic diagram used to illustrate a model of a neuron.

[0025] Figure 7 This is a diagram used to illustrate neural networks.

[0026] Figure 8 This is a flowchart illustrating an example of the processing flow performed by the aforementioned information processing system.

[0027] Figure 9 This is a functional block diagram illustrating an example of the configuration of the aforementioned information processing system, variant 1.

[0028] Figure 10 This is a functional block diagram illustrating an example of the configuration of the above-described information processing system, variant 2.

[0029] Figure 11 This is a functional block diagram illustrating an example of the configuration of the information processing system in Embodiment 2.

[0030] Figure 12 This is a flowchart illustrating an example of the processing flow performed by the aforementioned information processing system.

[0031] Figure 13 This is a functional block diagram illustrating an example of the configuration of the information processing system in Embodiment 3.

[0032] Figure 14 This is an example of a confirmation screen displayed on the display unit.

[0033] Figure 15 This is a functional block diagram illustrating an example of the configuration of the information processing system in Embodiment 4.

[0034] Figure 16 This is a diagram illustrating an example of the sensor configuration in Embodiment 4.

[0035] Symbol Explanation

[0036] 1. 1A Detection Device

[0037] 2, 2B, 2C, 2D, 2E server devices

[0038] 3. 3D Communication Device

[0039] 33 Display Department

[0040] 4. 4E Edge Server Device

[0041] 11 Sensors

[0042] Information processing systems 100, 100A, 100B, 100C, 100D, 100E

[0043] Signal Acquisition Units 101, 201, and 401

[0044] Output sections 102, 206, 403, and 403E

[0045] 202 First Prediction Department

[0046] 203 Diagnostic Results Acquisition Department

[0047] 204 Model Update Department

[0048] 205 Pneumonia Determination Department

[0049] 208 Sleep Status Assessment Department

[0050] 209 Second Forecasting Department

[0051] M2, M2C, M2E prediction models Detailed Implementation

[0052] [Implementation Method 1]

[0053] The following describes one embodiment of the present invention in detail. However, it should be noted that the following description is only used to illustrate one example of the information processing system 100 of the present invention, and the technical scope of the present invention is not limited to the following description and drawings.

[0054] (Overview of Information Processing System 100)

[0055] The information processing system 100 of Implementation Method 1 is a system in medical-related facilities such as hospitals and nursing facilities where doctors, nurses, and caregivers monitor individuals and conduct early detection of respiratory diseases. Individuals can be people receiving care or hospitalized patients. Individuals can be elderly or children. Individuals who wish to detect respiratory diseases early, such as elderly people with weakened immune systems, are also included.

[0056] The information processing system 100 outputs at least one of the following (1) and (2): (1) a first prediction information output based on a characteristic signal and a lung sound signal, which includes at least one of first information related to the likelihood of the subject developing a respiratory disease and second information related to the respiratory disease the subject has; (2) a medical recommendation information generated based on the first prediction information, which displays a recommendation to seek medical treatment at a medical institution. Here, the characteristic signal is a signal extracted from the detection signal output by the sensor 11 that detects vibrations emitted by the subject, which includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibration, and a body movement signal displaying the subject's body movement. The lung sound signal is the lung sound of the subject extracted from the above detection signal.

[0057] In this way, the information processing system 100 extracts characteristic signals closely related to physical condition and lung sound signals closely related to respiratory diseases from the detection signals output by the sensor 11 that detects vibrations emitted by the subject, thereby making predictions related to the likelihood of the subject developing a respiratory disease. Therefore, the information processing system 100 can predict the likelihood of the subject developing a respiratory disease, the type of respiratory disease the subject may have, before the subject shows symptoms, and can output this prediction and medical consultation recommendations based on the prediction. As a result, the information processing system 100 can detect respiratory diseases in subjects at an early stage and make early recommendations for medical consultations.

[0058] (Composition of Information Processing System 100)

[0059] For a brief overview of the information processing system 100, based on Figure 1 and Figure 2 Please provide an explanation. Figure 1 This is a conceptual diagram illustrating an example of the configuration of an information processing system 100. Figure 2 This is a functional block diagram illustrating an example of the configuration of an information processing system 100. For example... Figure 1 and Figure 2 As shown, the information processing system 100 includes a detection device 1, a server device 2, a communication device 3, and an edge server device 4. It should be noted that... Figure 1 The illustrations of server device 2 and edge server device 4 are omitted. Regarding the number of detection device 1, server device 2, communication device 3, and edge server device 4 included in the information processing system 100, each can be one or multiple.

[0060] In the information processing system 100, the edge server device 4 is not essential. For example, in the information processing system 100, the detection device 1 can also have the functions of the edge server device 4 (see reference). Figure 9 Alternatively, in the information processing system 100, the server device 2 may also have the functions of the edge server device 4 (see [reference]). Figure 10 ).

[0061] (Detection device 1)

[0062] The detection device 1 detects the vibrations emitted by the target and outputs the detection signal to an external device. The external device that serves as the destination for the output signal is typically an edge server device 4. The detection device 1 and the edge server device 4 can be directly connected, or they can be configured as follows: Figure 1 and Figure 2 As shown, the connection is made via communication network 9. The type of communication network 9 is not limited; it can be a local area network (LAN) or the Internet.

[0063] When the information processing system 100 has multiple detection devices 1, the server device 2 can be configured to process, store, and manage each of the multiple objects. In this case, medical personnel and others can use the communication device 3 to access the information of the objects managed in the server device 2.

[0064] As an example, the detection device 1 includes a sensor 11, a control unit 10, and a storage unit 12. The detection device 1 outputs a detection signal of the vibration emitted by the object output by the sensor 11 to the edge server device 4. It should be noted that the sensor 11 may not be disposed in the detection device 1, but may be disposed outside the detection device 1 and connected to the detection device 1.

[0065] (Sensor 11)

[0066] Sensor 11 can be a non-contact (non-invasive) sensor capable of detecting vibrations emitted by an object at a location where it does not come into contact with the object. There is no particular limitation on the type of sensor 11. For example, a piezoelectric sensor or a Doppler sensor can be used as sensor 11.

[0067] The sensor 11 may have multiple detection areas for outputting detection signals. When the sensor 11 has multiple detection areas, the sensor 11 may output the detection signals detected in each of the multiple detection areas.

[0068] When the sensor 11 is formed into a thin plate, multiple detection areas can be arranged side by side on the same plane. Figure 3 This is a diagram showing an example of a simplified configuration of the detection device 1. Figure 3The detection device 1 shown in 3001 includes a sensor 11 having a detection area D. Figure 3 The detection device 1 shown in 3002 has a sensor 11 having detection areas D1 to D3 arranged in three columns. Figure 3 The detection device 1 shown in 3003 has a sensor 11 having detection areas D1a to D3d arranged in 4 rows and 3 columns.

[0069] For example, in Figure 3 In the case of the detection device 1 shown in 3002, the detection signals detected in each of the detection areas D1 to D3 are output separately. Similarly, in Figure 3 In the case of the detection device 1 shown in 3003, the detection signals detected in each region of the detection area D1a to D3d are output separately. Each region of the detection area D1a to D3d can be, for example, 10 cm square.

[0070] If a configuration with multiple detection areas is adopted, the state of the object can be accurately determined based on the signal strength of the detection signal detected in each detection area.

[0071] Sensor 11 is preferably a sensor with a wide frequency band capable of detecting vibrations. This eliminates the need for multiple sensors with different frequency bands, making maintenance and management easier and more convenient for medical personnel.

[0072] The sensor 11 is preferably positioned in a location that does not cause discomfort to the person being treated, and is not in contact with them. The sensor 11 can be placed on the bed where the person is lying. Figure 1 As shown, sensor 11 can be placed, for example, between the bed where the person is lying and the mattress on the bed. Sensor 11 can also be placed between the sheet on the mattress and the mattress. In addition, if the person is wearing clothes, sensor 11 can be placed at the top of the bed.

[0073] When positioned in these locations, the sensor 11 can typically be formed as a thin plate (sheet). This allows for the detection of sounds and vibrations emitted by the user of the bed without causing discomfort.

[0074] The detection signal is either the signal (waveform data) of vibrations containing sound emitted by the subject itself, or a signal obtained by amplifying or noise-reducing the signal. Noise reduction processing can be performed, for example, by filtering the region above 2000 Hz. Sensor 11 can detect vibrations in various frequency domains originating from the subject. That is, the detection signal output from sensor 11 is a signal composed of the superposition of multiple vibrations with different frequency characteristics. In one example, the detection signal includes at least one of a heartbeat signal, a respiratory vibration signal, and a body movement signal with a frequency below 20 Hz. An example of the detection signal is shown below. Figure 4 .for Figure 4 The details will be described later.

[0075] In one example, the control unit 10 can be a CPU (Central Processing Unit). The control unit 10 reads the control program, which is software, stored in the storage unit 12, and expands it into memory such as RAM (Random Access Memory) to execute various functions. It should be noted that... Figure 2 For the sake of simplicity, the diagrams of the control program are omitted from the storage units 12 / 22 / 32 / 42 shown. Figure 9 , 10 And 11 also with Figure 2 Similarly, the diagram of the control procedure is omitted.

[0076] like Figure 2 As shown, the control unit 10 includes a signal acquisition unit 101 and an output unit 102. The signal acquisition unit 101 acquires a detection signal from the sensor 11. The output unit 102 outputs the detection signal acquired from the sensor 11 to the edge server device 4.

[0077] (Edge server device 4)

[0078] Edge server device 4 analyzes the vibration detection signal emitted by the display object output from detection device 1, extracts feature signals and lung sound signals, and outputs them to an external device. The external device that serves as the destination for the feature signals and lung sound signals is typically server device 2. Edge server device 4 and server device 2 are as follows... Figure 2 As shown, it is connected via communication network 9. As an example, the edge server device 4 includes a control unit 40 and a storage unit 42.

[0079] In one example, the control unit 40 may be a CPU. The control unit 40 reads the control program, which is stored as software, in the storage unit 42, expands it into a memory such as RAM, and executes various functions. The control unit 40 includes a signal acquisition unit 401, a signal extraction unit 402, and an output unit 403.

[0080] The signal acquisition unit 401 acquires a detection signal from the detection device 1. As described above, the sensor 11 may have multiple detection areas that output detection signals. In this case, the signal acquisition unit 401 can acquire a detection signal in each of the multiple detection areas.

[0081] The signal extraction unit 402 extracts the subject's characteristic signals and lung sound signals from the detection signals. The characteristic signals include at least one of the following: a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements. Additionally, the lung sound signals display the subject's lung sounds.

[0082] The signal extraction unit 402 applies known frequency separation methods to the detection signal to extract feature signals and lung sound signals from the detection signal. Figure 4 It is a diagram illustrating the various signals contained in the detected signal. For example... Figure 4 As shown, the detection signal can be separated into characteristic signals containing heartbeat signals, respiratory vibration signals and body movement signals with different frequency characteristics, and lung sound signals through frequency separation.

[0083] Lung sound signals can be extracted through frequency analysis of internal sound signals. Internal sound signals are signals that show internal sounds generated within the subject's body and are included in the detection signal.

[0084] Figure 5 This diagram illustrates the various sounds that can be included in lung sounds and the representative cases corresponding to each sound. Lung sounds include breath sounds and additional sounds. Compared to healthy individuals, if a subject's breath sounds are diminished or absent, the subject is more likely to have pneumothorax, pleural effusion, atelectasis, or other conditions. Furthermore, if a subject's expiratory interval is prolonged or if abnormal sounds are mixed into their breath sounds, the subject is more likely to have chronic obstructive pulmonary disease (COPD) and bronchial asthma.

[0085] On the other hand, additional sounds include "rales" and "pleural friction rubs," etc. "Rales" include wet rales and dry rales. Wet rales include coarse wet rales and fine wet rales, while dry rales include wheezing, snoring, stridor, and short wheezing. That is, the additional sounds contained in the subject's internal sounds include at least one of the subject's coarse wet rales, fine wet rales, wheezing, snoring, stridor, short wheezing, and pleural friction rubs.

[0086] If coarse moist rales are mixed in with the lung sounds of the subject, the subject may have emphysema or other similar conditions; if fine moist rales are mixed in, the subject may have pneumonia or other similar conditions. If wheezing is mixed in with the lung sounds of the subject, the subject may have bronchial asthma or other similar conditions; if snoring is mixed in, the subject may have chronic bronchitis or other similar conditions.

[0087] In this way, since lung sounds are closely related to respiratory diseases, by using lung sounds as an inferential factor for the likelihood of developing respiratory diseases, the predictive accuracy of the likelihood of a subject developing respiratory diseases can be improved. Furthermore, by employing the information processing system 100 of this disclosure, since the characteristic signals and lung sound signals of the subject can be measured with good accuracy, the likelihood of the subject developing respiratory diseases can be predicted with good accuracy.

[0088] The signal extraction unit 402 can store the feature signals and lung sound signals extracted from the detection signals in the storage unit 42. The output unit 403 outputs the feature signals and lung sound signals extracted by the signal extraction unit 402 to the server device 2. The output unit 403 can also output the feature signals and lung sound signals in each of the multiple detection areas.

[0089] (Server device 2)

[0090] Server device 2 processes information from various devices and sends the processed information to external devices. There can be one or more server devices 2. For example, server device 2 can be a server device implemented through cloud computing. Detection device 1, communication device 3, and edge server device 4 can be directly connected to server device 2, or they can be connected in a manner similar to... Figure 2 As shown, they are connected in a communicative manner via communication network 9.

[0091] Server device 2 includes a control unit 20 and a storage unit 22. In one example, the control unit 20 may be a CPU. The control unit 20 reads a control program, which is software, stored in the storage unit 22, and expands it into a memory such as RAM to execute various functions.

[0092] The control unit 20 includes a signal acquisition unit 201, a first prediction unit 202, a diagnosis result acquisition unit 203, a model update unit 204, a pneumonia determination unit 205, and an output unit 206.

[0093] The signal acquisition unit 201 acquires the feature signal and lung sound signal extracted from the detection signal output by the sensor 11 that emits vibrations emitted by the subject of detection from the edge server device 4.

[0094] Additionally, the signal acquisition unit 201 acquires information related to the sleep state of the subject from the communication device 3. This information includes, for example, information indicating whether the subject is asleep (whether they are in a sleep state). This sleep state information can be input to the communication device 3 and output to the server device 2 by medical personnel monitoring the subject, or it can output sleep determination results from other devices to the server device 2.

[0095] Server device 2 can also determine the subject's sleep state based on the detection signal output from sensor 11. In this case, for example, signal acquisition unit 201 can input the subject's detection signal acquired by signal acquisition unit 201 into a sleep determination model to determine the subject's sleep state. The sleep determination model is generated, for example, by machine learning using training data with the subject's detection signal as the explanatory variable and the determination result related to the subject's sleep state as the response variable. It should be noted that the determination result related to the subject's sleep state can also be a result of sleep / wakefulness determined based on electroencephalogram (EEG).

[0096] The detection signals used for sleep determination preferably include at least one of heartbeat signals, body movement signals, and respiratory vibration signals. For example, the server device 2 can accurately determine the sleep state of a subject by using the time variations of heartbeat signals, body movement signals, and respiratory vibration signals.

[0097] (First Prediction Department 202)

[0098] The first prediction unit 202 generates first prediction information, which includes at least one of first information and second information, based on the acquired feature signals and lung sound signals, and outputs it to the output unit 206. The first information is information related to the likelihood that the subject will develop a respiratory disease. The second information is information related to the respiratory disease the subject has. Here, a typical respiratory disease is pneumonia.

[0099] The first piece of information may include the confidence level of the subject's respiratory illness incidence. The confidence level of the subject's respiratory illness incidence is the probability of developing a respiratory illness, which can be calculated based on past data. Additionally, the first piece of information may include the probability that a subject with a respiratory illness will develop the illness within a given period P1 from the measurement time of the detection signal. The given period P1 can be set to any period, such as 10 days or 1 day. The second piece of information may include information regarding the severity of the subject's respiratory illness.

[0100] The first prediction unit 202 generates first prediction information related to the subject by inputting the feature signal and lung sound signal obtained from the subject into at least one of the prediction models M2 shown below (1), (2) and (3) obtained by machine learning using training data. (1) A prediction model M2 obtained by machine learning using training data with the feature signal and lung sound signal of the subject as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. (2) A prediction model M2 obtained by machine learning using training data with the feature signal and the aforementioned lung sound signal of the subject as explanatory variables and information related to the respiratory disease suffered by the aforementioned subject as a response variable. (3) A prediction model M2 obtained by machine learning using training data with the feature signal and lung sound signal of the subject suffering from a respiratory disease as explanatory variables and information related to whether the subject has developed a disease within a given period P1 from the measurement time of the detection signal as a response variable.

[0101] Here, the sample subject is someone who has previously used sensor 11 to measure the detection signal and who has received a diagnosis from medical personnel related to whether they have a respiratory illness and the severity of that illness. The sample subject may also include the subject themselves.

[0102] It should be noted that, in the following text, the "predictive model M2 obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample as explanatory variables and information related to whether the sample has developed a respiratory disease as the response variable" in (1) above is sometimes referred to as predictive model M21. The "predictive model M2 obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample as explanatory variables and information related to the respiratory disease suffered by the sample as the response variable" in (2) above is sometimes referred to as predictive model M22. The "predictive model M2 obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample suffering from a respiratory disease as explanatory variables and information related to whether the sample has developed a disease within a given period P1 from the measurement time of the detection signal as the response variable" in (3) above is sometimes referred to as predictive model M23. For example, to generate the first information, at least one of predictive model M21 and predictive model M23 can be used. In addition, to generate the second information, predictive model M22 can be used. It should be noted that the information related to the respiratory diseases suffered by the samplers, which serves as the response variable for the predictive model M22, may include information on the severity of the respiratory diseases suffered by the samplers.

[0103] The first prediction unit 202 can generate first prediction information using multiple prediction models M2 with different time series. For example, the first prediction unit 202 can use the current prediction model M2 for disease prediction to predict the probability of a subject having a respiratory disease at the time of measurement of the detection signal, and use the prediction model M2 for future disease prediction to predict the probability of a subject having a respiratory disease within a given period P2 from the time of measurement of the detection signal.

[0104] The current disease prediction model M2 can be, for example, a model obtained by machine learning using training data with the sample's characteristic signals and lung sound signals as explanatory variables and information related to whether the sample had a respiratory disease at the time of signal measurement as the response variable. Alternatively, the future disease prediction model M2 can be, as shown in the following variation 3, a model obtained by machine learning using training data with the sample's characteristic signals and lung sound signals as explanatory variables and information related to whether the sample had a respiratory disease within a given period P2 from the time of signal measurement as the response variable.

[0105] The first prediction unit 202 can further use information related to the sleep state of the subject to generate first prediction information. For example, the first prediction unit 202 can generate first prediction information by using a prediction model M2 obtained by machine learning on training data that includes information related to the sleep state of the subject as explanatory variables in addition to the subject's characteristic signals and lung sound signals.

[0106] Specifically, the first prediction unit 202 can generate first prediction information related to the subject by inputting the feature signal, lung sound signal, and information related to the subject's sleep state obtained from the subject into at least one of the prediction models M2 shown below (1) and (2) obtained by machine learning using training data. (1) A prediction model M2 obtained by machine learning using training data with the subject's feature signal, lung sound signal, and information related to the subject's sleep state as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. (2) A prediction model M2 obtained by machine learning using training data with the subject's feature signal, lung sound signal, and information related to the subject's sleep state as explanatory variables and information related to the respiratory disease the subject has developed as a response variable.

[0107] The location of the prediction model M2 is not limited; it can be as follows: Figure 2 It can be stored in storage unit 22 as shown, or it can be stored in a device other than server device 2.

[0108] In the machine learning used to generate the prediction model M2, well-known machine learning algorithms such as neural networks and support vector machines can be applied. Here, for the case where a neural network is used in the prediction model M2, we use... Figure 6 and Figure 7 This will be explained. For example, a neural network is simulated by an implementation. Figure 6 The model of the neurons shown consists of a neural network processor and memory, etc. Figure 6 This is a schematic diagram used to illustrate a model of a neuron. Figure 7 This is a diagram used to illustrate neural networks.

[0109] A neural network typically consists of an input layer composed of multiple neurons, a hidden layer (intermediate layer) composed of multiple neurons, and an output layer composed of multiple neurons. Neurons, such as... Figure 6 As shown, the output is the result y for multiple inputs x. Each input x is multiplied by its corresponding weight coefficient w. Figure 6 For example, input x1 is multiplied by weight coefficient w1, input x2 by weight coefficient w2, and input x3 by weight coefficient w3. The neuron adds the products of each input and substitutes the result, which takes into account the bias B, into the activation function f, thereby outputting the result y.

[0110] Next, use Figure 7 The neural network, which combines neurons, is explained. Figure 7 This is a schematic diagram illustrating a neural network with an input layer L1, a hidden layer L2, and an output layer L3. Figure 7 In the neural network shown, multiple inputs x are fed into the input layer L1, and the output y is output from the output layer L3. The number of hidden layers in a neural network can also be multiple. Figure 7 In this example, inputs x1 to x3 are multiplied by their corresponding weighting coefficients w. a Then, inputs are sent to three neurons N1a~N1c respectively. Here, neurons N1a~N1c output p respectively. 11 ~p 13 Vector (p) 11 , p 12 , p 13 This can be viewed as a feature vector that extracts the features of the input vector (x1, x2, x3). This feature vector (p) 11 , p 12 , p 13 ) is the feature vector between the input layer L1 and the hidden layer L2.

[0111] p 11 ~p 13 Multiplied by the corresponding weighting coefficient w bThen, inputs are fed into two neurons N2a and N2b respectively. Here, we assume that neurons N2a and N2b output p respectively. 21 and p 22 Vector (p) 21 , p 22 ) is the feature vector between the hidden layer L2 and the output layer L3.

[0112] p 21 and p 22 Multiplied by the corresponding weighting coefficient w c Then, inputs are sent to three neurons N3a~N3c respectively. Neurons N3a~N3c output results y1~y3 respectively.

[0113] The operation of a neural network includes a learning mode and an inference mode. In learning mode, the neural network uses training data containing explanatory and response variables to learn (adjust) the parameters representing the weight coefficients w. In inference mode, the neural network outputs an inference result based on the input data (e.g., feature signals) and the parameters that have been adjusted through learning.

[0114] In learning mode, the error between the output from the output layer L3 and the corresponding response variable is calculated when the explanatory variable contained in the training data is input to the input layer L1, and the parameters are adjusted in a way that reduces the error.

[0115] Any well-known method can be used to adjust the parameters, such as backpropagation. The parameters can be repeatedly adjusted until the error falls within a given range, or until all explanatory variables included in the input training data are included.

[0116] In addition, the first prediction unit 202 can set a threshold for judging the possibility that the subject has a respiratory disease and a threshold for judging whether the subject has a respiratory disease, respectively, for the feature signal and the lung sound signal, and output the first prediction information based on these thresholds and the feature signal and the lung sound signal.

[0117] The diagnosis result acquisition unit 203 acquires a diagnosis result containing at least one of the following: information related to whether the subject has developed a respiratory disease, and information related to the respiratory disease the subject has. For example, a doctor diagnoses whether the subject has developed a respiratory disease, the nature and severity of the subject's respiratory disease, inputs the diagnosis result to the communication device 3, and stores it as diagnosis result information M3 in the storage unit 32. The diagnosis result acquisition unit 203 acquires the subject's diagnosis result from the communication device 3.

[0118] The model update unit 204 updates the prediction model M2 by using machine learning on the training data, which uses the feature signal and lung sound signal used as the output of the first prediction information as explanatory variables and the diagnosis result obtained by the diagnosis result acquisition unit 203 as the response variable.

[0119] Specifically, the model update unit 204 updates the prediction model M21 using training data that uses the feature signals and lung sound signals from the output of the first prediction information as explanatory variables and the information in the diagnostic results related to whether the subject has developed a respiratory disease as a response variable. Additionally, the model update unit 204 updates the prediction model M22 using training data that uses the feature signals and lung sound signals from the output of the first prediction information as explanatory variables and the information in the diagnostic results related to the respiratory disease suffered by the subject as a response variable. Furthermore, the model update unit 204 updates the prediction model M23 using training data that uses the feature signals and lung sound signals from the output of the first prediction information as explanatory variables and the information in the diagnostic results related to whether the subject has developed a respiratory disease within a given period P1 from the measurement time of the detection signal as a response variable.

[0120] When the prediction model M2 is updated, the first prediction unit 202 uses the updated prediction model M2 to generate the first prediction information.

[0121] The pneumonia determination unit 205 determines the state of pneumonia in a subject based on lung sound signals. For example, the pneumonia determination unit 205 is based on... Figure 5 The pneumonia status of the subjects was determined based on the cases shown. The pneumonia status includes at least one of the following: whether or not one has pneumonia and the degree of progression (severity) of the pneumonia.

[0122] The output unit 206 outputs at least one of the first prediction information and the medical visit recommendation information to an external device. The medical visit recommendation information is generated based on the first prediction information and displays a recommendation to a medical institution. The medical visit recommendation information may include information displaying at least one of the urgency and reliability of the medical visit recommendation.

[0123] Urgency can indicate how quickly medical attention is needed. For example, the higher the likelihood of a respiratory illness progressing, the higher the urgency level can be set. Furthermore, the urgency level of the likelihood of developing a respiratory illness in the future can be determined based on the first prediction information from multiple time-series prediction models M2 (such as the current disease prediction model M2 and the future disease prediction model M2).

[0124] Reliability can be information indicating the degree of risk of ignoring a medical recommendation or the necessity of receiving a diagnosis from a healthcare professional. The same applies to urgency and reliability in the following implementations.

[0125] Additionally, the output unit 206 outputs the determination result based on the pneumonia determination unit 205. Furthermore, the output unit 206 can output the presence or absence of abnormal lung sounds in the current subject as information indicating the current subject's status to the communication device 3. The external device that serves as the destination for the information output from the output unit 206 is typically the communication device 3.

[0126] In nursing facilities and the like, it is sometimes undesirable to provide information about whether a patient has developed a respiratory illness, or information related to a patient's respiratory illness. In the information processing system 100, since only medical referral information can be output to the communication device 3, only the information desired by the nursing facility or the like can be output to the communication device 3.

[0127] The communication device 3 communicates with the server device 2, and displays various information on the display unit 33 of the communication device 3 based on the information output from the server device 2. The communication device 3 is typically a computer, smartphone, tablet terminal, etc., used by medical personnel, and may be installed at a nurse's station in a nursing facility. The communication device 3 includes an input unit 31, a control unit 30, a storage unit 32, and a display unit 33. The display unit 33 may also be external to the communication device 3 and connected to it.

[0128] Input unit 31 accepts input operations from medical personnel, and sends signals corresponding to the input operations to control unit 30. The input from medical personnel may include, for example, information related to the subject's sleep state and diagnostic results. As described above, the diagnostic results include at least one of the following: information related to whether the subject has developed a respiratory disease, and information related to the respiratory disease the subject has. For example, a doctor may diagnose the subject and input the information into input unit 31. Input unit 31 may be, for example, a keyboard, touch panel, or mouse.

[0129] In one example, the control unit 30 may be a CPU. The control unit 30 reads the control program, which is software, stored in the storage unit 32, expands it into a memory such as RAM, and executes various functions. The control unit 30 includes an acquisition unit 301 and a display control unit 302.

[0130] The acquisition unit 301 acquires at least one of the first prediction information and the medical visit recommendation information, as well as the pneumonia determination result from the pneumonia determination unit 205, from the server device 2. The acquisition unit 301 can save the acquired information to the storage unit 32.

[0131] The display control unit 302 causes the display unit 33 to display the information acquired by the acquisition unit 301. The display control unit 302 can display different information on the display unit 33 according to the attributes of the person viewing the display unit 33.

[0132] For example, when the display unit 33 is being viewed by a caregiver, the display control unit 302 can display on the display unit 33 a notification regarding whether a medical appointment is required and its reliability, as well as the patient's care history. The patient's care history can, for example, be retrieved from information stored in the storage unit 32 and displayed on the display unit 33.

[0133] When the display unit 33 is viewed by the subject or their family, the display control unit 302 can display the historical trend of the subject's lung sounds and lung sound data of healthy individuals of the same age as the subject, obtained from the server device 2. When the display unit 33 is viewed by a doctor, the display control unit 302 can display the subject's past medical history, as well as the historical characteristic signals and changes in lung sound signals of the subject obtained from the server device 2. The subject's past medical history can be obtained from information stored in the storage unit 32 and displayed on the display unit 33, for example.

[0134] (An example of the operation flow of information processing system 100)

[0135] Figure 8 This is a flowchart illustrating an example of the processing flow performed by the information processing system 100. For example... Figure 8 As shown, the signal acquisition unit 201 acquires the feature signal and lung sound signal extracted by the edge server device 4 (step S1, signal acquisition step). Next, the first prediction unit 202 generates first prediction information based on the feature signal and lung sound signal acquired by the signal acquisition unit 201 and the prediction model M2, and outputs it to the output unit 206 (step S2, first prediction step).

[0136] Next, the diagnosis result acquisition unit 203 acquires a diagnosis result containing at least one of the following: information related to whether the subject has developed a respiratory disease, and information related to the respiratory disease the subject has (step S3).

[0137] The model update unit 204 updates the prediction model M2 by using machine learning of the training data, with the feature signal and lung sound signal used as explanatory variables for the output of the first prediction information and the diagnosis result obtained by the diagnosis result acquisition unit 203 as the response variable (step S4).

[0138] Next, the pneumonia determination unit 205 determines the status of pneumonia based on lung sound signals (step S5). The output unit 206 outputs at least one of the first prediction information and the medical visit recommendation information, and the determination result of the pneumonia status based on the pneumonia determination unit 205, to the communication device 3 (step S6, output step).

[0139] In the information processing system 100, the sensor 11 can acquire the subject's detection signal at any time while the subject is present on the sensor 11. Therefore, the information processing system 100 can monitor the subject's lung sounds non-invasively for extended periods. Thus, according to the information processing system 100, early detection of respiratory diseases such as pneumonia, or the risk of developing respiratory diseases and the risk of severe illness, is possible, and recommendations for medical attention are made to caregivers and other monitors. Therefore, elderly individuals and other subjects can seek medical attention at medical institutions in the early stages of respiratory diseases such as pneumonia. Consequently, the information processing system 100 can prevent the subject from becoming severely ill, while simultaneously increasing facility revenue and reducing the burden on caregivers.

[0140] [Variation Example 1]

[0141] The information processing system 100A of Embodiment 1, Variation 1, will be described below. It should be noted that, for ease of explanation, in the following variations and embodiments, components having the same functions as those described above will be labeled with the same reference numerals, and their descriptions will be omitted.

[0142] Figure 9 This is a functional block diagram illustrating an example of the configuration of the information processing system 100A in Modified Example 1. For example... Figure 9 As shown, the information processing system 100A differs from the information processing system 100 in that it includes a detection device 1A instead of a detection device 1, and it does not include an edge server device 4; otherwise, the configurations are the same. In the information processing system 100A, since feature signals and lung sound signals can be extracted from the detection signal through the detection device 1A, the edge server device 4 can be omitted.

[0143] The detection device 1A differs from the detection device 1 in that the control unit 10A includes a signal extraction unit 103. The signal extraction unit 103 has a signal extraction unit 103 that... Figure 2 The signal extraction unit 402 of the edge server device 4 shown in the figure has the same function. The output unit 102 outputs the feature signal and lung sound signal extracted by the signal extraction unit 103 to the server device 2.

[0144] [Variation Example 2]

[0145] The information processing system 100B of the modified example 2 of embodiment 1 will be described below. Figure 10 This is a functional block diagram illustrating an example of the configuration of the information processing system 100B in Modified Example 2. For example... Figure 10 As shown, the information processing system 100B differs from the information processing system 100 in that it includes a server device 2B instead of a server device 2, and does not include an edge server device 4; otherwise, the configurations are the same. In the information processing system 100B, since the server device 2B can be used to extract feature signals and lung sound signals from the detection signals, the edge server device 4 can be omitted.

[0146] Server device 2B differs from server device 2 in that its control unit 20B includes a signal extraction unit 207. The signal extraction unit 207 has the same characteristics as... Figure 2 The signal extraction unit 402 of the edge server device 4 shown in the figure has the same function. In Modification 2, the signal acquisition unit 201 acquires the detection signal from the detection device 1.

[0147] [Variation Example 3]

[0148] The first prediction information may further include: first supplementary information related to the probability that a subject who does not have a respiratory disease will develop a respiratory disease within a given period P2 from the measurement time of the detection signal. The first prediction unit 202, for example, outputs the first supplementary information related to the subject by inputting the feature signal and lung sound signal obtained from the subject into the prediction model M2 obtained by machine learning using the training data described below.

[0149] The training data for machine learning used to predict model M2 consists of characteristic signals and lung sound signals of samples without respiratory diseases as explanatory variables, and information related to whether the sample has contracted a respiratory disease within a given period P2 from the measurement time of the detection signal as the response variable.

[0150] The response variable in the training data of the prediction model M2 used in Variation Example 3, which includes "information related to whether a sample without respiratory disease has contracted a respiratory disease within a given period P2 from the measurement time of the detection signal", includes: (1) information that the sample contracted and developed a respiratory disease within a given period P2 from the measurement time of the detection signal (information IN1); (2) information that the sample contracted the disease but did not develop symptoms within a given period P2 from the measurement time of the detection signal (information IN2); and (3) information that the sample did not contract the disease within a given period P2 from the measurement time of the detection signal (information IN3).

[0151] Information IN1 to information IN3 are information related to the likelihood of the sample subject developing a disease in the future. For example, by using information related to the likelihood of the sample subject developing a disease in the future as a response variable, the information processing system 100 can predict the likelihood of the subject developing a respiratory disease within a given period P2 from the measurement time of the detection signal, as well as the likelihood of developing a respiratory disease.

[0152] The given period P2 can be set to any period, such as 10 days or 1 day. For example, by setting the given period P2 to 10 days, the information processing system 100 can predict in advance the likelihood of the subject developing a respiratory disease. Conversely, by setting the given period P2 to 1 day, the information processing system 100C can predict the likelihood of the subject soon developing a respiratory disease.

[0153] Additionally, the diagnostic result acquisition unit 203 acquires a diagnostic result containing information related to whether a subject without a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal. "Information related to whether a subject without a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal" includes, for example, diagnostic results related to (1) the presence or absence of a respiratory disease in a subject without a respiratory disease within a given period P2 from the measurement time of the detection signal, and (2) the presence or absence of a respiratory disease in a subject without a respiratory disease within a given period P2 from the measurement time of the detection signal. Furthermore, the diagnostic result may also include time information regarding the time when the subject became ill or developed the disease.

[0154] In this case, the model update unit 204 uses training data to update the prediction model M2. The training data uses the feature signal and lung sound signal used for the output of the first prediction information as explanatory variables, and the information in the diagnosis results related to whether the subject who does not have a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal as the response variable.

[0155] Therefore, machine learning is performed on the prediction model M2, incorporating diagnostic results such as the presence or absence of respiratory diseases in the subjects within a given period P2 from the measurement time of the detection signal. Thus, in the future, it will be possible not only to improve the prediction accuracy of the information processing system 100 for respiratory diseases of the subjects at the measurement time of the detection signal, but also to improve the prediction accuracy for respiratory diseases of the subjects within a given period P2 from the measurement time of the detection signal.

[0156] In addition, the first prediction unit 202 can set a threshold for judging the likelihood of a subject developing a respiratory disease within a given period P2 based on the feature signal and lung sound signal, and output the first prediction information based on the threshold and the feature signal and lung sound signal.

[0157] [Implementation Method 2]

[0158] Other embodiments of the present invention will be described below. Figure 11 This is a functional block diagram illustrating an example of the configuration of the information processing system 100C according to Embodiment 2. For example... Figure 11 As shown, the information processing system 100C differs from the information processing system 100 in that it has a server device 2C instead of a server device 2, but the other components are the same.

[0159] (Server Device 2C)

[0160] The server device 2C differs from server device 2 in that it has a control unit 20C instead of a control unit 20, but the other configurations are the same. The control unit 20C includes a signal acquisition unit 201, a sleep state determination unit 208, a second prediction unit 209, a diagnosis result acquisition unit 203, a model update unit 204, a pneumonia determination unit 205, and an output unit 206. Similar to embodiment 1, the signal acquisition unit 201 acquires feature signals and lung sound signals extracted from the detection signals output by the sensor 11 that detects vibrations emitted by the subject from the edge server device 4.

[0161] The sleep state determination unit 208 determines the sleep state of the subject based on the feature signals acquired by the signal acquisition unit 201, and outputs the determination result related to the sleep state of the subject to the second prediction unit 209.

[0162] Sleep states include, for example, sleep and wakefulness. Sleep states can include, for example, sleep quality (light sleep, deep sleep, etc.), changes in sleep patterns, and the time of falling asleep. When determining the sleep state of a subject, the sleep state determination unit 208 can determine the sleep state by inputting the subject's feature signals acquired by the signal acquisition unit 201 into a prediction model obtained through machine learning, using training data with the subject's feature signals as explanatory variables and the subject's sleep state as a response variable. It should be noted that the sleep / wakefulness determination result based on the subject's electroencephalogram (EEG) can also be used as the response variable.

[0163] The second prediction unit 209 generates second prediction information, including at least one of the third and fourth information, based on the determination result of the sleep state determination unit 208 and the lung sound signal, and outputs it to the output unit 206. The third information is information related to the likelihood that the subject will develop a respiratory disease. The fourth information is information related to the respiratory disease that the subject has.

[0164] The third piece of information may include the confidence level of the subject's respiratory illness. Additionally, the third piece of information may include the probability that a subject with a respiratory illness will develop the illness within a given period P1 from the measurement time of the detection signal. The given period P1 can be any period, such as 10 days or 1 day. The fourth piece of information may include information about the severity of the subject's respiratory illness.

[0165] The second prediction unit 209 generates second prediction information related to the subject by inputting at least one of the judgment results based on the feature signal obtained from the subject and the lung sound signal into at least one of the prediction models M2C shown below (1), (2) and (3) obtained by machine learning using training data. (1) A prediction model M2C obtained by machine learning using training data with the judgment results based on the feature signal of the subject and the lung sound signal as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. (2) A prediction model M2C obtained by machine learning using training data with the judgment results based on the feature signal of the subject and the lung sound signal as explanatory variables and information related to the respiratory disease suffered by the subject as a response variable. (3) A prediction model M2C obtained by machine learning using training data with the judgment results based on the feature signal of the subject suffering from a respiratory disease and the lung sound signal as explanatory variables and information related to whether the subject has developed a disease within a given period P1 from the measurement time of the detection signal as a response variable.

[0166] It should be noted that, in the following text, the "predictive model M2C obtained by machine learning using training data with the judgment results of the characteristic signals of the sampler and the lung sound signals as explanatory variables and information related to whether the sampler has developed a respiratory disease as the response variable" in (1) above is sometimes referred to as predictive model M21C. The "predictive model M2C obtained by machine learning using training data with the judgment results of the characteristic signals of the sampler and the lung sound signals as explanatory variables and information related to the respiratory disease suffered by the sampler as the response variable" in (2) above is sometimes referred to as predictive model M22C. The "predictive model M2C obtained by machine learning using training data with the judgment results of the characteristic signals of the sampler suffering from a respiratory disease and the lung sound signals as explanatory variables and information related to whether the sampler has developed a disease within a given period P1 from the measurement time of the detection signal as the response variable" in (3) above is sometimes referred to as predictive model M23C. For example, to generate the third piece of information, at least one of predictive model M21C and predictive model M23C can be used. On the other hand, to generate the fourth piece of information, the predictive model M22C can be used. It should be noted that the information related to the respiratory disease suffered by the samplers, which serves as the response variable of the predictive model M22C, may include information on the severity of the respiratory disease suffered by the samplers.

[0167] The second prediction unit can further utilize the feature signals to output second prediction information. For example, the second prediction unit 209 can generate the second prediction information using a prediction model M2C obtained through machine learning, which uses the sleep state determination unit 208 based on the feature signals of the sampler and the lung sound signal as explanatory variables, and at least one of the following as response variables: information related to whether the sampler has developed a respiratory disease and information related to the respiratory disease the sampler has. In this case, by inputting the feature signals obtained from the subject and at least one of the judgment results of the sleep state determination unit 208 based on the feature signals, as well as the lung sound signal, into the prediction model M2C, the second prediction unit 209 generates second prediction information related to the subject.

[0168] The model update unit 204 updates the prediction model M2C by using machine learning of the training data. The training data uses the judgment result of the sleep state determination unit 208 based on feature signals and the lung sound signal as explanatory variables, and the diagnosis result obtained by the diagnosis result acquisition unit 203 as the response variable.

[0169] Specifically, the model update unit 204 updates the prediction model M21C using training data that includes the judgment result and lung sound signal used for the output of the second prediction information as explanatory variables and information in the diagnosis result related to whether the subject has developed a respiratory disease as a response variable. Additionally, the model update unit 204 updates the prediction model M22C using training data that includes the judgment result and lung sound signal used for the output of the second prediction information as explanatory variables and information in the diagnosis result related to the respiratory disease suffered by the subject as a response variable. Furthermore, the model update unit 204 updates the prediction model M23C using training data that includes the judgment result and lung sound signal used for the output of the second prediction information as explanatory variables and information in the diagnosis result related to whether the subject has developed a respiratory disease within a given period P1 from the measurement time of the detection signal as a response variable.

[0170] When the prediction model M2C is updated, the second prediction unit 209 uses the updated prediction model M2C to generate the second prediction information.

[0171] The output unit 206 outputs at least one of the second prediction information and the medical visit recommendation information to the communication device 3. The medical visit recommendation information is generated based on the second prediction information and displays a recommendation to a medical institution. The medical visit recommendation information may include information displaying at least one of the urgency and reliability of the medical visit recommendation. In addition, the output unit 206 outputs the determination result based on the pneumonia determination unit 205.

[0172] (The operation flow of Information Processing System 100C)

[0173] Figure 12 This is a flowchart illustrating an example of the processing flow performed by the information processing system 100C. For example... Figure 12 As shown, the signal acquisition unit 201 acquires the feature signal and lung sound signal extracted by the edge server device 4 (step S11, signal acquisition step).

[0174] Next, the sleep state determination unit 208 determines the sleep state of the subject based on the feature signals acquired by the signal acquisition unit 201, and outputs the determination result to the second prediction unit 209 (step S12, sleep state determination step).

[0175] Next, the second prediction unit 209 generates second prediction information using the prediction model M2C based on the judgment result of the sleep state determination unit 208 and the lung sound signal acquired by the signal acquisition unit 201, and outputs it to the output unit 206 (step S13, second prediction step).

[0176] Next, the diagnosis result acquisition unit 203 acquires a diagnosis result containing at least one of the following: information related to whether the subject has developed a respiratory disease, and information related to the respiratory disease the subject has (step S14).

[0177] The model update unit 204 updates the prediction model M2C by using machine learning of the training data. The training data uses the judgment result of the sleep state determination unit 208 based on feature signals and the lung sound signal as explanatory variables, and the diagnosis result obtained by the diagnosis result acquisition unit 203 as the response variable (step S15).

[0178] Next, the pneumonia determination unit 205 determines the status of pneumonia based on lung sound signals (step S16). The output unit 206 outputs at least one of the second prediction information and the medical visit recommendation information, and the determination result of the pneumonia status based on the pneumonia determination unit 205, to the communication device 3 (step S17, output step).

[0179] [Variation Example 4]

[0180] The second prediction unit 209 can further use the attribute information of the subject to generate second prediction information and output it to the output unit 206. For example, the second prediction unit 209 can use a prediction model M2C obtained by machine learning with the determination result of the sleep state determination unit 208 based on the characteristic signal of the sample subject, the lung sound signal and the attribute information of the sample subject as explanatory variables to generate second prediction information related to the subject.

[0181] In this scenario, by inputting the judgment result of the sleep state determination unit 208 based on feature signals obtained from the subject, lung sound signals, and the subject's attribute information into the prediction model M2C, the second prediction unit 209 generates second prediction information related to the subject. It should be noted that the response variable of this training data is at least one of information related to whether the subject has developed a respiratory disease and information related to the respiratory disease the subject has.

[0182] Attribute information may include, for example, at least one of the following: physical condition, age, gender, presence or absence of smoking history, duration of smoking, and frequency of smoking. Server device 2 may, for example, retrieve attribute information input and stored by medical personnel of the monitored subject to communication device 3 from storage unit 32.

[0183] The attributes of the subject can sometimes affect the detection signal. Therefore, by generating a second predictive information by taking into account the subject's attribute information, the predictive accuracy of the subject's likelihood of developing respiratory diseases, etc., can be improved.

[0184] [Variation Example 5]

[0185] The second prediction unit 209 can further use medical history information related to the subject's past respiratory diseases to generate second prediction information and output it to the output unit 206. For example, the second prediction unit 209 can use a prediction model M2C obtained by machine learning with the determination result of the sleep state determination unit 208 based on the characteristic signals of the sample, lung sound signals, and medical history information related to the sample's past respiratory diseases as explanatory variables, thereby generating second prediction information related to the subject.

[0186] In this scenario, by inputting the judgment result of the sleep state determination unit 208 based on feature signals obtained from the subject, lung sound signals, and medical history information related to the subject's past respiratory diseases into the prediction model M2C, the second prediction unit 209 generates second prediction information related to the subject. It should be noted that the response variable of this training data is at least one of information related to whether the subject has developed a respiratory disease and information related to the respiratory disease the subject has.

[0187] For example, server device 2 can obtain medical history information input and stored by medical personnel of the monitored subject to communication device 3 from storage unit 32. If the subject has a history of respiratory illness, they are more likely to develop pneumonia. Therefore, by generating second predictive information that considers medical history information related to past respiratory illnesses, the accuracy of predicting the subject's likelihood of developing respiratory illnesses can be improved.

[0188] The output unit 206 can output information about the subject's past respiratory illnesses along with the second prediction information.

[0189] [Variation Example 6]

[0190] The second prediction unit 209 can further use nursing history information to generate second prediction information and output it to the output unit 206. This nursing history information shows the execution history of nursing care provided by the subject to reduce the likelihood of developing respiratory diseases. For example, the second prediction unit 209 can use a prediction model M2C obtained by machine learning with the determination result of the sleep state determination unit 208 based on the characteristic signals of the sample, lung sound signals, and the nursing history information of the sample as explanatory variables, thereby generating second prediction information related to the subject.

[0191] In this scenario, by inputting the judgment result of the sleep state determination unit 208 based on feature signals obtained from the subject, lung sound signals, and the subject's nursing history information into the prediction model M2C, the second prediction unit 209 generates second prediction information related to the subject. It should be noted that the response variable of this training data is at least one of information related to whether the subject has developed a respiratory disease and information related to the respiratory disease the subject has.

[0192] Nursing history information includes, for example, information on past care for preventing aspiration, such as whether or not saliva was expelled and post-meal oral care. Server device 2 can, for example, obtain nursing history information input and stored by medical personnel of the monitored subject to communication device 3 from storage unit 32.

[0193] Without proper care to suppress aspiration, individuals are more susceptible to developing pneumonia. Therefore, by generating second predictive information that considers nursing history, the accuracy of predictions regarding the likelihood of developing respiratory illnesses can be improved. The output unit 206 can output the individual's nursing history information along with the second predictive information.

[0194] [Variation Example 7]

[0195] The second prediction information may further include: second supplementary information related to the probability that a subject who does not have a respiratory disease will develop a respiratory disease within a given period P2 from the measurement time of the detection signal. In this case, the second prediction unit 209 outputs the second supplementary information related to the subject, for example, by inputting the judgment result based on the feature signal and the lung sound signal obtained from the subject into the prediction model M2C obtained by machine learning using the training data described below.

[0196] The training data for the machine learning of the prediction model M2C used in Variation Example 7 consists of the judgment results based on the characteristic signals of the sample without respiratory diseases and the lung sound signals as explanatory variables, and the data related to whether the sample had a respiratory disease within a given period P2 from the measurement time of the detection signal as the response variable.

[0197] The response variable in the training data of the predictive model M2C in Variation 7, which includes "information related to whether a sample without respiratory disease has contracted a respiratory disease within a given period P2 from the measurement time of the detection signal", includes: (1) information that the sample contracted and developed a respiratory disease within a given period P2 from the measurement time of the detection signal (information IN1); (2) information that the sample contracted the disease but did not develop symptoms within a given period P2 from the measurement time of the detection signal (information IN2); and (3) information that the sample did not contract the disease within a given period P2 from the measurement time of the detection signal (information IN3).

[0198] Information IN1 to information IN3 are information related to the likelihood of the sample subject developing a disease in the future. For example, by using information related to the likelihood of the sample subject developing a disease in the future as a response variable, the information processing system 100C can predict the likelihood of the subject developing a respiratory disease within a given period P2 from the measurement time of the detection signal, as well as the likelihood of developing a respiratory disease.

[0199] The given period P2 can be set to any period, such as 10 days or 1 day. For example, by setting the given period P2 to 10 days, the information processing system 100C can predict early on the likelihood of a subject developing a respiratory illness. Conversely, by setting the given period P2 to 1 day, the information processing system 100C can predict the likelihood of a subject soon developing a respiratory illness.

[0200] Additionally, the diagnostic result acquisition unit 203 acquires a diagnostic result containing information related to whether a subject without a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal. "Information related to whether a subject without a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal" includes, for example, diagnostic results related to (1) the presence or absence of a respiratory disease in a subject without a respiratory disease within a given period P2 from the measurement time of the detection signal, and (2) the presence or absence of a respiratory disease in a subject without a respiratory disease within a given period P2 from the measurement time of the detection signal. Furthermore, the diagnostic result may also include time information regarding the time when the subject became ill or developed the disease.

[0201] In this case, the model update unit 204 uses training data to update the prediction model M2C. The training data uses the judgment result based on the feature signal and the lung sound signal used for the output of the second prediction information as explanatory variables, and the information in the diagnosis result related to whether the subject who does not have a respiratory disease has developed a respiratory disease within a given period P2 from the measurement time of the detection signal as the response variable.

[0202] Therefore, the prediction model M2C incorporates diagnostic results such as the presence or absence of respiratory diseases in the subjects within a given period P2 from the measurement time of the detection signal, and performs machine learning. Thus, in the future, it will be possible not only to improve the prediction accuracy of the information processing system 100C for respiratory diseases of the subjects at the measurement time of the detection signal, but also to improve the prediction accuracy for respiratory diseases of the subjects within a given period P2 from the measurement time of the detection signal.

[0203] [Implementation Method 3]

[0204] Doctors and others diagnose a patient's illness based on whether the sounds (so-called auscultatory sounds) emitted from one or more parts of the patient's body (hereinafter referred to as auscultation sites) are abnormal. When using sensor 11 to detect the sounds emitted by the patient, the site where sensor 11 detects the sounds may sometimes differ from the auscultation site used by doctors and others to make diagnoses related to the patient's illness. That is, there is a possibility that the sound corresponding to the detection signal detected by sensor 11 is different from the auscultatory sounds familiar to doctors and others. In this case, even if doctors and others hear the sound detected by sensor 11, because the sound is different from auscultatory sounds, it is difficult for them to make diagnoses related to the patient's illness based on that sound.

[0205] In cases where a subject relapses into a respiratory illness of the same severity as a previous illness, the auscultation sounds confirmed by doctors during the previous examination are similar to those confirmed during the re-emergence examination. However, the detection signal detected by sensor 11 may differ even for the same subject and with the same degree of respiratory illness, depending on the environment in which sensor 11 is used. For example, the detection signal detected by sensor 11 may vary depending on the bed used by the subject, the subject's clothing, and the operating status of the air conditioning.

[0206] Therefore, in order to enable doctors and others to understand the respiratory diseases and risks of the subjects determined based on the detection results of sensor 11, it is effective to enable doctors and others to grasp the correspondence between the detection signals detected by sensor 11 or the relevant information of the respiratory diseases of the subjects based on the detection signals and the auscultation sounds familiar to doctors and others.

[0207] Therefore, the information processing system 100D in Embodiment 3 has the following function: converting the detection signal obtained from the subject using the sensor 11 into an auscultatory equivalent sound that corresponds to the auscultatory sound that can be heard when listening to the lung sounds of the subject using a stethoscope, and reproducing the converted auscultatory equivalent sound. By reproducing the auscultatory equivalent sound, the information processing system 100D makes it easier for medical personnel to grasp the correspondence between the detection signal detected in the information processing system 100D and the sound that medical personnel can hear when listening to the lung sounds with a stethoscope.

[0208] Figure 13 This is a block diagram illustrating the configuration of the information processing system 100D according to Embodiment 3. Hereinafter, using... Figure 13 The details of the information processing system 100D in Embodiment 3 will be described below. Figure 13 As shown, the information processing system 100D differs from the information processing system 100 of Embodiment 1 in that it includes a server device 2D and a communication device 3D instead of a server device 2 and a communication device 3.

[0209] like Figure 13 As shown, the server device 2D includes a control unit 20D and a storage unit 22. The control unit 20D includes the processing units and the audio conversion unit 210 included in the control unit 20 of the server device 2 of Embodiment 1.

[0210] The sound conversion unit 210 acquires the detection signal of the subject detected by the sensor 11 and converts the acquired detection signal into an auscultatory equivalent sound. An auscultatory equivalent sound is a sound that is similar to the auscultation sound (e.g., lung sound) that can be heard when a person is auscultated with a stethoscope. For example, an auscultatory equivalent sound can be at least one of the following (1) to (3).

[0211] (1) Sound extracted from the intervals with less noise, such as additional sound, in the detected signal.

[0212] (2) The sound is divided into segments of the detection signal by specific time periods (e.g., time periods with higher or lower signal values ​​compared to a given threshold).

[0213] (3) The sound formed by segmenting the detection signal according to the presumed sound type (coarse wet rales, fine wet rales or wheezing, etc.), or the sound in the detection signal that emphasizes at least one of the above sound types.

[0214] As an example, the tone conversion unit 210 converts the detection signal into a signal corresponding to a stethoscope-equivalent tone by making the frequency characteristics of the detection signal approximate the frequency characteristics of the sound heard when listening with a stethoscope. For example, the tone conversion unit 210 converts the detection signal into a stethoscope-equivalent tone by amplifying or attenuating the detection signal at each frequency, taking into account the characteristics of the sensor 11 and in a way that makes the detection signal match the sound of a stethoscope. Specifically, the tone conversion unit 210 amplifies the signal value in the 200-300Hz range of the detection signal by a factor of 1.6 and attenuates the signal value in the 500-600Hz range by a factor of 0.9.

[0215] Alternatively, the tone conversion unit 210 can generate auscultatory equivalent sounds based on the detection signal. Specifically, the tone conversion unit 210 generates auscultatory equivalent sounds by inputting the detection signal obtained from the subject into a prediction model that takes the detection signal as input data and the auscultatory equivalent sounds as output data. Here, the prediction model can, for example, use machine learning with training data that takes the detection signal obtained from the sensor 11 from a patient suffering from a respiratory disease of various degrees of severity (e.g., pneumonia) as the explanatory variable and the auscultatory sound data when the patient is auscultated with a stethoscope as the response variable.

[0216] Furthermore, the server device 2D can store multiple sample sound data of typical auscultatory sounds obtained in advance from patients with respiratory diseases of various degrees of severity, and corresponding data for respiratory diseases determined based on the detection results of sensor 11. In this case, the sound conversion unit 210 can select the sample sound data corresponding to the detection signal as the auscultatory equivalent sound after converting the detection signal, based on the detection signal. For example, if the detection signal includes a signal showing coarse moist rales, the sound conversion unit 210 can select sample sound data containing typical coarse moist rales as the auscultatory equivalent sound.

[0217] 3D communication devices are terminal devices such as computers used by medical personnel. For example... Figure 13 As shown, the communication device 3D includes a control unit 30D, an input unit 31D, a storage unit 32, a display unit 33, and a playback unit 34. The playback unit 34 is a sound playback device such as headphones, capable of playing sound according to the control of the playback control unit 303 (described later). The playback unit 34 can be installed in the communication device 3D or connected to the communication device 3D in a communicable manner. The communication device 3D can play stethoscope-like sounds generated or selected by conversion of detection signals in the server device 2D.

[0218] like Figure 13As shown, the control unit 30D includes an acquisition unit 301, a display control unit 302D, and a reproduction control unit 303. The display control unit 302D causes the display unit 33 to display a confirmation screen. The confirmation screen is a screen that displays data of auscultatory equivalent sounds generated or selected in the server device 2.

[0219] Figure 14 This is an example of a confirmation screen. (For example...) Figure 14 As shown, the confirmation screen contains an anomaly coordinate graph (symbol 1401) with the anomaly score as the vertical axis and the moment when the signal was detected by sensor 11 as the horizontal axis. The anomaly score is a value representing the degree of anomaly of the sound detected from the subject. The anomaly score can be the signal value of the detected signal or the signal value of the lung sound signal extracted from the detected signal.

[0220] like Figure 14 As shown, the confirmation screen contains icons (symbols 1402 to 1404) for a regeneration button used to reproduce auscultatory sounds. The regeneration button is displayed in the anomaly degree coordinate graph corresponding to the area of ​​time period (hereinafter referred to as the anomaly range) where there is a possibility that anomalies such as additional sounds have occurred in the sound emitted by the subject.

[0221] For example, if the subject's breath sounds include additional sounds related to respiratory diseases, the abnormality score increases at the moment the sound is detected. Regarding the display control unit 302D, if the signal value of the detection signal or lung sound signal exceeds a given threshold, the time period exceeding that threshold (symbols 1405 to 1407) is defined as an abnormal range, and icons of the corresponding regeneration buttons (symbols 1402 to 1404) are displayed on the confirmation screen. Alternatively, the display control unit 302D may define all time periods in the detection signal from which the additional sounds were extracted as abnormal ranges.

[0222] The icons for the regeneration buttons were mapped to the corresponding auscultatory sounds within the time period of the abnormal range associated with each button. For example, in Figure 14 In the confirmation screen shown, the abnormal range indicated by symbol 1405 corresponds to the time period before and after 0:10. In this case, the regeneration button (symbol 1402) corresponding to this abnormal range is associated with the following sound data: sound data of the auscultatory equivalent tone obtained by converting the detection signal detected at least at 0:10.

[0223] Here, when the confirmation screen is displayed, the input unit 31D can receive an input operation to select a playback button included in the confirmation screen. When the input unit 31D receives input to select a playback button, the playback control unit 303 controls the playback unit 34, which may be a speaker or the like, to reproduce the stethoscope tone corresponding to that playback button. For example, in Figure 14 In the example shown, when the replay button indicated by symbol 1402 is selected, the replay control unit 303 replays the auscultatory tone corresponding to the abnormal range indicated by symbol 1405. The replay control unit 303 can replay the auscultatory tone converted from the detection signal within the abnormal range, or it can replay the auscultatory tone converted from the detection signal detected within a given period before and after the abnormal range, in addition to the detection signal within the abnormal range.

[0224] Values ​​within the abnormal range indicate signal values ​​that exceed a given threshold range for the detection signal or lung sound signal. Therefore, during time periods encompassing the abnormal range, there is a high probability that the subject will emit sounds containing additional sounds related to respiratory illnesses. Thus, by having the regeneration control unit 303 regenerate the auscultatory equivalent sound corresponding to the detection signal within the time period confirmed to have peaked, medical personnel can confirm sounds emanating from the subject that contain additional sounds.

[0225] It should be noted that, as Figure 14 As shown, the confirmation screen can display an icon (symbol 1408) for a regeneration button that is different from the regeneration button corresponding to the abnormal range. This regeneration button is used to regenerate auscultatory sounds that correspond to the patient's normal lung sounds. The input unit 31D can also receive the selection of this regeneration button, and when the regeneration button is selected, the regeneration control unit 303 can regenerate auscultatory sounds that correspond to the patient's normal lung sounds. Thus, medical personnel can hear auscultatory sounds that correspond to the patient's normal lung sounds and auscultatory sounds that correspond to the patient's lung sounds when the patient makes a sound including additional sounds.

[0226] Furthermore, after the regeneration control unit 303 regenerates the auscultatory tone, the input unit 31D can receive input operations from medical personnel. For example, the input unit 31D can receive input operations for setting a threshold in the abnormality score. If the threshold is changed, the display control unit 302D displays a confirmation screen according to the changed threshold. For example, if the threshold increases, the display control unit 302D defines the score range exceeding the increased threshold as an abnormal range and displays the regeneration button corresponding to that abnormal range.

[0227] Additionally, the input unit 31D can receive input from medical personnel who have confirmed the auscultation results of the subject. In this case, the control unit 30D can send the received information as diagnostic result information to the server device 2D. Upon receiving the diagnostic result information, the diagnostic result acquisition unit 203 of the server device 2D can update the predictive model used to estimate the subject's medical recommendation information based on this information and the subject's detection signal or lung sound signal.

[0228] Generally, medical personnel assess a patient's condition based on the sounds produced when a stethoscope is used. Therefore, it is sometimes difficult to determine a patient's condition solely based on the anomaly score (detection signal or lung sound signal) determined by the information processing system 100D. Furthermore, even if first or second predictive information is generated based on lung sound signals and presented to medical personnel, there is a possibility that the medical personnel may perceive a low level of trust in this information.

[0229] Here, in the information processing system 100D, a sound corresponding to the sound heard during auscultation, i.e., an auscultatory equivalent sound, can be generated and reproduced based on the detection signal. This makes it easier for medical personnel to grasp the patient's condition. Furthermore, by reproducing the auscultatory equivalent sound, the information processing system 100D can improve the reliability of predictive information generated based on the detection signal corresponding to the auscultatory equivalent sound.

[0230] Furthermore, it is practically difficult for medical personnel to constantly monitor the condition of patients. Here, the information processing system 100D can continuously detect the sounds emitted by the patient via sensor 11, displaying a confirmation screen showing the detection results and reproducing the corresponding auscultatory sounds. Therefore, by reviewing the confirmation screen and auscultatory sounds, medical personnel can monitor long-term lung sound detection results and trace back to an earlier point before their initial examination of the patient to confirm whether additional sounds were produced. Additionally, medical personnel can use a communication device to hear the auscultatory sounds corresponding to the sounds detected by sensor 11. That is, even when medical personnel are far from the patient, they can monitor the patient's condition. Therefore, the information processing system 100D can be used for remote diagnosis and treatment.

[0231] Furthermore, as mentioned above, the 3D communication device can receive input that displays information from auscultation results received by medical personnel. Medical personnel can confirm the auscultation findings and set a threshold value that should be displayed as an abnormal range.

[0232] [Variation Example 8]

[0233] In the above embodiment, the sound conversion unit 210 is provided in the server device 2D, but it is not limited thereto. For example, the sound conversion unit 210 may also be provided in the edge server device 4. In this case, the edge server device 4 sends the feature signal and lung sound signal extracted based on the detection signal, as well as information displaying the auscultatory equivalent sound obtained by converting the detection signal to the communication device 3D.

[0234] [Implementation Method 4]

[0235] Other embodiments of the present invention will be described below. Figure 15 This is a functional block diagram illustrating an example of the configuration of the information processing system 100E according to Embodiment 4. For example... Figure 15 As shown, the information processing system 100E includes a detection device 1E, a server device 2E, an edge server device 4E, and a communication device 3. In Embodiment 4, the information processing system 100E detects lung sounds from multiple sites of a subject and makes predictions related to the subject's respiratory diseases based on the detection signals obtained from these multiple sites.

[0236] The information processing system 100E according to Embodiment 1 of the present invention includes: a signal acquisition unit 201 that acquires a lung sound signal, which displays the lung sounds of the subject, extracted from a detection signal output by a sensor 11E that detects vibrations emitted by the subject; a third prediction unit 211 that outputs third prediction information based on the acquired lung sound signal, the third prediction information including at least one of the following: fifth information related to the likelihood that the subject has developed a respiratory disease, and sixth information related to the respiratory disease the subject has; and an output unit 206 that outputs at least one of the third prediction information and consultation recommendation information generated based on the third prediction information, which displays a recommendation to visit a medical institution. Furthermore, the sensor 11E has multiple detection areas, and the third prediction unit 211 outputs the third prediction information using the lung sound signal extracted from the detection signal based on vibrations generated at multiple locations on the subject.

[0237] <Detection Device 1E>

[0238] like Figure 15 As shown, the detection device 1E differs from the detection device 1 in that it has a sensor 11E instead of a sensor 11.

[0239] (Sensor 11E)

[0240] The sensor 11E has multiple detection areas D and outputs detection signals detected in each of the multiple detection areas D. For example, the sensor 11E can be placed between the mattress and the sheet on the bed where the subject is lying. With this configuration, detection can be performed when the detection area D of the sensor 11E is close to the subject.

[0241] Figure 16 This is a diagram illustrating an example of the configuration of sensor 11E in this embodiment. (As shown...) Figure 16 As shown, the sensor 11E is configured such that when the subject is lying on a bed equipped with the detection device 1E, the subject's lungs are located on the detection area D of the sensor 11E.

[0242] As an example, sensor 11E can have four detection areas D4 to D7. Detection areas D4 to D7 can be arranged side-by-side on the same plane. For example... Figure 16 As shown, when the subject lies supine on the sensor 11E, the detection areas D4-D7 are configured such that they lie between the lower end of the costal arch and the iliac crest, and within the transverse diameter of the subject's trunk. Furthermore, the detection areas D4-D7 are configured such that when the subject lies supine on the sensor 11E, multiple parts of the subject are each located within detection areas D4-D7. Specifically, when the subject lies supine, the upper lung field of the subject's right lung is located in detection area D4, the lower lung field of the subject's right lung is located in detection area D5, the upper lung field of the subject's left lung is located in detection area D6, and the lower lung field of the subject's left lung is located in detection area D7.

[0243] The detection signals detected in each of the detection areas D4 to D7 are output from the detection areas D4 to D7 respectively. The signal acquisition unit 101 of the detection device 1E establishes a correspondence between the detection signals output from each of the detection areas D4 to D7 and the information of the detection area D in which the detection signal was detected, and sends it to an external device, such as the edge server device 4E.

[0244] <Edge Server Device 4E>

[0245] like Figure 15 As shown, the edge server device 4E includes a control unit 40E instead of a control unit 40. The control unit 40E includes a signal acquisition unit 401, a signal extraction unit 402E, and an output unit 403E. The signal extraction unit 402E desorbs the detection signals detected in multiple detection areas D4 to D7 and extracts lung sound signals.

[0246] The output unit 403E displays information about each extracted lung sound signal, information about the body part of the subject corresponding to the detection area D of the lung sound signal extraction source (i.e., the detection signal detection area), and information about the time when the detection signal was detected. This information is then output to an external device, such as the server device 2E. Hereinafter, the lung sound signal extracted from a detection signal detected in a certain detection area D may also be referred to as the lung sound signal obtained from the body part of the subject corresponding to that detection area D. For example, the lung sound signal extracted from a detection signal detected in detection area D4 may also be referred to as the lung sound signal obtained from the upper lung field of the subject's right lung.

[0247] <Server Device 2E>

[0248] Server device 2E differs from server device 2 in that it has a control unit 20E instead of control unit 20. Control unit 20E also differs from control unit 20 in that it has a third prediction unit 211 instead of the first prediction unit 202.

[0249] The third prediction unit 211 generates third prediction information, including at least one of the fifth and sixth information, based on lung sound signals obtained from multiple sites of the subject, and outputs it to the output unit 206. The fifth information is information related to the likelihood that the subject will develop a respiratory disease. The sixth information is information related to the respiratory disease the subject has.

[0250] The fifth piece of information may include the confidence level of the subject's respiratory illness. Additionally, the third piece of information may include the probability that a subject with a respiratory illness will develop the illness within a given period P1 from the measurement time of the detection signal. The given period P1 can be any period, such as 10 days or 1 day. The sixth piece of information may include information about the severity of the subject's respiratory illness as shown in the detection signal, or the extent to which it will worsen in the future.

[0251] The third prediction unit 211 generates third prediction information related to the subject by inputting lung sound signals obtained from multiple sites of the subject into at least one of the prediction models M2E shown below (1), (2), and (3) obtained by machine learning using training data. (1) A prediction model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the subject as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. (2) A prediction model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the subject as explanatory variables and information related to the respiratory disease suffered by the subject as a response variable. (3) A prediction model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the subject as explanatory variables and information related to whether the subject has developed a disease within a given period P1 from the measurement time of the detection signal as a response variable. As an example, the doctor's examination results of the subject can be used as the response variable of the training data.

[0252] It should be noted that, in the following text, the "predictive model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the sample as explanatory variables and information related to whether the sample has developed a respiratory disease as a response variable" in (1) above is sometimes referred to as predictive model M21E. The "predictive model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the sample as explanatory variables and information related to the respiratory disease suffered by the sample as a response variable" in (2) above is sometimes referred to as predictive model M22E. The "predictive model M2E obtained by machine learning using training data with lung sound signals obtained from multiple sites of the sample as explanatory variables and information related to whether the sample has developed a disease within a given period P1 from the measurement time of the detection signal as a response variable" in (3) above is sometimes referred to as predictive model M23E. For example, to generate the fifth piece of information, at least one of predictive model M21E and predictive model M23E can be used. On the other hand, to generate the sixth piece of information, the predictive model M22E can be used. It should be noted that the information related to the respiratory disease suffered by the samplers, which serves as the response variable of the predictive model M22E, may include information on the severity of the respiratory disease suffered by the samplers.

[0253] The third prediction unit 211 can further utilize the feature signals to output third prediction information. For example, the third prediction unit 211 can generate the third prediction information using a prediction model M2E obtained through machine learning, which uses the sample's feature signals as explanatory variables, information related to whether the sample has developed a respiratory disease, and information related to the sample's respiratory disease as response variables, in addition to lung sound signals obtained from multiple sites of the sample. In this case, by inputting the feature signals obtained from the subject and the lung sound signals obtained from multiple sites of the subject into the prediction model M2E, the third prediction unit 211 generates third prediction information related to the subject.

[0254] The third prediction unit 211 can generate the third prediction information using lung sound signals obtained from multiple parts of the subject at one time point, or it can generate the third prediction information using lung sound signals obtained from multiple parts of the subject at multiple time points with different time series.

[0255] The third prediction unit 211 can use other information related to the subject's respiratory system disease, along with lung sound signals, as explanatory variables. For example, characteristic signals related to respiratory rate, heart rate, and body movement extracted from the test signals obtained from the subject can be used as explanatory variables. In addition, the subject's body temperature, auscultation data, vital signs such as SpO2, information related to the subject's past medical history, and information related to the subject's lifestyle habits such as sleep duration can also be used as explanatory variables.

[0256] The following examples illustrate the correspondence between respiratory diseases and the characteristics of lung sounds emitted by subjects suffering from those diseases.

[0257] (Difference between left and right lung sounds)

[0258] In patients with respiratory diseases, a sound pressure difference can sometimes occur between the left and right lungs. Specifically, in cases of respiratory diseases, the difference in airflow between the affected and healthy lungs can create a sound pressure difference between the left and right lungs.

[0259] The third prediction unit 211 uses lung sound signals based on lung sounds obtained from the left and right lungs of the subject as explanatory variables. Therefore, the third prediction unit 211 is able to predict whether there is a sound pressure difference between the left and right lung sounds and whether the subject has a respiratory disease.

[0260] (Bronchoalveolar phonation at the site of lung vesicle detection)

[0261] In patients with respiratory diseases, lung sounds at the site where vesicles are detected may sometimes exhibit bronchial breath sounds. In cases of respiratory diseases, lung swelling leads to increased water content in the lung tissue (hydration). This improves sound conduction within the lungs, making it easier for high-frequency sounds contained in lung sounds to reach the detection area D of sensor 11E. Furthermore, bronchial breath sounds primarily occur in the lower lung fields.

[0262] In cases of bronchial breath sounding, lung sound signals obtained from the lower lung field will contain more high-frequency signals compared to lung sound signals obtained from the same location without bronchial breath sounding.

[0263] The third prediction unit 211 uses lung sound-based signals obtained from the lower lung field of the subject as explanatory variables. Therefore, the third prediction unit 211 is able to predict whether bronchial breath sounds have occurred in the lower lung field of the subject, and whether the subject has a respiratory disease.

[0264] In addition, when bronchial breath sounds are detected at the lung sound detection site, the lung sound-based signal obtained from the lower lung field will contain more high-frequency signals compared to lung sound signals obtained from other sites.

[0265] Therefore, by using lung sound-based signals acquired from the subject's upper and lower lung fields respectively, the third prediction unit 211 can determine that the lung sounds in the lower lung field contain signals with more high-frequency regions compared to the lung sounds in the upper lung field. In this way, the third prediction unit 211 uses the lung sound-based signals acquired from the subject's upper and lower lung fields respectively as explanatory variables. Thus, the third prediction unit 211 can more accurately predict the likelihood that a subject has bronchial breath sounds in the lower lung field and therefore has a respiratory disease.

[0266] (Bronchoalveolar phonation at the site of lung vesicle detection)

[0267] Furthermore, lung sounds originating from expiration are difficult to observe in patients without respiratory illnesses. Conversely, in patients with respiratory illnesses, sound conduction within the lungs becomes better, making lung sounds originating from expiration easier to observe. Lung sounds originating from expiration can be observed throughout the chest wall.

[0268] The third prediction unit 211 uses lung sound signals based on lung sounds obtained from multiple sites of the subject as explanatory variables. Therefore, the third prediction unit 211 can predict whether the subject's lung sounds contain clearer exhaled lung sounds compared to when the subject does not have a respiratory disease, and whether the subject has a respiratory disease.

[0269] (Temporal changes in lung sounds associated with aspiration pneumonia)

[0270] In cases of aspiration followed by pneumonia, the patient's lung sounds may include sounds originating from the aspirated foreign object, as well as sounds from inflammation. Furthermore, the location of these sounds may change over time. For example, if the aspirated foreign object is in the bronchus, high-frequency sounds will be produced in the bronchus. Conversely, if the aspirated foreign object has reached the lower lobe of the lung, additional sounds will be produced due to fluid secreted in the alveoli. In cases of aspiration followed by pneumonia, additional sounds and high-frequency sounds will be produced at sites of inflammation.

[0271] The third prediction unit 211 uses lung sound signals acquired from multiple parts of the subject at multiple time points. Therefore, the third prediction unit 211 can determine changes in the location and type of sounds associated with the subject's respiratory disease, and predict the likelihood that the subject has a respiratory disease and the degree of progression of that disease.

[0272] For example, if the lung sound signal acquired from the detection area D near the subject's bronchus contains high-frequency sounds, the third prediction unit 211 can predict that the subject has experienced aspiration. Subsequently, if the lung sound signal acquired from the detection area D near the lower lobe of the lung contains additional sounds, the third prediction unit 211 can predict that a foreign object has reached the lung and the risk of aspiration pneumonia is increasing. Furthermore, if the lung sound signal acquired from any detection area D begins to contain additional or high-frequency sounds, the third prediction unit 211 can predict that inflammation has occurred at the location where the additional or high-frequency sounds were detected, and that the subject has developed aspiration pneumonia.

[0273] Previously, when doctors examined patients with respiratory diseases, they auscultated specific locations on the lungs sequentially. In respiratory diseases, specific lung sounds are sometimes heard corresponding to the symptoms and depending on the location. However, conventional methods could not simultaneously hear and compare lung sounds from multiple locations, making it impossible to establish a correspondence between lung sounds from different locations and respiratory diseases. Furthermore, the areas and types of lung sounds audible from respiratory diseases sometimes change over time, while conventional methods could only detect lung sounds at the time of auscultation, making it difficult to grasp these historical changes in lung sounds.

[0274] Here, in the information processing system 100E of this embodiment, the sensor 11E is capable of simultaneously and sequentially acquiring detection signals from multiple sites. Furthermore, the third prediction unit 211 is capable of generating third prediction information, including at least one of fifth and sixth information, based on the lung sound signals acquired from multiple sites of the subject. In this way, by using the lung sound signals acquired from multiple sites of the subject, the information processing system 100E can more accurately and effectively predict information related to the subject's respiratory system diseases.

[0275] [Variation Example 9]

[0276] In the above embodiments, the sensor 11E may further have multiple detection areas D. For example, Figure 16 The detection areas D4-D7 shown can each be divided into two or more detection areas, allowing for the acquisition of more detection signals. By acquiring more detection signals within a given range, the location of complex sounds contained in lung sounds can be more accurately determined. This improves the predictive accuracy related to respiratory diseases in the subject.

[0277] [Variation Example 10]

[0278] Furthermore, in the above-described embodiments, it is possible to... Figure 16 The detection areas D4-D7 shown are further set at different locations. For example, a detection area can be set below the subject's larynx. The prediction model M2E can perform machine learning using training data with lung sound signals including additional sounds that can be obtained from the subject's larynx as explanatory variables. In this case, the third prediction unit 211 can use the lung sound signals obtained from the subject's larynx as explanatory variables to make predictions related to the subject's respiratory diseases.

[0279] [Variation Example 11]

[0280] Furthermore, in the above-described embodiment, the information processing system 100E can be a system capable of responding to actions such as turning over by the subject. For example, multiple detection areas D can be provided throughout the bed where the subject is lying. Additionally, the third prediction unit 211 can determine a feature signal with a high signal intensity corresponding to a sound originating from heart sounds among the acquired feature signals, and determine the detection area surrounding the detection area D where the extraction source of that feature signal, i.e., lung sounds, is acquired, as the position where the subject is lying. The third prediction unit 211 can be used to predict lung sound signals originating from the sound acquired from the detection area D at the subject's lying position. Furthermore, the information processing system 100E can include a camera (not shown) capable of photographing the bed where the subject is lying, and determine the subject's position based on the image captured by the camera. The method for determining the subject's lying position is not limited to this; any method can be used.

[0281] [Software-based implementation example]

[0282] The functions of the information processing systems 100, 100A~100E (hereinafter referred to as "the System") can be realized by a program that enables the computer to function as the System, and the program enables the computer to function as the various control modules of the System (especially the various units contained in the various control units).

[0283] In this case, the system described above, as hardware for executing the program, includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory). By executing the program using this control device and storage device, the functions described in the above embodiments can be realized.

[0284] The aforementioned program can be recorded in a non-transitory manner on one or more computer-readable recording media. This recording media may or may not be present in the aforementioned device. In the latter case, the program can be provided to the aforementioned device via any wired or wireless transmission medium.

[0285] Furthermore, some or all of the functions of the aforementioned control modules can also be implemented using logic circuits. For example, integrated circuits that form the logic circuits that enable the functions of the aforementioned control modules are also included within the scope of this invention. In addition, the functions of the aforementioned control modules can also be implemented using, for example, a quantum computer.

[0286] Furthermore, the processes described in the above embodiments can also be performed by AI (Artificial Intelligence). In this case, the AI ​​can run in the control device described above, or it can run in other devices (such as edge computers or cloud servers).

[0287] [Summarize]

[0288] The information processing system (100 / 100A / 100B / 100D) according to Embodiment 1 of the present invention includes: a signal acquisition unit (201) that acquires a feature signal extracted from a detection signal output by a sensor (11) of vibrations emitted by a subject, and a lung sound signal displaying lung sounds of the subject, wherein the feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; a first prediction The first prediction unit (202) outputs first prediction information based on the acquired feature signal and lung sound signal. The first prediction information includes at least one of first information related to the possibility that the subject has a respiratory disease and second information related to the respiratory disease the subject has. The output unit (206) outputs at least one of the first prediction information and consultation recommendation information generated based on the first prediction information, which displays a recommendation to visit a medical institution.

[0289] Based on the above configuration, the information processing system extracts characteristic signals closely related to the subject's physical condition and lung sound signals closely related to respiratory diseases from the detection signals output by a sensor emitting vibrations emitted by the subject, thereby predicting the likelihood of the subject developing a respiratory disease. Therefore, the information processing system can predict the likelihood of the subject developing a respiratory disease and the specific respiratory disease the subject may have before symptoms appear, and can output this prediction or a medical consultation recommendation based on the prediction. As a result, an information processing system capable of early detection of respiratory diseases and early recommendations for medical consultation can be achieved.

[0290] In the information processing system (100 / 100A / 100B / 100D) of the second embodiment of the present invention, in the first embodiment 1, the first prediction unit (202) can further use information related to the sleep state of the subject to output the first prediction information.

[0291] For example, the detection signal can sometimes be affected by whether the subject is asleep or not. Based on the above configuration, since the first prediction unit can take into account the sleep state when outputting the first prediction information, the prediction accuracy of the likelihood of the subject developing respiratory diseases can be improved.

[0292] In the information processing system (100 / 100A / 100B / 100D) of the present invention in the third embodiment, in the first embodiment or the second embodiment, the first prediction information related to the subject can be output by inputting the feature signal and the lung sound signal obtained from the subject into at least one of the prediction models (1) to (3) below: (1) a prediction model (M2) obtained by machine learning using training data with the feature signal and the lung sound signal of the subject as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. 1); (2) A predictive model (M22) obtained by machine learning using training data with the above-mentioned characteristic signals and lung sound signals of the above-mentioned sample as explanatory variables and information related to the respiratory diseases suffered by the above-mentioned sample as response variables; and (3) A predictive model (M23) obtained by machine learning using training data with the above-mentioned characteristic signals and lung sound signals of the above-mentioned sample suffering from respiratory diseases as explanatory variables and information related to whether the above-mentioned sample has developed the disease within a given period (P1) from the measurement time of the above-mentioned detection signal as response variables.

[0293] Based on the above configuration, the first prediction unit can output first prediction information based on the prediction model obtained by machine learning using training data, and thus can predict the likelihood of a subject developing respiratory diseases with higher accuracy.

[0294] In the information processing system (100 / 100A / 100B / 100D) of Embodiment 4 of the present invention, in Embodiment 3 above, it may further include: a diagnostic result acquisition unit that acquires a diagnostic result including at least one of the following: information related to whether the subject has developed a respiratory disease, information related to the respiratory disease the subject has, and information related to whether the subject has developed a disease within the given period (P1) from the measurement time of the detection signal; and a model update unit (204) that updates the prediction model (M2) by using machine learning of training data, wherein the training data uses the feature signal and the lung sound signal used for the output of the first prediction information as explanatory variables and the diagnostic result as a response variable.

[0295] Based on the above structure, the prediction model is updated by using machine learning with training data that uses diagnostic results related to respiratory diseases of actual subjects as explanatory variables. Therefore, the accuracy of predictions such as the probability of subjects developing respiratory diseases can be improved each time the information processing system is used.

[0296] In the information processing system (100 / 100A / 100B) of Embodiment 5 of the present invention, in any of Embodiments 1 to 4, the first prediction information may further include first additional information related to the probability that the subject who does not suffer from a respiratory disease will develop a respiratory disease within a given period (P2) from the measurement time of the detection signal. The first prediction unit (202) outputs the first additional information related to the subject by inputting the feature signal and the lung sound signal obtained from the subject into the prediction model (M2). The prediction model is obtained by machine learning using training data with the feature signal and the lung sound signal of the subject who does not suffer from a respiratory disease as explanatory variables and information related to whether the subject has developed a respiratory disease within the given period (P2) from the measurement time of the detection signal as a response variable. According to the above configuration, the probability of future illness can be predicted.

[0297] The information processing system (100C) of Embodiment 6 of the present invention includes: a signal acquisition unit (201) that acquires a feature signal extracted from a detection signal output by a sensor (11) of vibrations emitted by a subject, and a lung sound signal displaying the subject's lung sounds, the feature signal including at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; and a sleep state determination unit (208) that determines the sleep state based on the acquired feature signal. The system outputs a determination result related to the sleep state of the subject; a second prediction unit (209) outputs second prediction information based on the determination result and the lung sound signal, the second prediction information including at least one of a third information related to the possibility that the subject has a respiratory disease and a fourth information related to the respiratory disease the subject has; and an output unit (206) outputs at least one of the second prediction information and a consultation recommendation information generated based on the second prediction information, which displays a recommendation to visit a medical institution.

[0298] Based on the above configuration, the information processing system extracts lung sound signals closely related to respiratory diseases from the detection signals output by a sensor emitting vibrations emitted by the subject, and sleep states closely related to physical condition based on characteristic signals extracted from the aforementioned detection signals, thereby predicting the likelihood of the subject developing a respiratory disease. Thus, the information processing system can predict the likelihood of the subject developing a respiratory disease before symptoms appear, and if the subject does develop a respiratory disease, it can output the prediction or a medical recommendation based on the prediction. As a result, an information processing system capable of early detection of respiratory diseases and early recommendations for medical visits can be achieved.

[0299] In the information processing system (100C) of the present invention in the seventh embodiment, in the above embodiment 6, the second prediction unit (209) can further use the above feature signal to output the second prediction information.

[0300] The characteristic signals are closely related to physical condition. Based on the above configuration, since the second prediction unit can take into account the characteristic signals and output second prediction information, the prediction accuracy of the likelihood of the subject developing respiratory diseases, etc., can be improved.

[0301] In the information processing system (100C) of the present invention according to embodiment 8, in embodiment 6 or 7, the second prediction unit (209) can output second prediction information related to the subject by inputting at least one of the determination results based on the feature signals obtained from the subject and the lung sound signal into at least one of the prediction models in (1) to (3) below: (1) using a prediction model (M2) obtained by machine learning with the determination results based on the feature signals of the subject and the lung sound signal as explanatory variables and information related to whether the subject has developed a respiratory disease as a response variable. (1C); (2) A prediction model (M22C) obtained by machine learning using training data with the above-mentioned judgment result based on the above-mentioned characteristic signal of the above-mentioned sample and the above-mentioned lung sound signal as explanatory variables and information related to the respiratory disease suffered by the above-mentioned sample as response variable; and (3) A prediction model (M23C) obtained by machine learning using training data with the above-mentioned judgment result based on the above-mentioned characteristic signal of the above-mentioned sample suffering from respiratory disease and the above-mentioned lung sound signal as explanatory variables and information related to whether the above-mentioned sample has developed the disease within a given period (P1) from the measurement time of the above-mentioned detection signal as response variable.

[0302] Based on the above configuration, the second prediction unit can output second prediction information based on the prediction model obtained by machine learning using training data, and thus can predict the likelihood of a subject developing respiratory diseases with higher accuracy.

[0303] In the information processing system (100C) of embodiment 9 of the present invention, in embodiment 8, it may further include: a diagnostic result acquisition unit that acquires a diagnostic result including at least one of the following: information related to whether the subject has developed a respiratory disease, information related to the respiratory disease the subject has, and information related to whether the subject has developed a disease within the given period (P1) from the measurement time of the detection signal; and a model update unit (204) that updates the prediction model (M2C) by using machine learning of training data, wherein the training data uses the determination result based on the feature signal and the lung sound signal for the output of the second prediction information as explanatory variables and the diagnostic result as a response variable.

[0304] Based on the above structure, the prediction model is updated by using machine learning with training data that uses diagnostic results related to respiratory diseases of actual subjects as explanatory variables. Therefore, the accuracy of predictions such as the probability of subjects developing respiratory diseases can be improved each time the information processing system is used.

[0305] In the information processing system (100C) of Embodiment 10 of the present invention, in any of Embodiments 6 to 9, the second prediction information may further include second additional information related to the likelihood that the subject who does not suffer from a respiratory disease will develop a respiratory disease within a given period (P2) from the measurement time of the detection signal. The second prediction unit (209) outputs the second additional information related to the subject by inputting the determination result based on the feature signal and the lung sound signal obtained from the subject into the prediction model. The prediction model is obtained by machine learning using training data with the determination result based on the feature signal and the lung sound signal of the subject who does not suffer from a respiratory disease as explanatory variables and information related to whether the subject has developed a respiratory disease within the given period (P2) from the measurement time of the detection signal as the response variable. According to the above configuration, the likelihood of future illness can be predicted.

[0306] In the information processing system (100C) of the present invention according to embodiment 11, in any of the embodiments 6 to 10, the second prediction unit (209) may further use the attribute information of the object to output the second prediction information.

[0307] The attributes of the subject can sometimes affect the detection signal. Based on the above configuration, since the second prediction unit can take into account the attribute information of the subject and output the second prediction information, the prediction accuracy of the subject's likelihood of developing respiratory diseases, etc., can be improved.

[0308] In the information processing system (100C) of the present invention according to embodiment 12, in any of the embodiments 6 to 11, the second prediction unit (209) may further use medical history information related to the respiratory diseases suffered by the subject in the past to output the second prediction information.

[0309] If a subject has a history of respiratory illness, that subject is more likely to develop pneumonia. Based on the above configuration, since the second prediction unit can output second prediction information taking into account medical history information related to past respiratory illnesses, the prediction accuracy of the subject's likelihood of developing respiratory illnesses can be improved.

[0310] In the information processing system (100C) of embodiment 13 of the present invention, in any of the embodiments 6 to 12, the second prediction unit (209) may further use nursing history information to output the second prediction information, which shows the execution history of nursing care provided by the subject to reduce the likelihood of the occurrence of the respiratory disease.

[0311] Without proper care to suppress aspiration, individuals are more susceptible to developing pneumonia. Based on the above structure, by generating second predictive information that takes into account nursing history, the accuracy of predicting the likelihood of developing respiratory diseases in individuals can be improved.

[0312] In the information processing system (100 / 100A / 100B / 100C / 100D) of Embodiment 14 of the present invention, in any of Embodiments 1 to 13 above, the respiratory disease can be pneumonia.

[0313] Based on the above structure, it is possible to detect respiratory diseases in individuals with pneumonia at an early stage and recommend them to medical institutions for treatment.

[0314] In the information processing system (100 / 100A / 100B / 100C) of Embodiment 15 of the present invention, in any of the embodiments 1 to 14, a pneumonia determination unit (205) may be further provided to determine the pneumonia status of the subject based on the lung sound signal, and the output unit (206) outputs the determination result based on the pneumonia determination unit.

[0315] Based on the above structure, the individual can also determine whether they have actually contracted pneumonia.

[0316] In the information processing system (100 / 100A / 100B / 100C / 100D) of Embodiment 16 of the present invention, in any of the embodiments 1 to 15 above, the medical visit recommendation information may include information displaying at least one of the urgency and reliability of the medical visit recommendation.

[0317] Based on the above structure, the subject can make a judgment on whether to actually go to a medical institution for treatment, taking into account both urgency and reliability.

[0318] In the information processing system (100 / 100A / 100B / 100C / 100D) of Embodiment 17 of the present invention, in any of the embodiments 1 to 16 above, the sensor can be in the form of a thin plate.

[0319] Based on the above configuration, it is possible to detect vibrations emitted by the user without causing discomfort to the user of the bed.

[0320] In the information processing system (100 / 100A / 100B / 100C / 100D) of the present invention in any of the above-mentioned methods 1 to 17, the sensor (11) may have multiple detection areas (D) that output the above-mentioned detection signal, and the signal acquisition unit (201) acquires the above-mentioned detection signal in each of the multiple detection areas.

[0321] Based on the above configuration, the detection signal can be accurately determined based on the signal strength of the detection signal detected in each detection area.

[0322] In the information processing system (100 / 100A / 100B / 100C / 100D) of the present invention, in any of the above-mentioned methods 1 to 18, the sensor (11) can detect the vibration emitted by the object at a position where it does not come into contact with the object.

[0323] Based on the above configuration, it is possible to detect vibrations emitted by the subject without causing discomfort to the subject.

[0324] The information processing method of Embodiment 20 of the present invention includes: a signal acquisition step, acquiring a feature signal extracted from a detection signal output by a sensor (11) of vibrations emitted by a subject, and a lung sound signal displaying the lung sounds of the subject, wherein the feature signal includes at least one of a heartbeat signal displaying the heartbeat of the subject, a respiratory vibration signal displaying the respiratory vibrations of the subject, and a body movement signal displaying the body movements of the subject; a first prediction step, outputting first prediction information based on the acquired feature signal and the lung sound signal, wherein the first prediction information includes at least one of first information related to the possibility that the subject has developed a respiratory disease, and second information related to the respiratory disease suffered by the subject; and an output step, outputting at least one of the first prediction information and consultation recommendation information generated based on the first prediction information, displaying a recommendation to visit a medical institution. According to the above configuration, the same effect as Embodiment 1 is achieved.

[0325] The program of Embodiment 21 of the present invention is a program for enabling a computer to function as the information processing system described in Embodiment 1, wherein the program enables the computer to function as the signal acquisition unit (201), the first prediction unit (202), and the output unit (206). According to the program, the same effects as in Embodiment 1 are achieved.

[0326] The information processing method of embodiment 22 of the present invention includes: a signal acquisition step, acquiring a feature signal extracted from a detection signal output by a sensor (11) emitting vibrations emitted by a subject, and a lung sound signal displaying the subject's lung sounds, wherein the feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibrations, and a body movement signal displaying the subject's body movements; a sleep state determination step, outputting a determination result related to the subject's sleep state based on the acquired feature signal; a second prediction step, outputting second prediction information based on the determination result and the lung sound signal, wherein the second prediction information includes at least one of third information related to the likelihood of the subject developing a respiratory disease, and fourth information related to the respiratory disease the subject has; and an output step, outputting at least one of the second prediction information and consultation recommendation information generated based on the second prediction information, displaying a recommendation to visit a medical institution. According to the above configuration, it achieves the same effect as embodiment 6.

[0327] The program of embodiment 23 of the present invention is a program for enabling the computer to function as the information processing system described in embodiment 6, wherein the program enables the computer to function as the signal acquisition unit (201), the sleep state determination unit (208), the second prediction unit (209), and the output unit (206). According to the program described above, the same effects as in embodiment 5 are achieved.

[0328] The information processing systems of the various embodiments of the present invention can also be implemented by a computer. In this case, the control program of the information processing system implemented by the computer as the various parts (software elements) of the information processing system, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

[0329] This invention is not limited to the embodiments described above, and various modifications can be made within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included within the technical scope of this invention. Furthermore, new technical features can be formed by combining the technical means disclosed in each embodiment.

Claims

1. An information processing system, comprising: The signal acquisition unit acquires a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds. The feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibration, and a body movement signal displaying the subject's body movement. A first prediction unit outputs first prediction information based on the acquired feature signal and the lung sound signal. This first prediction information includes at least one of the following: first information related to the likelihood of the subject developing a respiratory disease, and second information related to the respiratory disease the subject is suffering from; and The output unit outputs at least one of the first prediction information and the medical visit recommendation information generated based on the first prediction information, which displays a recommendation to a medical institution.

2. The information processing system according to claim 1, wherein, The first prediction unit further uses information related to the sleep state of the subject to output the first prediction information.

3. The information processing system according to claim 1, wherein, The first prediction unit outputs the first prediction information related to the subject by inputting the feature signal and the lung sound signal obtained from the subject into at least one of the prediction models (1) to (3) below: (1) A prediction model obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample as explanatory variables and information related to whether the sample has developed a respiratory disease as a response variable; (2) A predictive model obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample as explanatory variables and information related to the respiratory diseases suffered by the sample as response variables; and (3) A predictive model obtained by machine learning using training data with the characteristic signals and lung sound signals of the sample suffering from respiratory diseases as explanatory variables and information related to whether the sample has developed the disease within a given period from the measurement time of the detection signal as the response variable.

4. The information processing system according to claim 3, further comprising: A diagnostic result acquisition unit acquires a diagnostic result including at least one of the following: information related to whether the subject has developed a respiratory disease, information related to the respiratory disease the subject has, and information related to whether the subject has developed the disease within the given period from the measurement time of the detection signal; and The model update unit updates the prediction model using machine learning with the training data, wherein the training data uses the feature signal and the lung sound signal used as the output of the first prediction information as explanatory variables and the diagnosis result as the response variable.

5. The information processing system according to claim 1, wherein, The first predictive information further includes: first additional information relating to the likelihood that the subject, who does not have a respiratory disease, will develop a respiratory disease within a given period from the measurement time of the detection signal. The first prediction unit outputs the first additional information related to the subject by inputting the feature signal and the lung sound signal obtained from the subject into the prediction model. The prediction model is obtained by machine learning using training data with the feature signal and the lung sound signal of the sample without respiratory disease as explanatory variables and information related to whether the sample has a respiratory disease during the given period from the measurement time of the detection signal as the response variable.

6. An information processing system, comprising: The signal acquisition unit acquires a feature signal extracted from a detection signal output by a sensor of vibration emitted by a subject, and a lung sound signal displaying the subject's lung sounds. The feature signal includes at least one of a heartbeat signal displaying the subject's heartbeat, a respiratory vibration signal displaying the subject's respiratory vibration, and a body movement signal displaying the subject's body movement. A sleep state determination unit outputs a determination result related to the sleep state of the subject based on the acquired feature signals; The second prediction unit outputs second prediction information based on the determination result and the lung sound signal. The second prediction information includes at least one of the following: third information related to the possibility that the subject has a respiratory disease, and fourth information related to the respiratory disease the subject has. as well as The output unit outputs at least one of the second prediction information and the consultation recommendation information generated based on the second prediction information, which displays a consultation recommendation to a medical institution.

7. The information processing system according to claim 6, wherein, The second prediction unit further uses the feature signal to output the second prediction information.

8. The information processing system according to claim 6, wherein, The second prediction unit outputs second prediction information related to the subject by inputting at least one of the determination results based on the feature signals obtained from the subject and the lung sound signal into at least one of the prediction models (1) to (3) below. (1) A prediction model obtained by machine learning using training data with the determination result based on the characteristic signal of the sample and the lung sound signal as explanatory variables and information related to whether the sample has developed a respiratory disease as a response variable. (2) A predictive model obtained by machine learning using training data with the determination result based on the characteristic signals of the sample and the lung sound signal as explanatory variables, and information related to the respiratory disease suffered by the sample as a response variable; and (3) A predictive model obtained by machine learning using training data with the determination result based on the characteristic signal of the sample person suffering from respiratory disease and the lung sound signal as explanatory variables, and information related to whether the sample person has developed the disease within a given period from the measurement time of the detection signal as the response variable.

9. The information processing system according to claim 8, further comprising: A diagnostic result acquisition unit acquires a diagnostic result including at least one of the following: information related to whether the subject has developed a respiratory disease, information related to the respiratory disease the subject has, and information related to whether the subject has developed the disease within the given period from the measurement time of the detection signal; and The model update unit updates the prediction model by using machine learning on the training data, wherein the training data uses the determination result based on the feature signal and the lung sound signal as explanatory variables and the diagnosis result as response variables as explanatory variables for the output of the second prediction information.

10. The information processing system according to claim 6, wherein, The second predictive information further includes: second supplementary information relating to the likelihood that the subject, who does not have a respiratory disease, will develop a respiratory disease within a given period from the measurement time of the detection signal. The second prediction unit outputs the second additional information related to the subject by inputting the determination result based on the feature signal and the lung sound signal obtained from the subject into the prediction model. The prediction model is obtained by machine learning using training data with the determination result based on the feature signal and the lung sound signal of the sample without respiratory disease as explanatory variables and information related to whether the sample has a respiratory disease during the given period from the measurement time of the detection signal as the response variable.

11. The information processing system according to claim 6, wherein, The second prediction unit further uses the attribute information of the object to output the second prediction information.

12. The information processing system according to claim 6, wherein, The second prediction unit further uses medical history information related to the subject's past respiratory diseases to output the second prediction information.

13. The information processing system according to claim 6, wherein, The second prediction unit further uses nursing history information to output the second prediction information, which shows the execution history of nursing care provided by the subject to reduce the likelihood of developing the respiratory disease.

14. The information processing system according to claim 1 or 6, wherein, The respiratory disease mentioned is pneumonia.

15. The information processing system according to claim 14, further comprising: A pneumonia determination unit that determines the pneumonia status of the subject based on the lung sound signals. The output unit outputs the determination result based on the pneumonia determination unit.

16. The information processing system according to claim 1 or 6, wherein, The medical appointment recommendation information includes information showing at least one of the urgency and reliability of the recommendation.

17. The information processing system according to claim 1 or 6, wherein, The sensor is in the shape of a thin plate.

18. The information processing system according to claim 1 or 6, wherein, The sensor has multiple detection areas for outputting the detection signal. The signal acquisition unit acquires the detection signal in each of the plurality of detection areas.

19. The information processing system according to claim 1 or 6, wherein, The sensor detects vibrations emitted by the object at a position where it does not come into contact with the object.

20. An information processing method, comprising: The signal acquisition step involves acquiring a feature signal extracted from a detection signal output by a sensor that detects vibrations emitted by the subject, and a lung sound signal that displays the subject's lung sounds. The feature signal includes at least one of a heartbeat signal that displays the subject's heartbeat, a respiratory vibration signal that displays the subject's respiratory vibrations, and a body movement signal that displays the subject's body movements. The first prediction step outputs first prediction information based on the acquired feature signal and lung sound signal. The first prediction information includes at least one of the following: first information related to the possibility that the subject has a respiratory disease, and second information related to the respiratory disease the subject has. as well as The output step outputs at least one of the first prediction information and the medical recommendation information generated based on the first prediction information, which displays a recommendation to visit a medical institution.

21. A program for enabling a computer to function as the information processing system of claim 1, wherein, The program is used to enable the computer to function as the signal acquisition unit, the first prediction unit, and the output unit.

22. An information processing method, comprising: The signal acquisition step involves acquiring a feature signal extracted from a detection signal output by a sensor that detects vibrations emitted by the subject, and a lung sound signal that displays the subject's lung sounds. The feature signal includes at least one of a heartbeat signal that displays the subject's heartbeat, a respiratory vibration signal that displays the subject's respiratory vibrations, and a body movement signal that displays the subject's body movements. The sleep state determination step outputs a determination result related to the sleep state of the subject based on the acquired feature signals; The second prediction step outputs second prediction information based on the determination result and the lung sound signal. The second prediction information includes at least one of the following: third information related to the possibility that the subject has a respiratory disease, and fourth information related to the respiratory disease the subject has. as well as The output step outputs at least one of the second prediction information and the medical recommendation information generated based on the second prediction information, which displays a recommendation to visit a medical institution.

23. A program for enabling a computer to function as the information processing system of claim 6, wherein, The program enables the computer to function as the signal acquisition unit, the sleep state determination unit, the second prediction unit, and the output unit.

Images