WO2021152710A1

WO2021152710A1 - Information transmission device and information transmission method

Info

Publication number: WO2021152710A1
Application number: PCT/JP2020/003048
Authority: WO
Inventors: 野中　修; 智子後町; 弘達藤原; 亮櫻井
Original assignee: オリンパス株式会社
Priority date: 2020-01-28
Filing date: 2020-01-28
Publication date: 2021-08-05
Also published as: CN114830255A

Abstract

Provided are an information transmission device and information transmission method whereby it is possible to take subject conditions into account to accurately ascertain state of health, and provide customized information, such as advice, that corresponds to the state of health.　The present invention comprises: an ID identification unit 1b that acquires a time-series first test data group for a subject via a first device 2a, and acquires a time-series second test data group for the subject via a second device 2b that is capable of testing that enables interpolation of the first test data group; and an information provision unit 1c that uses the first test data group and the second test data group to determine transmission information to be provided to the subject. The first test data group and the second test data group complement each other in terms of test timing or test items.

Description

Information transmission device and information transmission method

The present invention relates to an information transmission device and an information transmission method capable of providing a user with customized information such as advice according to an inspection result that can be obtained in daily life.

The Internet has become widespread in recent years, and by using the Internet, it is possible to easily obtain information closely related to the user's life. By using this acquired information, various customized information (valid information) suitable for each user is generated, and services that provide this customized information are increasing. For example, services that introduce health foods are often interesting information that is common to many people, so this type of service is often found.

In addition, since the network environment has been improved, various proposals have been made to perform remote inspections outside specialized institutions such as hospitals. For example, Patent Document 1 discloses a remote inspection method for transmitting inspection data by using a sensor chip and a mobile phone as a reader / writer and using a public communication network. Then, in Patent Document 1, it is proposed to store past inspection data and its evaluation results in a database and use these information.

Further, Patent Document 2 discloses a biometric information measuring device that integrates personal authentication data and image data photographed by excrement photographing means and transmits the integrated data by communication means. Further, in Patent Document 3, a study list related to an examination is displayed, a medical image based on the medical image information of the subject in the selected study is displayed on the image display screen, and when a history browser is requested, the study of the subject is requested. A method of displaying medical information that displays a history browser representing a list of medical information is disclosed.

JP-A-2009-258886 Japanese Unexamined Patent Publication No. 2014-031655 Japanese Patent No. 5294947

The above-mentioned Patent Document 1-3 describes that biometric information is acquired and this information is remotely transmitted and used. If the subject has no subjective symptoms, it is useful to know that he / she is ill and that he / she may become ill. The user may be inspected not only by a plurality of hospitals and inspection facilities but also by using inspection equipment provided at home or at work. As the inspection device, various devices may be used even if the inspection items are the same, but the above-mentioned Patent Documents 1-3 do not consider this point.

The present invention has been made in view of such circumstances, and it is possible to grasp an accurate health condition by considering the situation of the subject and to provide customized information such as advice according to the health condition. It is an object of the present invention to provide a possible information transmission device and information transmission method.

In order to achieve the above object, the information transmission device according to the first invention includes a first inspection data acquisition unit that acquires a time-series first inspection data group of a subject by a first device, and the first inspection data acquisition unit. A second inspection data acquisition unit that acquires a time-series second inspection data group of the subject by a second device capable of performing an inspection capable of interpolating the inspection data group of the above, and the first inspection data. It has a transmission information determination unit that determines transmission information to be provided to the subject by using the group and the second inspection data group, and has the first inspection data group and the second inspection data group. Complement each other's inspection timing or inspection items.

The information transmission device according to the second invention, in the first invention, the transmission information determination unit transmits the transmission information according to an inference model learned according to a change pattern of a test data group acquired by a plurality of devices. decide.
Regarding the information transmission device according to the third invention, in the first invention, the transmission information determination unit is acquired by the first inspection data group acquired by the first device and the second device. The second inspection data group is corrected for each of the first and second inspection data groups, the reliability when the corrected inspection data group is inferred as an input is calculated, and the above transmission is performed according to the reliability. Determine the information.
The information transmission device according to the fourth invention is a numerical value common to each of the data included in the inspection data group when the transmission information determination unit corrects each inspection data group in the third invention. Performs four arithmetic operations on.

Regarding the information transmission device according to the fifth invention, in the first invention, the transmission information determination unit is acquired by the first inspection data group acquired by the first device and the second device. The second inspection data group is corrected for each of the first and second inspection data groups, and the plurality of corrected inspection data groups are combined into one inspection data group, and the combined inspection data group is combined. Inference is performed by inputting to the inference model, and the above-mentioned transmission information is determined based on the inference result.
Regarding the information transmission device according to the sixth invention, in the first invention, the transmission information determination unit is acquired by the first inspection data group acquired by the first device and the second device. The second inspection data group is corrected for each of the first and second inspection data groups, and each of the corrected plurality of inspection data groups is input to the inference model, and the inference result by each inference model is displayed. A comprehensive judgment is made, and the above-mentioned transmission information is determined based on the judgment result.

Regarding the information transmission device according to the seventh invention, in the first invention, the first and second inspection data acquisition units are either the inspection data group from the subject or a person other than the subject. If it is the inspection data group from the subject, it is acquired as the first inspection data group or the second inspection data group.

The information transmission device according to the eighth invention is the first to seventh inventions, wherein the first and second inspection data groups are a color sensor, a shape sensor, a hardness sensor, and an olfactory sensor (line) for defecation. The data is obtained according to one of the output results of (including the reaction determination of insects and animals), the gas component sensor, the color change detection sensor when a specific reagent is added, and the shape determination by the magnified observation image.

The information transmission method according to the ninth invention is capable of performing an inspection capable of acquiring a time-series first inspection data group of a subject by a first device and interpolating the first inspection data group. The second inspection data group in time series of the subject is acquired by the second device, and the transmission information to be provided to the subject is transmitted using the first inspection data group and the second inspection data group. Determined, the first inspection data group and the second inspection data group complement each other's inspection timing or inspection item.

According to the present invention, an information transmission device and an information transmission method capable of grasping an accurate health condition by considering the situation of a subject and providing customized information such as advice according to the health condition are provided. can do.

It is a block diagram which shows the structure of the information transmission system which concerns on one Embodiment of this invention. It is a graph which shows the time-series change of the inspection data of a subject in the information transmission system which concerns on one Embodiment of this invention. It is a graph which shows the time-series change of the inspection data of a subject in the case where there are a plurality of devices which measure the inspection data of a subject in the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows an example of the operation of the inspection result transmission in the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows another example of the operation of the inspection result transmission in the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows the operation of inference by combining the historical data in the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows the operation of the inference of making an inference specification in the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows the operation of the inference model creation in the information transmission system which concerns on one Embodiment of this invention. It is a graph which shows the time-series change of the inspection data of a subject in the case where there are a plurality of devices which measure the inspection data of a subject in the modification of the information transmission system which concerns on one Embodiment of this invention. It is a flowchart which shows the operation of the inference of making an inference specification in the modification of the information transmission system which concerns on one Embodiment of this invention.

Hereinafter, an example in which the present invention is applied to an information transmission system will be described as an embodiment of the present invention. In the present embodiment, as an example of grasping an accurate health condition by considering the situation of the subject and providing customized information, daily inspection data on the health condition is monitored by a first device, a second device, or the like. , An information transmission system capable of providing health-related information based on this information will be described. This information transmission system monitors test data on the health condition of the subject using multiple devices on a daily basis. If information is collected only by a single device, there is a restriction that information can be collected only by that device, and the information that can be collected is limited. Further, in the case of a single device, if an error occurs due to the device, the usage environment of the device, the installation environment, or other restrictions peculiar to the device, the error or the like affects at the time of determination. Therefore, in the present embodiment, the inspection data is collected from the devices of a plurality of devices, and the inspection data is combined to make a judgment. As a result, we are trying to grasp the accurate health condition of the subject by taking measures against the situation where the accuracy could not be improved by a single device.

That is, acquiring inspection data using a plurality of devices is easier than the case of forcing an inspection using the same device, without the burden of binding on the user, and the data acquisition is easy, and the data is unknowingly acquired. It is convenient because it can be provided. In addition, since inspection data can be acquired under various circumstances, it is possible to increase the amount of data. For example, in order to acquire environment-dependent stress such that blood pressure rises only in the workplace and physical condition changes depending on the season and the course of the day, it is preferable to effectively utilize the data of various devices. That is, in order to effectively utilize the data of various devices, by treating the data as a group of inspection data, it is possible to grasp the health condition of the user in more detail. In addition, according to the information acquired from the device specified at a specific time of a specific person, it is possible to identify which device the information is from on a common time axis by color coding or the like, and the transition of the information value is shown. By creating a graph, it is possible to grasp the tendency of a person's health condition.

In other words, compared to a general graph that only displays the data changes of the same item, the one that displays the data changes of different items in different colors on the same graph provides more comprehensive information. May be obtained. Therefore, if the first inspection data group and the second inspection data group are acquired so as to complement each other's inspection timing or inspection item, and can be expressed in a comprehensive and comprehensive expression in an identifiable manner, this expression is used. It is possible to make a judgment or judgment using, and it is possible to make an inference by an inference model using this expression. After obtaining the inference result, a judgment / judgment may be made based on the inference result, and this judgment / judgment result may be connected to another control. Judgment and judgment may adopt a rule-based method or a pattern matching method. As a method of inference using deep learning, simply, the graph itself is represented as an image, an inference model of the result of learning using this as teacher data is prepared, and a similar graph is input to this inference model. You just have to make inferences.

However, in this case, there is a possibility that the measurement results may deviate due to differences in the equipment, specifications, installation environment, etc. of multiple devices. Therefore, we take measures against this measurement deviation and take advantage of the large amount of data. In addition, it is possible to consider the situation of the user at the time of inspection, generate customized information such as advice in consideration of the situation, and inform the user of this customized information.

Further, in the information transmission system according to the present embodiment, the inspection data group consisting of time-series inspection data acquired by different devices depends on the device and the environment, so even if there is a difference in the level of each inspection data. As long as the data of the same person is used, the tendency is the same when viewed as a similar temporal change pattern of biological information. Therefore, the deviation of the inspection data when the inspection is performed by a plurality of devices is performed by addition / subtraction calculation or multiplication / division calculation for each inspection data group so that the levels of the individual data of each inspection data group are substantially the same. The correction is performed to eliminate the problem (see, for example, Graph 34 in FIG. 3). By performing this correction, a plurality of inspection data groups can be treated as a single inspection data group. Therefore, it is possible to increase the data and grasp the accurate health condition of the subject.

In addition, it is possible to reduce the difference between devices by changing the meaning of data for each device and giving a difference in data weighting. If the number of target devices increases, it becomes possible to reduce the influence of errors of individual devices while increasing the number of data when evaluating in chronological order. Also, if you know which data (including time information) comes from which device, even if a device becomes unreliable at a specific timing for some reason, only the data before that specific timing Can be devised to make a judgment using the adopted data. For example, if abnormal data is output due to a malfunction of a specific device, a problem may occur that causes the user who uses that device to provide incorrect information, but the results of other devices may be reported. By reference, this problem can be prevented. In addition, if it is possible to determine which device the user frequently uses, it is possible to use it so that the inspection results using other devices are reflected as necessary while making a judgment solely on that device. It becomes. In this case as well, by "reflecting" this, it is possible to prevent problems specific to the device.

Further, in the information transmission system according to the present embodiment, the deviation of the inspection data when the inspection is performed by a plurality of devices is applied to each inspection data group so that the levels of the individual data of each inspection data group are substantially the same. On the other hand, correction is performed by addition / subtraction calculation or multiplication / division calculation, and the corrected inspection data group is input to the inference model to calculate the reliability of the inference at this time (see, for example, S73 and S77 in FIG. 10). .. If the amount of correction in the correction calculation is changed little by little, the reliability also changes little by little. The information transmission system adopts the inference result when this reliability is the highest.

Further, in the information transmission system according to the present embodiment, the reliability is calculated while changing the correction amount in the correction calculation little by little, and the inference result when the reliability is the highest is adopted as the inference result of each inspection data group. do. The information transmission system makes a comprehensive judgment based on each inference result, and uses this judgment result as the inference result (see, for example, FIG. 9).

The subject in this embodiment is a person who may become a patient depending on the re-examination. In addition, depending on the results of the re-examination, this subject is also a person who can regain self-confidence in health, do not have to worry about illness, and enjoy daily life. In addition, he is also a person who can become healthy by simple improvement and treatment of life.

The information transmission system according to the present embodiment is composed of, for example, a server, but may be configured by a personal computer capable of exchanging information with the server, a mobile information device such as a smartphone, or the like.

Further, in the present embodiment, in determining the information to be transmitted to the target person, the inspection result of the target person, the equipment required for further inspection based on the inspection result, and the facility with this equipment are used. To search and / or infer. In order to perform this search / inference, it is advisable to provide a database (DB) for storing facilities having facilities. In addition, in providing the transmitted information, information such as the facility name, access method including telephone, e-mail, map, etc., consultation time, free time, cost estimation, etc. may be included. Further, the number of facilities is not limited to one, and may be multiple.

Next, the configuration of the information transmission system according to the embodiment of the present invention will be described with reference to FIG. This information transmission system includes a control unit 1, a first device 2a, a second device 2b, a third device 3, a terminal 4, a learning unit 5, a learning request unit 6, an inference engine 7, a database (DB) unit 8, and a related inspection. It consists of 9 institutions (including medical institutions). Of these units, the control unit 1 is arranged in the server, and is the first device 2a, the second device 2b, the third device 3, the terminal 4, the learning unit 5, the learning request unit 6, the inference engine 7, and the DB unit. 8. The related inspection organization 9 can connect to the server through a network such as the Internet. However, the present embodiment is not limited to this configuration, and for example, the control unit 1, the first device 2a, the second device 2b, the third device 3, the learning unit 5, the learning request unit 6, the inference engine 7, and the like. Any one or more of the DB units 8 may be arranged in the server, and the others may be arranged in another server or an electronic device such as a personal computer. Further, the related inspection agency 9 may have a server function.

The control unit 1 is a controller (processor) that controls an information transmission system according to the present embodiment, and is a CPU (Central Processor Unit), a memory, and an HDD that provide files and data to a server or the like or other terminals via a network. It is assumed that the IT device is composed of (Hard Disc Drive) and the like. However, the control unit 1 is not limited to this configuration, and when it is constructed as a small-scale system, it can be configured with something like a personal computer. The control unit 1 has various interface circuits, can cooperate with other devices, and can perform various arithmetic controls by a program.

The control unit 1 receives information from each linked device, organizes the information, generates necessary information, and provides this information to the user. The control unit 1 also has a function of outputting a request to each of the linked devices and operating each device. In the present embodiment, assuming a high degree of freedom and usability of the system, wireless communication or wire communication is performed between the device such as the first device 2a or the terminal 4 or the like owned by the target person (also referred to as a user) and the control unit 1. It is possible to connect by communication. As the communication for this purpose, a wireless LAN or a mobile phone communication network is assumed, and short-range wireless communication such as Bluetooth (registered trademark) or infrared communication may be used in combination depending on the situation. The description of the communication unit including the communication circuit, the antenna, the connection terminal, etc. is complicated, so it is omitted in FIG. 1, but the communication unit having the communication circuit, etc. is provided in the part of the arrow indicating the communication in the figure. Has been done.

The control unit 1 includes a communication control unit 1a, an ID determination unit 1b, an information provision unit 1c, an inference model specification determination unit 1d, an inference request unit 1e, and a search unit 1f. Each of these parts may be realized by software by a CPU, a program, or the like in the control unit 1, may be realized by a hardware circuit, or may be realized by coordinating software and a hardware circuit. You may. Further, in FIG. 1, since each unit in the control unit 1 cooperates with each other to perform each function, the direction of the signal is omitted, but this will be described separately with a flowchart. For example, in a step like S1 in FIG. 5, the ID determination unit 1b collects information from the first device 2a, the second device 2b, and the like for each of the same users.

The communication control unit 1a has a communication circuit and the like, and has a first device 2a, a second device 2b, a third device 3, a terminal 4, a learning unit 5, a learning request unit 6, an inference engine 7, and a database (DB) unit 8. , And the communication unit provided in the related inspection organization 9, and sends and receives data and the like. Each device / part such as the first and second devices 2a and 2b, the third device 3, and the terminal 4 also has a communication unit, but the illustration is omitted in FIG. 1 because it is complicated.

The ID determination unit 1b collects information for each same user from the first device 2a and the like. An ID is assigned to each individual in order to identify the individual whose information has been acquired by the first device 2a, the second device 2b, the third device 3, and the related inspection organization 9. In the present embodiment, since the data of each user is handled, the ID determination unit 1b manages which user's information is received and which user is given the guide. In the determination of the specific user, the first device 2a, the second device 2b, and the third device 3 have a biometric authentication function, the user inputs an ID by the terminal 4, and the user uses the first and second devices 2a. This is performed by transmitting an ID through the communication unit in 2b or by reading a unique code from the terminal 4. In addition, in order to protect personal information, management will be strict by encrypting the necessary parts, but since these are general-purpose technologies, detailed description will be omitted.

The ID of each device may include information on the model name of the device and unique information indicating which individual it is. The function and performance of the sensor to be mounted may be known from the model name, and the installation location and usage environment may be known from the individual information, and such information may be searchable through a network or the like. If the model name is known, it is possible to determine information on similar devices, and from the installation location and usage environment, determine latitude / longitude, indoor / outdoor, season, weather, temperature characteristics, etc., and take this determination result into consideration. Then, the output information of the device may be corrected.

The ID determination unit 1b functions as a first inspection data acquisition unit that acquires a time-series first inspection data group of the subject by the first device (see, for example, S101 in FIG. 4 and S1 in FIG. 5). ). Further, the ID determination unit 1b acquires the second inspection data group in time series of the subject by the second device capable of performing the inspection so as to interpolate the first inspection data group. It functions as an acquisition unit (see, for example, S105 in FIG. 4 and S1 in FIG. 5). The ID determination unit 1b functions as a second inspection data acquisition unit that acquires a time-series second inspection data group of the subject by the second apparatus capable of performing the same inspection as the first apparatus. The first inspection data group and the second inspection data group complement each other's inspection timing or inspection item.

The first inspection data acquisition unit and the second inspection data acquisition unit determine whether it is an inspection data group from a target person or an inspection data group from a person other than the target person, and inspect from the target person. If it is a data group, it is acquired as a first inspection data group or a second inspection data group. The acquired first inspection data group and second inspection data group are recorded in the recording unit. The test data group other than the target person may be recorded so that it can be converted into teacher data by associating with the health information of the person other than the target person.

The information providing unit 1c has a function of acquiring user information (may refer to the result acquired by another device) in order to provide correct information to the user. In addition, the information providing unit 1c acquires the inspection data of the user (specified by the ID) acquired from the first device 2a and the like and the related inspection organization 9. Further, the information providing unit 1c uses the acquired inspection data, various information acquired from the related inspection institution 9, information on the possessed device stored in the DB unit 8, user profile information, and the like to determine the user's health condition. to decide. The health condition includes a disease that is currently present and a disease that may develop in the future, and when the health condition is determined, the user is provided with information related to the health condition. In addition, when the user's illness or the like is determined, information on the facility to be examined or treated is provided to the user as necessary.

In addition, in order to confirm the health condition of the user in a specific situation, the control unit 1 checks the current hospital visit status, information such as prescription drugs, past health examination results, etc. by the user's ID and the like. If the institution 9 can be referred to, it becomes easy to determine the association with the device data. This is a security problem because the user who operates the terminal 4 permits the cooperation, or the doctor who operates the related inspection institution (IT device) 9 permits the cooperation. Can be dealt with.

That is, the information providing unit 1c recommends information on health to the user, for example, information that when the user will visit the facility to receive the test or treatment, or a facility suitable for receiving the test or treatment. Provide information about. The information providing unit 1c acquires the inspection data transmitted from the first device 2a and the like and the related inspection organization 9. As will be described later, this data is inspection data (time series information) with time information, and is accumulated in a data structure that can be graphed as shown in FIGS. 2 and 3. In the present embodiment, it is assumed that the control unit 1 provides information to the user by using the information from the devices in the first device 2a and the like and the related inspection organization 9 and the like. The server may be a modified example in which information is collected in the same manner.

Further, in order to provide such information, the information providing unit 1c collects inspection data from the first device 2a, the second device 2b, and the like, and records the inspection data in the DB unit 8. The frequency of information acquisition and the number of data may differ depending on the first device 2a, the second device 2b, and the like. In other words, the increase and decrease of specific health-related numerical values obtained with various devices are arranged in chronological order, and the numerical values measured by changing the devices can be arranged for each device.

For example, even if the blood pressure is the same, the test data measured by a wearable simple device and the test data measured by a dedicated device at the event venue are recorded separately. Assuming that the wearable device is worn almost all the time, test data that increases or decreases depending on the condition can be obtained in the morning, day and night, before meals, before bedtime, after bedtime, and before waking up. In addition, data measured with a dedicated device can be obtained in a form with higher accuracy or more incidental information (such as an inspector or a subjective symptom obtained by an inspector's hearing), although it is sporadic. In addition, in the case of measurement with a dedicated device, the test data is accurately obtained by taking off clothes, skipping meals, and eating a dedicated meal. Under such circumstances, there are few differences such as the position and items of the human body to be measured differ from person to person, and the error of the equipment is strictly controlled, so it is an appropriate situation to compare the absolute value with other people. ing.

In addition, the information providing unit 1c may acquire lifestyle habits such as the user's address, behavioral style at the place of work, eating habits, bedtime, and meal timing on the Internet, and the acquired information is also taken into consideration. Therefore, information such as facilities to be provided to the user may be generated. The acquisition of this information can be complemented by general-purpose or well-known technology. The information providing unit 1c may also customize the information of the facility or the like generated by acquiring the information. Profile information about this facility is obtained as medical institution information from the related inspection institution 9.

In addition, the information providing unit 1c uses the information of the possessed equipment, etc. stored in the DB unit 8 in addition to the information collected from the related inspection organizations 9 such as the first equipment 2a, etc., when providing the information of the recommended facilities, etc. .. Of course, the information recorded in the DB 8 may be recorded in a different recording unit other than the DB unit 8. In this case, there are a plurality of DB parts in FIG. 1, but they are omitted because they are complicated. The information providing unit 1c collects various kinds of information in providing the information. That is, the information providing unit 1c functions as an acquisition unit for acquiring the user's examination data, the user's profile information, and the possessed device information for each examination / medical institution.

The information providing unit 1c functions as a transmission information determination unit that determines transmission information to be provided to the target person by using the first inspection data group and the second inspection data group (for example, S107 in FIG. 4 and FIG. 5). S9, S79 in FIG. 10 and the like). The transmission information determination unit determines transmission information according to an inference model learned according to a change pattern of a test data group acquired by a plurality of devices (for example, S107 in FIG. 4, S9 in FIG. 5, S79 in FIG. 10 and the like. reference).

The transmission information determination unit selects the first inspection data group acquired by the first device and the second inspection data group acquired by the second device for each of the first and second inspection data groups. The corrected inspection data group is input to, the reliability of the inference result at this time is calculated, and the transmission information is determined according to the reliability (for example, S107 in FIGS. 3 and 4 and S9 in FIG. 5). , FIG. 9 and FIG. 10). When making corrections for each inspection data group, the transmission information determination unit performs four arithmetic operations on the numerical values common to each of the data included in the inspection data group.

As the four arithmetic operations, for example, in the case of a device that acquires the same biological information, the reliability of the inference may be calculated while making corrections for each device, and the highly reliable one may be used as the inference result. When performing the four arithmetic operations, the constants of the specific four arithmetic operations are changed little by little in the time series data for each device. By this processing, the reliability is increased in the situation where the error is corrected, so that correct inference is possible. With such a device, if the device acquires similar biological information, it is possible to provide reliable information regardless of the error and noise that may be affected by the sensitivity and usage environment. In addition, it is not always necessary to have the same biological information, and even if different data such as pulse, heart rate, and respiratory rate are used, if they have similar changes including the magnitude relationship of the data, correction and reliability judgment are performed. Allows comprehensively correct inference.

The transmission information determination unit selects the first inspection data group acquired by the first device and the second inspection data group acquired by the second device for each of the first and second inspection data groups. The corrected test data group is combined into one test data group, the combined test data group is input to the inference model to perform inference, and the transmitted information is determined based on the inference result. (See, for example, Graph 34 in FIG. 3, S107 in FIG. 4 and the like). The transmission information determination unit selects the first inspection data group acquired by the first device and the second inspection data group acquired by the second device for each of the first and second inspection data groups. Is corrected to, and each of the corrected plurality of inspection data groups is input to the inference model, the inference result by each inference model is comprehensively judged, and the transmission information is determined based on the judgment result (for example, FIG. 9. See FIG. 10).

As described above, the information providing unit 1c acquires inspection data that is a time-series pattern of the user for a specific period. This acquired time-series pattern is composed of individual inspection data acquired by measurement at a plurality of different timings, not simply data obtained by one-time measurement, and even changes in the inspection data pattern are used as information. .. By using a time-series pattern consisting of multiple inspection data, it is less susceptible to errors caused by changes in the measurement environment and conditions. Furthermore, it infers the health condition from the end of the specific period to the future period (when the specific period is extended), and makes it possible to predict the future.

In addition, teacher data can be created by adding the timing information of the user's examination / visit to the medical institution as annotation information to the acquired time series pattern. If there is an inference part that has an inference model generated by learning using this teacher data, what is the timing (when the specific period is extended) beyond the specific period (period for acquiring the time series change pattern)? Can be inferred if Further, if the user's disease name or the like is known, teacher data to which this information is added as annotation information can be generated. By learning using this teacher data, it is possible to generate an inference model that infers health information such as illness. When generating the inference model used here, the specifications of specific input / output information are specified and learning is performed.

Therefore, in the present embodiment, the time-series change pattern of the user's inspection data is input to the inference unit, the inference unit makes an inference, and based on this inference result, the transmission information at the timing beyond the specific period is determined. A transmission information determination unit is provided. Therefore, it is possible to provide a system, an apparatus, a method, a program, or the like capable of transmitting the prediction information at the timing after the inspection acquisition of the time series pattern.

The reliability of user inspection data will decline if there are differences in mechanical performance for each inspection device. For example, when the same user acquires test data (biological information) from a plurality of devices at the same time, the same test data may not be obtained. Therefore, if a large amount of change pattern information of inspection data is acquired by repeating inspections at different dates and times using multiple inspection devices (inspection devices with specific specifications) that can inspect the same inspection item, it is treated as big data. It becomes possible. In this case, the specific period does not have to be a fixed period, and may be a different time width (specific period) depending on the situation. The "time width" here is not the time between measurement timings (inspection interval / measurement interval), but the time interval from the first measurement to the last measurement when acquiring a series of inspection data. say. The "time width" may be rewritten as a specific time width including a lot of time series data in this time width and change pattern information of the inspection data.

In the present embodiment, the information providing unit 1c inputs a change putter of the inspection data into the inference engine 7 in which the inference model generated by the learning unit 5 is set, obtains an inference result regarding advice, and inputs the input inspection data. Provide to users corresponding to. This service may use personal information, and may require a contract for personal information in order to receive advice. In that sense, the user's profile information may be important. In addition, when the user is an infant or an elderly person, advice may be delivered to a person who takes care of the user, a caregiver, or the like. This also receives valid information such as advice according to the information managed by the user's profile information.

The inference model specification determination unit 1d determines the specifications of the inference model to be generated when the inference request unit 1e requests the learning unit 5 to generate the inference model through the learning request unit 6. The control unit 1 acquires the biometric information of the user from the first device 2a and the like, and accumulates the biometric information. The control unit 1 requests the learning unit 5 to generate various inference models through the learning requesting unit 6 using the accumulated biological information as teacher data. The inference model specification determination unit 1d determines what kind of specification the inference model is requested in generating the inference model. For example, as shown in FIG. 2A, which will be described later, when time-series biometric information is accumulated, what kind of inspection data (value) does the inference model specification determination unit 1d have, and what is the user? Determine the specifications of an inference model to infer whether you will be treated in a medical facility in a day. In addition, the inference model specification determination unit 1d may, based on time-series biometric information, what kind of disease it currently has, what kind of disease it may have in the future (when), and whether it will further suffer it. Determine specifications to generate inference models that infer the recommended facilities to receive the necessary tests and treatments if they do not.

The inference request unit 1e requests the learning unit 5 to generate an inference model of the specifications determined by the inference model specification determination unit 1d through the learning request unit 6. That is, the inference requesting unit 1e requests the learning unit 5 to generate an inference model through the learning requesting unit 6 when a predetermined number of biological information acquired by the first device 2a or the like is accumulated, and the generated inference is generated. The model is received through the learning request unit (or directly from the learning unit 5). This received inference model is transmitted to the inference engine 7. The control unit 1 may prepare a plurality of inference models and appropriately select the inference model according to the information to be provided to the user.

Based on the user's biometric information acquired by the first device 2a, the second device 2b, and the third device, the search unit 1f may be affected by the current disease or any disease in the future (when). Or, when it is found that further examination or treatment is necessary, the examination institution or medical institution having the equipment necessary for the examination or treatment is searched in the database stored in the DB unit 8. These pieces of information may be obtained by inference using the inference engine 7, but may match the accumulated data. Since there are such cases, in the present embodiment, the search unit 1f can be used for searching.

The first device 2a and the second device 2b are devices for acquiring test data such as user health-related information such as vital information and sample information. The first device 2a and the second device 2b are inspection devices having specific specifications, and are devices capable of inspecting the same type (similar) health-related information. It suffices if the inspection data groups acquired by the first device 2a and the second device 2b can perform an inspection that can interpolate both data when the inspection timings are different from each other. Further, the first device 2a and the second device 2b do not have to inspect exactly the same inspection items. For example, even when the heart rate is measured while measuring the blood pressure, both data are interpolated with each other. Can be done. Note that FIG. 1 shows only two devices, the first device 2a and the second device 2b, as devices for acquiring the user's inspection data, but the number is not limited to two and is three or more. May be good. Further, as will be described later, in the present embodiment, the third device 3 is assumed as a device for acquiring inspection data of a person other than the user.

There are various kinds of health-related information acquired by the first device 2a and the like, and for example, there is vital information such as the user's body temperature, blood pressure, and heartbeat. The health-related information includes various sample information such as excrement such as urine and stool of the user, sputum and blood. In the case of stool, the first device 2a and the second device 2b acquire the color, shape, amount, and date / time information. The first device 2a and the second device 2b may acquire information according to an instruction from the control unit 1, may acquire information according to a user's operation, or automatically acquire information. May be good. Furthermore, the first device 2a, etc. is used for daily life, work / school activities, meals, sports activities, etc. in the information "Personal Health Records (PHR)" which is medical / health information. Personal Life Records (PLR), which includes various activity data of daily life, may be collected and utilized. The acquired information is obtained through the communication unit (not shown) in the first device 2a, etc. , Is transmitted to the control unit 1.

The inspection data of the subject detected by the first device 2a or the like is obtained by acquiring the inspection data in time series using an inspection device having a specific specification and extracting the change pattern information of the inspection data in a specific time width. Is. That is, as the first device 2a and the second device 2b, inspection devices of specific specifications (inspection devices of the same type) are used, and the first device 2a and the like measure the inspection items of the same subject at different timings. By doing so, data is acquired in chronological order. Using this time-series data, a change pattern can be obtained by drawing the measured values on a graph according to the inspection timing. By extracting this change pattern in a specific time width, a test data group can be obtained. The inspection data is based on a color sensor for defecation, a shape sensor, a hardness sensor, an olfactory sensor (including reaction judgment of nematodes and animals), a gas component sensor, a color change detection sensor when a specific reagent is added, and a magnified observation image. It is the data obtained according to one of the output results of any of the shape determinations.

When the first device 2a and the second device 2b obtain the information about the specific user, the information about the facility recommended by the information providing unit 1c of the control unit 1 is presented to the information terminal 4 of the specific user. The explanation is given on the assumption that this presentation assists the user's behavior, but various variations can be considered.

The extent to which the information determination performed in the first device 2a or the like is determined may be changed in relation to the control unit 1. For example, it may be transmitted to the control unit 1 without determining only the result of sensing by the first device 2a or the like. However, in this case, it is necessary to attach information on what kind of data of what kind of person to the sensing signal and transmit this signal. It is preferable that the attached information is associated with which person and which sensing result, but it may be associated with the information of another terminal by adding it.

The third device 3 is a device that acquires data of a person different from the user who uses the first device 2a and the second device 2b. Although only one third device 3 is shown in FIG. 1, there may be a plurality of the third devices 3, and an unspecified number of devices are collectively represented in FIG. This makes it possible to record and manage what kind of person has what kind of illness and what kind of health value is in big data.

The third device 3 composed of this unspecified majority may acquire different numerical values with different performances. The more such a device participates in this system as a health management device, the more various data can be used as a health monitoring device. In an extreme case, if the results of daily selfies taken by each person's smartphone and other health figures of that person are combined into big data, the complexion will look long before the person became ill. It is possible to utilize data such as whether it has become worse. This data can be used to advise others on early health care, abstinence, and treatment in the event of similar complexion changes.

When a wearable terminal is used as the first device 2a, the second device 2b, and the third device 3, it adheres to the skin or the vicinity of the body depending on the wearing part of the wearable terminal, and the body temperature, heart rate, blood pressure, brain wave, line of sight, etc. It is possible to obtain vital information such as breathing and exhalation. In addition, as a scale, a sphygmomanometer, and a measuring instrument for measuring arterial stiffness, which means the hardness of the arterial wall, dedicated precision equipment is installed in health facilities, public baths, pharmacies, shopping malls, etc. The measurer may also be assigned. In such facilities, users often use the measuring device comfortably in their spare time and manage their physical condition based on the measurement results at that time. These measuring devices may be the first device 2a, the second device 2b, and the third device 3.

In addition, the first device 2a, the second device 2b, and the third device 3 may be requested to fill out a questionnaire before and after the user uses a dedicated terminal or the like. In such cases, the user's profile information and other information can be identified based on the description in this questionnaire. Such information collection is not limited to the first device 2a and the like, and may be performed by the control unit 1. This information can be used when determining whether or not the specific information has been acquired in step S3 of FIG. 5, which will be described later. If information such as when the doctor went to the doctor can be heard, it can be used as the time Tc information in FIGS. 2 (a) and 2 (b) described later.

The first device 2a, the second device 2b, and the third device 3 may be a thermometer or a sphygmomanometer that is already suffering from a specific disease and is used under the guidance of a doctor. In addition, when the color and facial expression of the face and claws taken by the camera of the smartphone, the image of the affected area, and the voice when the throat becomes strange are picked up by the microphone, the mobile terminal (smartphone) is used as it is. It can be 1 device 2a, 2nd device 2b, and 3rd device 3.

Recently, simple health management devices and health information acquisition devices have been developed, and these devices may be installed in wearable devices. Such devices are also treated as peripheral devices for smartphones, not stand-alone devices. Since there are many cases, this may also be assumed as a mobile terminal. In addition, even if it is not a wearable device, a simple measuring device may be installed in a place where people gather to provide a health information service. Such a device may be used as the first device 2a, the second device 2b, and the third device 3.

The related inspection institution 9 is a facility where the user is inspected, and there are, for example, an inspection facility and a medical facility. Of course, the related inspection organization 9 may be a mobile type, for example, a type in which general medical equipment or inspection equipment is mounted on an automobile, train, ship, helicopter, drone, or the like and the patient goes to the patient. The control unit 1 can acquire which medical institution the patient went to, what kind of test result was obtained, and the like from the server that operates the system of the related test institution 9. Of course, the server of the related inspection organization 9 may be the same as that of the control unit 1, or some functions may be shared.

As described above, the terminal 4 is a mobile information terminal, and is a device for receiving information that can be confirmed by the user and related persons. As information, there are health information and facilities recommended according to the health condition. The terminal 4 may be, for example, a smartphone or a tablet PC. In this case, the built-in camera or microphone can be used as an information acquisition unit. Further, a wearable terminal or other home appliances that can be linked may be used as the terminal 4, and information may be acquired by the wearable terminal or the like. Therefore, the first device 2a or the second device 2b and the terminal 4 may be the same, or may be dedicated devices, respectively. The terminal 4 linked with the wearable terminal may acquire information and manage the information. Further, depending on the situation, the functions of the control unit 1 may be possessed by the first device 2a, the second device 2b, the third device 3, and the terminal 4, and the detection, control, and information provision are shared. The configuration may be different.

The database (DB) unit 8 has an electrically rewritable non-volatile memory. The DB unit 8 has a data history list for each ID, and this list records the relationship between the acquired data and the inspection date. As described above, since the ID determination unit 1b receives the inspection data from the first device 2a and the like, the related inspection organization 9 and the like, the DB unit 8 records the inspection data for each ID. At this time, the inspection date, inspection equipment (first equipment, second equipment, third equipment, related inspection organization, etc.), inspection location, inspection items, etc. are also recorded. In addition, record how and what the test is for. The DB unit 8 organizes the acquired data into 5W1H, that is, WHO (who), WHERE (where), WHERE (date and time), WHAT (which inspection), WHY (why), and HOW (how). This organized data may be recorded.

Further, the DB unit 8 may have a owned facility recording unit that records a list of owned equipment for each facility and a visit history recording unit that records an ID and the visit of the user for each facility. The holding facility recording department records a list of equipment owned by facilities such as hospitals, clinics, and inspection institutions. The information providing unit 1c can present the information of the facility having the most suitable equipment for the inspection to the user by searching the possessed facility recording unit. In order to update the information when the medical facility or the like replaces the device, the information may be linked with the information of the related inspection institution 9. In addition, the visit history recording unit records visit information indicating which person (identified by ID) came at what time for each facility.

The DB unit 8 is constructed as a part of an information transmission system in which medical facilities cooperate, and the DB unit 8 may also be able to access the related inspection institution 9 through the control unit 1. In this case, when the DB unit 8 receives a search command from the control unit 1, the DB unit 8 searches the data in the related inspection organization 9 in addition to the data recorded in the DB unit 8 to search. Output the result. The DB unit 8 functions as a storage unit that stores the user's profile information and the possessed device information for each examination / medical institution. The storage unit is not limited to the DB unit 8, and all or part of its functions may be arranged in the control unit 1 or the like.

When the learning request unit 6 receives a request for generating an inference model from the inference request unit 1e in the control unit 1, it conveys the specifications of the inference model to the learning unit 5 and requests the generation of the inference model according to the specifications. The learning request unit 6 includes a data classification recording unit 6a, a specification setting unit 6d, a communication unit 6e, and a control unit 6f.

The control unit 6f is a controller (processor) that controls the inside of the learning request unit 6, and is a CPU (Central Processor Unit), a memory, and an HDD (Hard Disc) that provide files and data to a server or the like or other terminals via a network. It is assumed that the IT equipment is composed of Drive) and the like. However, the control unit 6f is not limited to this configuration, and when it is constructed as a small-scale system, it can be configured with something like a personal computer. The control unit 6f has various interface circuits, can be linked with other devices, and can perform various arithmetic controls by a program.

The data classification unit 6a has an object type A image group 6b, and the teacher data 6c is recorded in the image group 6b. The object type A image group 6b is an image group used when the learning unit 5 generates an inference model, and has a large number of image groups such as type A, type B, and so on. Teacher data 6c is generated based on this image group. That is, as shown in FIGS. 2 and 3, a graph can be drawn by plotting the inspection data for each inspection date, and this graph can be treated as an image. The data record classification unit 6a records the teacher data 6c based on the data history list recorded in the DB unit 8.

The specification setting unit 6d sets what kind of inference model is generated based on the inference model specifications determined by the inference model specification determination unit 1d. Further, the teacher data is generated from the data recorded in the history list of the DB unit 8 so as to satisfy this specification.

The communication unit 6e has a communication circuit for communicating with the control unit 1 and the learning unit 5. Through the communication unit 6e, the control unit 1 requests the generation of the inference model, and the learning unit 5 requests the generation of the inference model.

The learning unit 5 has an input / output modeling unit 5a, and generates an inference model by machine learning or the like according to the specifications from the learning request unit 6. The input / output modeling unit 5a has a specification collation unit 5b. The specification collation unit 5b determines whether or not the specifications received from the learning request unit 6 and the inference model generated by the input / output modeling unit 5a match. That is, the specification collating unit 5b defines not only the input / output relationship but also the learning method so as to perform learning according to the "required specifications" such as the time required for inference of this inference model, energy, and circuit configuration. Is.

The inference model is generated by learning the relationship between the acquired information such as acquired biometric information and biopsy information and the disease, and specifically by learning the relationship between the acquired information and the clinical department / department. Similar to the inference engine 7, the input / output modeling unit 5a has an input layer, a plurality of intermediate layers, and an output layer, obtains the strength of the connection of neurons in the intermediate layer by learning, and generates an inference model.

In generating such an inference model, the learning requesting unit 6 extracts the change pattern of the test data acquired from the subject using the test device in a specific time width, and uses the extracted change pattern in the inference engine 7. From the timing when the subject inspects the data, the teacher data is generated with the health advice to be output at a later timing as the inference information. Then, the learning unit 5 generates an inference model by performing learning using the teacher data. In this embodiment, the time width that goes back from the time when the test result is obtained has been described, but if the data improves after the test result due to treatment or the like, the treatment may not be successful. You may learn and output prognostic (after getting sick) advice.

In addition, the learning unit 5 can generate an inference model capable of giving future prediction advice on lifestyle-related improvement and treatment and medication effects by learning using the test data sequence after examination, hospital visit, and medication. You can also do it. In this case, the time-series data is used starting from the time of examination, outpatient visit, and medication. Use the previous time-series data when giving advice on tests, hospital visits, medications, etc.

Here, deep learning will be described as an example of learning performed by the learning unit 5. "Deep learning" is a multi-layered structure of the process of "machine learning" using a neural network. A typical example is a "forward propagation neural network" that sends information from front to back to make a judgment. The simplest forward-propagating neural network is an input layer consisting of N1 neurons, an intermediate layer consisting of N2 neurons given by parameters, and N3 corresponding to the number of classes to be discriminated. It suffices if there are three layers of output layers composed of the above neurons. Each neuron in the input layer and the intermediate layer, and each neuron in the intermediate layer and the output layer is connected by a connection weight, and a logic gate can be easily formed in the intermediate layer and the output layer by applying a bias value.

The neural network may have three layers as long as it makes a simple discrimination, but by increasing the number of intermediate layers, it is possible to learn how to combine a plurality of features in the process of machine learning. In recent years, those having 9 to 152 layers have become practical from the viewpoints of learning time, determination accuracy, and energy consumption. Alternatively, a "convolutional neural network" that compresses the feature amount of the image, performs a process called "convolution", operates with the minimum processing, and is strong in pattern recognition may be used. In addition, a "recurrent neural network" (fully connected recurrent neural network) that can handle more complicated information and whose meaning changes depending on the order or order may be used.

In order to realize these technologies, a conventional general-purpose arithmetic processing circuit such as a CPU or FPGA (Field Programmable Gate Array) may be used. However, not limited to this, since most of the processing of neural networks is matrix multiplication, even if a processor called GPU (Graphic Processing Unit) or Tensor Processing Unit (TPU) specialized in matrix calculation is used. good. In recent years, such a "neural network processing unit (NPU)" dedicated to artificial intelligence (AI) has been designed so that it can be integrated and incorporated together with other circuits such as a CPU, and has become a part of processing circuits. In some cases.

Other methods of machine learning include, for example, support vector machines and support vector regression. The learning here is to calculate the weight of the discriminator, the filter coefficient, and the offset, and there is also a method using logistic regression processing. When making a machine judge something, humans need to teach the machine how to make a judgment. In the present embodiment, a method of deriving the judgment of the image by machine learning is adopted, but in addition, a rule-based method of applying the rules acquired by humans by empirical rules / heuristics may be used.

The inference engine 7 has an input / output layer and a neural network similar to the input / output modeling unit 5a of the learning unit 5. The inference engine 7 makes inferences using the inference model generated by the learning unit 5. For example, the inference engine 7 is measured by the first device 2a or the like, inputs time-series biometric information, and infers, for example, an appropriate inspection institution / medical institution for inspecting, treating, or the like the user's health condition. Ask. In addition, based on time-series biometric information, it may be inferred when a medical institution will receive a medical examination.

As described above, the control unit 1 may provide information on the recommended facility by using the inference engine 7 in addition to the search unit 1f searching the DB unit 8. The inference engine 7 infers information about the recommended facility using the inference model generated by the learning unit 5. This inference model is generated by learning the relationship between acquired information such as acquired biometric information and biopsy information and the disease, and specifically by learning the relationship between the acquired information and the clinical department / department. In this way, the control unit 1 may output the guide information to be presented by the inference by the inference engine 7.

When the control unit 1 guides medical facilities, etc. with a single judgment based on the acquired information obtained at one time by searching or inference, it unnecessarily brings medical information into life and lives soundly and with peace of mind. May interfere with. Therefore, the accuracy may be improved by using the history (time-series information) of the acquired information a plurality of times.

FIG. 2A shows a graph using personal health-related historical data (time-series data) recorded in the recording unit 8. The recording unit 8 records, for example, the specific data A among the data acquired by the specific device A or the data of the device having various inspection functions, and when and which facility the individual visited. is doing. The control unit 1 manages the recording of the recording unit 8.

The horizontal axis of the graph shown in FIG. 2 (a) is time, and the vertical axis is health-related data. The data posted on this graph can be treated as if the information were arranged two-dimensionally. Therefore, inference can be made by a method similar to image search, in which a specific object is found from this image. In other words, the graph can be used as input and the output can be used as health advice. As advice, the measurement method, the name of the specific disease that is currently occurring, the name of the specific disease that may occur in the future, the guidance of the treatment / testing facility for the specific disease, etc. are assumed. Details of the graph showing the historical data shown in FIG. 2 will be described later.

The recording unit 8 records what kind of facility the person whose health-related information has changed is going to or is going to, and when the control unit 1 centrally manages this record, these data are collected. It is possible to have the learning unit 5 create an inference model by using it as teacher data.

The inference engine 7 has an inference model obtained by inference model specification determination unit 1d of the control unit 1 designating an inference model specification and learning according to this specification. The inference engine 7 may have a plurality of inference models because the device data may be different when a new device appears. A plurality of inference models may be prepared and appropriately selected according to the user's inspection data to determine the inference model. In addition, since new learning is required each time a new device appears, it is assumed that the inference model is often improved or newly created through the learning unit 5 by the designation of the control unit 1. There is. However, when the first device 2a or the like is dedicated and specialized only in inference of a specific disease, it may be a single dedicated inference model.

The inference engine 7 is a circuit block centered on an AI chip such as a CPU, GPU, and DSP, like the input / output unit 5a of the learning unit 5, and also includes a memory and the like to form a neural network. In the present embodiment, the inference engine 7 and the learning unit 5 are connected to a network or the like operated in cooperation with a hospital or the like, and the control unit 1 may be used in cooperation with these. I'm assuming. In this case, there is a possibility that learning and inference information can be exchanged via the related inspection organization 9.

When it is determined that the inference request unit 1e of the control unit 1 has sufficiently acquired the information of a specific user from the first device 2a or the like, the inference engine 7 is requested to make an inference. The inference engine 7 inputs an information group showing the time-series transition of similar data, infers based on this information group, and outputs medical institution information (visit information, examination information, etc.) suitable for a specific user. It is possible to do. If a group of data similar to a person with a chronic medical condition is input to the inference engine 7, it is better to display a guide to a facility where the same treatment can be performed. The inference engine 7 can estimate the medical institution that has the equipment to inspect a specific user.

In addition, since it is more likely that home appliances and other devices will be equipped with health monitoring functions in the future, these devices will provide various information in daily life without the user having to go to a place where special equipment is installed. It is acquired and effective health management can be performed without the user being aware of it. For example, many methods have been proposed in which a sensor attached to a warm water toilet seat or a toilet bowl detects the amount and color of stool, and the detection result is used for diagnosis.

Next, using FIG. 2, the biological information (examination data) acquired in time series by the first device 2a and the like will be described. As described above, in the present embodiment, two devices, the first device 2a and the second device 2b, are assumed in order to acquire the inspection data of the user. In FIG. 2, biological information (examination data) acquired from one of the two devices will be described.

FIG. 2 is a graph created using inspection data. The DB unit 8 records test data organized in chronological order for each patient ID, and FIG. 2 shows the test data in a graph. In FIG. 2, the horizontal axis represents time T, and the vertical axis plots time-series inspection data. The vertical axis plots the test data, the biological data, the vital data, and the sample data, and any one of these is plotted based on the numerical value D of the test output result of the device to be inspected. The numerical value D is, for example, a value indicating the degree of red color of stool.

Further, in FIG. 2, it is assumed that the date and time of visit and the like are automatically updated systematically. Of course, there may be a plurality of visit dates and times, but in order to avoid complication, for example, the date and time of the first visit of a specific clinical department may be used. The example shown in FIG. 2A is a case in which the time-series data changes in the direction of deterioration of health and the user eventually goes to the hospital, as will be described later. For the patient in the situation shown in FIG. 2 (a), provide the result of inferring how long and what kind of clinical department the patient will go to before the time T1. Is possible. In addition to this example, as shown in FIG. 2B, even if the patient has already visited the hospital because he / she is aware of other signs and vital data can be obtained, this information is stored in the DB. Record in Part 8. However, some people have only vital data even if they do not go to the hospital at all.

As explained earlier, Fig. 2 (a) is a case in which it is presumed that the patient will go to the hospital in the future. The graph shown in FIG. 2A shows changes in time series of examination data (equipment data) of users who are not currently visiting the hospital. From this time-series test data, it is possible to obtain information on whether or not to visit a medical institution when a specific test result (specific information) is obtained. Therefore, based on the time-series test data, it is possible to guide health information that can grasp one's own health condition before it deteriorates as one goes to the hospital. For example, in FIG. 2A, in the case of the test data at time T1, it can be inferred that the medical institution is visited at time Tc when time + ΔT has elapsed. That is, if the examination data, medical institution information (clinical name, clinical department, date and time information) and the like are accumulated in the DB unit 8, the period until receiving medical treatment at the medical institution can be estimated.

Fig. 2 (b) shows a case in which the patient has already visited the hospital and has deteriorated during the hospital visit due to factors other than treatment. The graph shown in FIG. 2B is an example in which a person who goes to the hospital due to illness receives treatment at a clinic when specific information appears at times Tc1 and Tc2. The time-series inspection data as shown in FIG. 2B can be sufficiently utilized to learn such a situation. This example is useful for a guide to the effect that "people with this number usually cannot be treated on their own." It is effective as information that can prevent further deterioration.

Figure 2 (c) shows a case where it is not necessary to go to the clinic. In this case, the test data D is lower than the predetermined value (indicated by the broken line in the graph) and there is no need to go to the clinic. In this case, in the database of the DB unit 8 shown in FIG. 2, the column of the visit date and time is blank.

In the DB section 8, information on the clinics and clinical departments visited, the names of diseases, information on owned equipment, etc. are organized and recorded. For this reason, it is possible to recommend the optimal facility even for patients who do not even think about the facility. This database may hold the relationship between the type of acquired information (toilet bowl occult blood test information), the clinic, and the owned equipment Mod, and the patient-specific time-series data may be managed in a separate database. Further, by searching a plurality of DBs and organizing the search results, information corresponding to the database recorded in the DB unit 8 may be obtained.

FIG. 2 is a graph showing the time-series information for each patient recorded in the DB unit 8, where the horizontal axis is time and the vertical axis is the numerical value of the acquired information. Therefore, the information is two-dimensionally visual. Since it is a two-dimensional diagram, the following two things can be said. First, since it is a diagram, it can be handled in the same way as image judgment, and a general-purpose and easy-to-build AI chip or system such as an inference model for image recognition can be easily diverted, and inference can be easily realized. .. In addition, since the horizontal axis is time, it is possible to effectively use the time-varying information of physical information, and it is possible to easily make a prediction. In addition, it can include information on the characteristics of time changes peculiar to the living body, such as fluctuations and frequency.

For example, it is easy for the user to consider information such as when sleeping, when waking up, morning and day and night, before meals, after meals, and before and after bathing. In addition, there is a study that it is more relaxed and healthier to have appropriate fluctuations in heart rate and breathing.

Further, in FIG. 2, a certain period within the period for acquiring the historical data corresponds to a specific period, and the inspection data during this specific period is extracted. The extracted time-series inspection data is input to the inference engine 7, the inference engine 7 outputs advice by inference, and the advice is provided to the user. In addition, the specific time width may be suitable for giving some advice at a time corresponding to the future after the end of the time width, and a plurality of information retroactive to the time of this advice is acquired. It should be as wide as possible. Further, the specific time width does not have to be strict as determined by the standard, and it is sufficient if a sufficient amount of data can be obtained. Since the time interval of each data is also important information, it is better that the data is obtained with a regular time width and is not discrete. However, even if the measurement time points of the data are discrete within the time width, it is effective as long as the time width is such that the data can be supplemented by interpolation to obtain meaningful data. It may be determined according to some health or medical information.

It is also possible to switch the accuracy of prediction etc. by the inference model by appropriately defining the width of the horizontal axis of the graph. If the time width is about one year, it is possible to predict orders for several months, and if the time width is about one week, it is suitable for forecasting orders for several days, and the appropriate width is changed according to the characteristics of the disease. Can be done. For example, when it is necessary to distinguish between a tumor that progresses gradually and an infectious disease such as influenza that cures rapidly and worsens, the appropriate time range differs. That is, the information transmission device according to the present embodiment extracts the change pattern of the user's inspection data within a predetermined time width, and determines the transmission information to the user according to the inference model learned together with the time information. It has a decision part.

Next, with reference to FIG. 3, a case where the user measures with the first device 2a and the second device 2b and acquires a plurality of inspection data will be described. FIG. 3 graphically displays the acquired inspection data. In FIG. 3, the horizontal axis T represents the measurement timing, and the vertical axis represents the value of inspection data (DA for the first device, DB for the second device).

FIG. 3 illustrates that even if the inspection data has the same items for the same subject, the same value may not be output depending on the device due to the difference in the machine and the installation / measurement environment. For example, if the threshold value for determining whether a person is healthy or not requires a detailed examination is indicated as DAT in the first device and DBR in the second device, according to the results shown in this figure, the first device may be determined by time. The measured value is above the threshold value and below the threshold value in the second device. That is, it is not judged only by the fact that the threshold value is exceeded at the timing of T0, but an accurate judgment is made by searching the inspection data as T1, T2 ... T4, going back from the timing T0 that exceeds the threshold value. Can be done.

However, since there is only one inspection history using the first device 2a, it is preferable to refer to the data accumulated in other devices when making a more accurate judgment. Therefore, by referring to the data of the device (second device 2b) capable of measuring the same measurement items as the first device 2a, as shown in the graph 33, the amount of information is increased so that the change with time can be correctly determined. I have to. For example, body fluids such as excrement and blood change depending on the physical condition, meal time, drinking and taking medicine, before and after bathing, before and after sleeping, and the excess and deficiency (living conditions). It is better not to judge. In particular, in the case of equipment that is monitored on a daily basis, various error factors are faced, so it is desirable to take measures as described here.

Before and after a specific treatment such as surgery, vital data may change significantly due to a large physical burden. In addition, the test data of such a patient may change significantly at a certain time, and when such a situation is determined and detected, the correction as shown in the present embodiment may be performed. It is better not to do it. Furthermore, by managing and judging the inspection information obtained in daily life from this point onward separately, the value as data for follow-up observation is increased.

In order to reduce the error due to the living situation explained above, the judgment is made by increasing the data measured in various situations, and the judgment is not made only in the special situation for that time (before and after the treatment such as the above-mentioned surgery). Data is not limited to this). That is, even if the inspection data group of the subject is acquired by the first device 2a, it is better to be able to confirm the status of changes over time, but in the example shown in FIG. 3, the amount of data is small. It's too much. Therefore, by acquiring the time-series second inspection data group of the subject by a second device (which may further have a third and a fourth) capable of interpolating the first inspection data group. , By supplementing the data, it is possible to make highly reliable judgments. This second device has a second inspection data acquisition unit, and records the history of the data retroactively by a predetermined time (up to time T4 in the example shown in FIG. 3). When the transmission information determination unit determines the transmission information to be provided to the target person by using the first inspection data group and the second inspection data group, the reliability is enhanced. By supplementing the inspection timing or inspection items with each other, the first inspection data group and the second inspection data group can provide information with less influence of the special living conditions as described above. Here, a specific time is traced back to T4, but in the above-mentioned follow-up, some measures may be taken such as tracing back to the postoperative period.

A little more concretely, an example considering prevention of digestive system diseases will be explained. A technique called stool test is known to prevent this disease. In the example of FIG. 3, it is assumed that this stool test is performed by the toilet performed at the time of excretion without the subject (user) being aware of it. The graph 31 is created based on the inspection data acquired in the toilet at home by the first device 2a, for example, and the graph 32 is created based on the inspection data acquired in the toilet at the workplace by the second device 2b, for example. Is.

In this way, if the data of two toilets can be used in a private and official environment as a pattern of life, many inspection opportunities will be created and a wealth of information can be obtained. In the workplace, it is possible to correct cases where high data is obtained due to tension, or there are more opportunities to judge trends that were not noticed outside the workplace.

Even if the same user is inspected with the same inspection items, measurement errors and accuracy will differ due to differences in the equipment used and management methods. Therefore, it is not always correct to describe the two inspection data on the same graph as shown in the graph 33. For example, the first device 2a used at home may be a simple sensor, and the second device 2b used at work may be a high-performance sensor with an emphasis on employee health management. Under these circumstances, the two inspection data may not be simply comparable because they have similar numerical values but different measurement methods and sensors. From another point of view, if the first and second devices are placed in the toilets of stations and public facilities, it is difficult to control the temperature and humidity in this case, and there are many users, so the temperature, etc. It is difficult to perform numerical analysis on the same scale because error factors such as environmental errors, deterioration of parts, and dirt are likely to occur. However, in order to collect health management information of a specific person, it is better to increase the number of information and increase the value of the information by comprehensively using the data acquired by various devices. In addition, if the toilet equipment is updated frequently, the data cannot be handled equally before and after the equipment change. It is better to handle the data in consideration of this point as well.

Therefore, in the present embodiment, the inspection data acquired in the first and second devices 2a and 2b are not simply arranged as shown in the graph 33, but the same devices have the same numerical values as shown in the graph 34. Considering that it has characteristics, correction calculation is performed on the two inspection data. In FIG. 3, as shown in Graph 33, the tendency that the output from the first device tends to be low is increased by a predetermined number or emphasized by applying a gain, and these processes are performed. It is shown as graph 34. ) That is, the inspection data is used as a database in which the device ID and the test result are recorded together, and the data of the same device is uniformly shift-corrected (addition / subtraction) and / or gain-corrected (multiplication / division). This correction calculation makes it possible to compare the vertical movement patterns of data transitions obtained by a plurality of devices, which change according to the acquisition time, in the same manner as other data.

However, if the increase / decrease pattern of the test data is reversed depending on the degree of health, or if the pattern is different even after correction, record the measured value so that it can be analyzed so that it will not be erroneously compared. Further, information such as the type of the device used for the measurement, its model number, and the sensor value may be used. With this information, it is possible to compare numerical change patterns from other devices that acquire similar biometric information. In general, biological information rarely changes in minute increments, so the error in the clock information (which determines the accuracy of the horizontal axis of the graph) of each device should be as accurate as minutes.

Based on the above-mentioned way of thinking, abundant health-related numerical values can be obtained within an appropriate time range for changes in physical condition and the onset or worsening of illness. Based on this information, the information providing department 1c advises the user as soon as possible. It becomes possible to do. When making inferences using discrete time-series data of a specific device, enrich the data by converting it into data while correcting the information that supplements the discrete time intervals, and make inferences using this data. This enables highly accurate advice.

Next, an example of the operation of transmitting the inspection result in the information transmission system will be described using the flowchart shown in FIG. This flow is mainly executed by the CPU in the control unit 1 controlling the entire information transmission system according to the program stored in the memory.

In the example shown in FIG. 4, the first device 2a is assumed to be a wearable device, and the second device 2b is assumed to be a dedicated device. Both are devices for acquiring health-related information of the target person (specific user). Wearable devices have low measurement accuracy, but since they are worn by users on a daily basis, they can make frequent measurements and collect a large amount of information. On the other hand, the dedicated device has high measurement accuracy, but cannot perform frequent measurement as compared with the wearable device. By acquiring inspection data using a plurality of devices, the user can compensate for each other's shortcomings, improve the accuracy of the data, and increase the number of data. That is, it is possible to reinforce the advantages of each of the devices having different characteristics and to manage the health of each individual in close contact with daily life.

When the flow of transmitting the test result shown in FIG. 4 starts, the mobile terminal acquires the health-related numerical value (S101). Here, the wearable type first device 2a routinely acquires a numerical value (test data) for health management such as blood pressure. When the inspection data acquired by the first device 2a is abnormal, the data is transmitted to the control unit 1 and the process proceeds to step S103. If there is no problem with the loss of energy and time spent on communication and inference, the inspection data acquired in step S101 may be transmitted to the control unit 1. As will be described later, since the inspection data by the first device 2a is used when the history data is combined in step S107, the first device 2a transmits the inspection data to the control unit 1 at predetermined time intervals.

When the terminal acquires the health-related numerical value, the test result of the same person is searched (S103). Here, the ID determination unit 1b of the control unit 1 searches the DB unit 8 for the inspection result of the same person as the user measured in step S101. For terminals owned by individuals, dedicated sensors may be simple, or errors may easily be included due to restrictions on individual carrying and handling. Therefore, we would like to check and verify whether the health-related numerical values acquired by the terminal include errors, etc., using the results of the health-related numerical values acquired by the dedicated device. As described above, if the inspection data is not abnormal and the inspection data is not transmitted, this step may be skipped.

Next, acquire the corresponding data using a dedicated device (S105). Here, the second device 2b of the dedicated device type acquires the user's health management numerical value (examination data). When the second device 2b acquires the inspection data, the second device 2b transmits the inspection data to the control unit 1. Dedicated equipment is calibrated equipment handled by a specific specialized institution or specialist, and is often installed and used in a stable environment, and it is possible to obtain highly reliable results depending on the results of personal terminals. Be expected. Therefore, based on the data acquired by the dedicated device, for example, the result acquired in step S101 can be corrected and handled.

Subsequently, the history of the inspection data acquired in steps S101 and S105 is adjusted, and inference is performed using the adjusted history data (S107). As described with reference to FIG. 3, even if the first device 2a and the second device 2b are the same, the values of the inspection data may deviate due to an error of the device or the like. However, if the same person is used, the tendency of the inspection data acquired in time series will be the same even if the equipment is different. Therefore, the control unit 1 performs a correction operation on the inspection data acquired by the first device 2a and the second device 2b, and generates time-series inspection data (history data) acquired by the same device. The inference engine 7 inputs this historical data and makes an inference to output advice on changes in physical condition and onset or worsening of illness.

After making an inference, the inference result is displayed next (S109). Here, the control unit 1 transmits advice based on the inference result to the terminal 4 owned by the user, and the terminal 4 displays the advice. At this time, the advice can be changed in various ways depending on what kind of annotation information the inference model has to learn. For example, the annotation information determines the type of clinical department and prescription drug information of the hospital to be visited. If included, these results can be presented as inference output. In addition to these inference results, a database or the like may be searched on the Internet from the inference results, and additional precautions or the like may be presented based on the search results. It is also possible to determine what kind of inspection data tends to be under what circumstances from the history of the user's inspection data and the history of the mobile terminal in possession, and this determination is by reasoning. You can do it, or you can do it on a rule basis. In other words, in the case of a person with signs of hypertension, it is possible to analyze and present when the tendency to hypertension occurs by judging the situation of the user who outputs a high value in the historical data. Depending on the situation, it can be shown that stress in the workplace has an effect.

In addition, even if the user simply presents the date and time when the high number appears, the user sees it and what kind of tendency his / her health depends on what kind of environment, season, what time of day, etc. It can be inferred whether it will be. According to this judgment, the user can take actions such as resting, taking medicine, and going to the hospital, so that it is possible to take early measures before becoming aware of it, and it is possible to increase the number of people who continue to live a healthy life. In addition, if the user knows an environment in which the numerical value of the inspection data tends to deteriorate, feedback control such as causing a mobile device or an inspection device to perform control to acquire various data when the user comes under the environment. By doing this, it is possible to increase the amount of clues to identify the cause.

In the flow of transmitting the test result in the present embodiment, it is considered that the error in the dedicated device and the health diagnosis result can be almost ignored, and the result acquired in the wearable device (S1) having many variation factors is corrected. Therefore, it is possible to determine whether the health data obtained on a daily basis is different or the same as that at the time of the inspection with the dedicated device.

As a result, when measured with a dedicated device, the numerical value happens to be good, while when the numerical value measured in the daily life scene gives a worse result, it becomes possible to accurately judge. In this case, for example, it is possible to give health advice such as caution after eating or caution when waking up. This health advice may be presented to a device such as a user's terminal 4 that is easily accessible by the user. Recently, it is possible to display on the user's TV, and technologies for transmitting information to specific individuals other than mobile terminals such as AI speakers and health care wash basins have been widely provided, so we will use these. You may.

The explanation of the flow in FIG. 4 focused on a device provided with a dedicated sensor such as a sphygmomanometer, but it is not limited to blood pressure and can be applied to information such as pulse and heart rate. These can be complemented by general-purpose or widely known technologies that can be mounted on mobile terminals other than special applications such as image sensors and acceleration sensors. It is not necessary for the two devices to measure the same item (for example, blood pressure and blood pressure), one may be blood pressure and the other may be pulse. Regardless of whether there is a correlation or no correlation, such as at what heart rate and what kind of blood pressure, if the heart rate continues to be high, the risk of developing heart disease increases in hypertensive patients, so it is comprehensive. It is often useful for making a good judgment. Since various devices can acquire various health-related data, it is possible to change the monitoring device and the data requiring attention depending on the user, and it is possible to give advice such as which hospital to go to depending on the symptomatology.

In this way, the information providing unit 1c may customize the information for each user. Specifically, it is conceivable to provide information on an appropriate clinic in the neighborhood of the user's place of residence. In addition to this, the medical staff of the family facility may decide which numerical value to monitor, and the system of this facility manages the data. If the medical facility handles a sufficient number of cases, it is possible to diagnose that this person and this person have similar health information, and that each has a similar tendency to illness. For this purpose, the DB unit 8 may be provided on the server in the hospital to store data. In this case, it is possible to make inferences, considerations, and advice that automatically reflect the environment and eating habits peculiar to the area.

Next, another example of the operation of transmitting the inspection result in the information transmission system will be described using the flowchart shown in FIG. This flow is also mainly executed by the CPU in the control unit 1 controlling the entire information transmission system according to the program stored in the memory. Further, the flow shown in FIG. 5 shows a case where the search in the DB unit 8 shown in FIG. 1 and the inference by the inference engine 7 are individually used. You may want to use one of the functions, or you may use both of them in layers, but here is the simplest example.

In explaining the flow of FIG. 5, as the first device 2a, the second device 2b, and the third device 3, the toilet bowl has an image sensor, a magnified image judgment device such as a microscope, and a sensor that detects special light reflection. The description will be made on the assumption that an array of crystalline nanowires, an olfactory sensor that applies changes in electrical characteristics such as a molecular film, a gas component sensor, and the like are arranged so that the characteristics of the user's excrement can be confirmed.

When the flow of inspection result transmission shown in FIG. 5 starts, first, each ID is determined based on the sensor output result (S1). Here, the control unit 1 may acquire the output of the first device 2a or the like through the communication control unit 1a, or the control unit 1 may receive the data transmitted by the first device 2a or the like in the communication control unit 1a. Further, it is assumed that the control unit 1 collects the data recorded by the first device 2a or the like through the communication control unit 1a at a specific timing. At this time, the inspection result is determined based on the sensor output for each ID attached to the sensor output result. The sensors are a color sensor, a shape sensor, a hardness sensor, an olfactory sensor (including reaction judgment of nematodes and animals), a gas component sensor, and a color change detection sensor when a specific reagent is added, based on the output of the image sensor. Then, the shape may be determined by the magnified observation image.

When confirming the characteristics of the user's excrement described above, for example, stool with occult blood can be determined by a color sensor. Further, the amount, shape, hardness, etc. of excretion may be determined by an image sensor / color sensor, or a method of measuring the color distribution by performing special dyeing may be used. Alternatively, the composition may be detected in a magnified image of the object, or the result of culturing for a specific time may be determined. For example, when the amount of blood mixed in the stool increases, the red color of red blood cells becomes conspicuous, but if this is quantified, the difference from the healthy case can be seen. These are detected in step S1.

In step S1, when the control unit 1 determines the sensor output result by a determination using a specific program or the like, it then determines whether or not specific information can be obtained (S3). Here, based on the determination result in step S1, it is determined whether or not specific information related to the disease, for example, a feature such as a numerical value different from the healthy state is detected.

If the specific information cannot be acquired as a result of the determination in step S3, the process returns to step S1. On the other hand, if the specific information can be acquired as a result of the determination in step S3, it is determined whether or not there is a progress inference model (S5). Here, based on the specific information acquired in step S3, whether or not a database capable of suspecting a specific disease and conducting a detailed examination related to this disease is accumulated, and using this data, infer the future progress. Determines if an inference model capable of is set in the inference engine 7.

If there is no transitional inference model as a result of the determination in step S5, an inference specification is created (S13). If it is determined in step S5 that the database has not been accumulated, the control unit 1 requests the DB unit 8 to construct the database. By constructing a database capable of searching for a specific disease, it is possible to quickly construct a system as the number of users increases, even if it is the first device. Also, if you simply build a system that allows people who are sick or not sick to send that fact according to the test data, it is possible to know whether or not they are likely to get sick. You will be able to judge from. In addition, it is better to be able to make more accurate inferences beyond such simple predictions, so in order to obtain inferences for this purpose, first, inference specifications are created in this step. The detailed operation of creating the inference specification will be described later with reference to FIG. 7.

When the inference specification is created in step S13, the inference model specification is requested to be created (S15). Here, the specifications of the created inference model are transmitted to the learning unit 5 through the learning request unit 6. The learning unit 5 generates an inference model according to the specifications. The control unit 1 receives the generated inference model through the learning request unit 6. When the inference model creation request is made, the process returns to step S1. The detailed operation of creating the inference model will be described later with reference to FIG.

If the result of the determination in step S5 is Yes, that is, if there is a database for search and there is also an inference model, the "history search" method is determined (S7). If the result of the determination in step S5 is Yes, it means that there is a database for searching. In this case, the information on the specific disease determined in step S3 is stored in the DB unit 8. Search from inside. In this step, a method of historical search, i.e., how to search the database for facilities for further testing for a user's specific disease is determined. For example, in the case of a digestive system disease, the excrement system test data is mainly searched. In addition, if it is an emergency disease, a short-term history is sufficient, and if it is a chronic disease, a long-term history is searched. At this time, if the period is too long and there is a large amount of data, the data may be thinned out.

It is difficult to predict the future from the previous pattern on the time axis using only one type of data at a time. Therefore, in step S7, the historical data in the specific time range that goes back to the past on the time axis of the user's health data is searched and used. The results of this search include time information, which makes it possible to predict the future. The specific time width depends on the disease. For example, for the latest future forecast of sudden changes in the medical condition, the latest past change data is important, but for lifestyle-related diseases that gradually worsen, the history over a long span is important. .. Therefore, the time range for history acquisition may be changed depending on the disease to be considered.

After deciding the history search method, next, one history data is matched with the other history data, and inference is performed (S9). In the present embodiment, the user's inspection data is acquired by both the first device 2a and the second device 2b. As described with reference to FIG. 3, the levels of the inspection results of the first device 2a and the second device 2b do not match. Therefore, in step S9, the control unit 1 inputs the inspection data so that one level (the level of the inspection data of the first device 2a) and the other level (the level of the inspection data of the second device 2b) match. On the other hand, the correction calculation is performed. In this case, the control unit 1 may combine one historical data with the other historical data.

When the historical data is combined, in step S9, the inference engine 7 uses this historical data to make an inference to output advice on changes in physical condition, onset or worsening of illness. The detailed operation of "inference by combining historical data" in step S7 will be described later with reference to FIG.

When the historical data is combined and the inference is performed, the inference result is displayed next (S11). Here, the control unit 1 transmits the inference result in step S9 to the user's terminal 4, and causes the display unit of the terminal 4 to display the inference result. This step S11 is a step of providing information on examinations and medical assistance to the user who became the information source acquired in step S1 and related persons thereof, and it is assumed that a display or a warning is issued on the terminal 4. .. When the inference result is displayed, the process returns to step S1.

As described above, in the flow of transmitting the inspection result in the present embodiment, the control unit 1 acquires the sensor detection results from the first device 2a and the second device 2b (S1), and the health state (health state) from these detection results (S1). It is determined whether or not there is specific information related to (disease) (S3). When the specific information is acquired, it is determined whether or not there is a database related to the specific information and whether or not there is an inference model (S5), and if there is a database, this is searched. Then, the correction calculation is performed on the data acquired from the first device 2a so that the levels of the values acquired from the first device 2a and the second device 2b match. When the two historical data are combined by the correction calculation, inference is performed using this historical data (S9), and the inference result is displayed. Therefore, when the user can acquire the inspection data by a plurality of devices, the output levels of the respective devices can be matched, and therefore the inspection data can be enriched. Highly accurate prediction and inference can be performed using abundant inspection data. As a result, the user can perform a health check in daily life and receive advice according to the health condition.

If there is no inference model (S5 → No), an inference specification for performing inference according to specific information is created (S13), and the inference model is sent to the learning unit 5 (through the learning request unit 6). Requesting generation (S15). Therefore, it is possible to sequentially add inference models according to the user's health condition.

In the flowchart shown in FIG. 5, for example, in step S3, when the specific information is not acquired, the user's profile, behavior, lifestyle, etc. may be determined. By acquiring this information, it is possible to provide appropriate information. In addition, as information, information such as age, gender, and pre-existing illness, address, eating habits, and food information are also effective. This information can be obtained by taking a questionnaire on the terminal 4, inputting and acquiring the information when setting up the information judgment device 2, and inputting by the related inspection institution 9 at the time of going to the hospital. These devices and their devices Information existing on the network may be collected and prepared through.

Further, in the flowchart shown in FIG. 5, DB search (S7) and inference (S9) are treated independently as separate processes. However, the present invention is not limited to this, and these may be treated comprehensively. For example, there is a method of searching the DB after making an inference, and using the inference model learned including the information in the DB including the inspection device information at the time of learning, the equipment such as the equipment owned is output. You may make inferences. In this case, it is possible to display such as "This clinic has XX inspection equipment".

Next, the operation of "inference by combining historical data" in step S9 of FIG. 5 will be described using the flowchart shown in FIG. In this flow, as described above, the first device 2a and the second device 2b are used to extract the change pattern of the user's inspection data within a predetermined specific time width so that the output levels of the devices match. The correction calculation is performed to match the two historical data. Based on this combined historical data, infer health advice. This process is performed by the control unit 1 in cooperation with the inference engine 7, the DB unit 8, and the like through the communication control unit 1a.

When the flow shown in FIG. 6 starts, time series data is acquired (S21). Here, the time-series data corresponding to the specific ID recorded in the DB unit 8 is acquired. The time width of the time-series data to be acquired is set to a specific time width, but if the data of the specific time width cannot be acquired, it is set to the time range that can be acquired. This is because if there is no specific time width, the judgment will be based only on the data under a specific situation, and the reliability will be inferior. The specific time span differs between diseases that progress over time, such as colorectal cancer, and diseases that progress in a short period of time, such as influenza. Further, the type of test data depends on the learning of the inference model, but it is desirable that the graph is a specific item used at the time of learning. For example, it is preferable not to infer body weight and blood pressure together. Therefore, it is preferable to make an inference after considering the auxiliary information of the data such as what kind of sensor information of what kind of device.

When the time series data is acquired in step S21, it is next determined whether or not the time series data for a specific time width can be acquired (S23). For example, if the status of occult blood is detected, it is determined whether or not occult blood was obtained within a range of several months. That is, the specific time span depends on the disease involved.

As a result of the determination in step S23, if the data of the specific time width has not been acquired, no inference is performed (S35). Inference is possible depending on the expected reliability without information of a specific time width, but it can be difficult. Therefore, if it is determined in step S23 that the data of the specific time width has not been acquired, no inference is made. However, there are cases where a clearly dangerous situation can be detected, in which case emergency information may be output before inference.

In other words, if the acquired data is a numerical value that has a significant problem, there is no time to make a long-term forecast by reasoning, so a warning is displayed when the change is significant. With this response, it is possible to prevent the system from being unable to respond in an emergency without making inferences in step S35, and to make a highly reliable system in which information is output after sufficient data has been collected. That is, in the present embodiment, in the case of a numerical change that is contained in a specific change, the change pattern of the user's inspection data is cut out within a predetermined time width, and inference is performed according to the inference model learned together with the time information. Processing step S35 ends this flow and returns to the original flow.

As a result of the determination in step S23, when the data for a specific time width is acquired, then the data acquisition device information and the acquisition time information are associated with each data (S25). Here, the control unit 1 associates the data acquired in step S21 with information on which device among the first device 2a, the second device 2b, etc., and the acquisition time, and records the data in the DB unit 8. do. By associating these information with the acquired data, it is possible to position each data in the graphs shown in FIGS. 2 and 3. If the number of target devices increases, it becomes possible to reduce the influence of errors of individual devices while increasing the number of data when evaluating in chronological order. Also, if you know which data (including time information) comes from which device, even if a device becomes unreliable at a specific timing for some reason, only the data before that timing will be displayed. It can be adopted and devised to make a judgment based on the adopted data. In addition, if it is possible to determine which device the user frequently uses, it is possible to use it to reflect the inspection results using other devices as necessary while making a judgment solely on that device. It becomes.

Subsequently, increase / decrease the time series data for each device (S27). As described with reference to FIG. 3, when the user's time-series inspection data is acquired by a plurality of devices, there are errors and characteristic differences between the individual devices. Therefore, the plurality of time-series inspections are performed. The data cannot be plotted on the same graph (see Graph 33 in FIG. 3). However, since the data is time-series inspection data of the same person, the tendency of the change pattern of the data is the same. Therefore, by performing a correction calculation on a plurality of time-series inspection data, it is possible to plot the plurality of time-series inspection data on the same graph. As the correction operation, each data may be added / subtracted or multiplied / divided based on the difference between the average values of the two time-series data. As a result of this correction calculation, time-series data corrected for each device can be obtained. Further, when certain specific data is important and other specific data is not important, the reflection of the information may be different by changing the weighting of the data.

When the time series data is increased or decreased in step S27, the data are collectively input to the inference model (S29). Since the time-series data was generated for each device in step S27, the control unit 1 inputs the time-series data to the inference engine 7. In this case, the inference will include the error of each device, and it is judged that the inference is not consistent and the reliability is low in the specific inference model that learned the increase / decrease information in the same time range as the specific time as teacher data. May be done. Therefore, the reliability of the inference is calculated while making corrections for each device, and the one with high reliability is used as the inference result (S31). Here, the constants of the specific four arithmetic operations are changed little by little for the time series data for each device. By this processing, the reliability is increased in the situation where the error is corrected, so that correct inference is possible.

The process in step S31 may be performed after, for example, the control unit 1 determines whether the device is a device that detects similar biological information based on the device ID or the like of each information. As a result, the relationship of increase / decrease in data due to changes in health is guaranteed, so it is only necessary to assume reduction of sensitivity and environmental error. However, if the inspection items have the same increase / decrease relationship of data due to changes in health, they may be treated uniformly. This is because if the information is treated uniformly, it can be more effective information if the number of data is a medical condition that is effective as a temporal density or a temporal range. In this case, it is sufficient for the data to have information on what kind of test item it is, and whether or not it can be treated equally, or how much data should be handled collectively assuming a specific disease, etc. The control unit 1 may provide a step for determining the determination.

That is, the information output by the device is not only the judgment result data, but also the inspection timing (date and time) information, the information corresponding to the target individual, the inspection content information, and the device-specific information in a specific format. If you have some of these information, such as information on the type of device, you can use this information to correct or select the data. Moreover, not only correction but also weighting is possible. Data acquired by an unreliable device may be lightly weighted so that it is not treated in the same way as other devices. Furthermore, if there are many people who use the same multiple devices, for example, if all the time series data is color-coded and arranged on a graph, the information of multiple devices is mixed instead of the information of a single device. People with similar health tend to have the same tendency.

That is, the transition of information so that the information from which device can be identified on a common time axis by color coding or the like according to the information from the specified device acquired at a specific time of a specific person. By creating a graph showing the above, it is possible to grasp the tendency of the person's health condition and convey it. The judgment may be made by adding various corrections and the like already described on the graph. Identification by color coding is a device that makes it easy for humans to see, but in addition, the shape of the data points drawn at the plot points may be changed so that the device can be identified. , The same effect can be obtained by making additional information displayable or readable in the data at that point.

When inference is performed in step S31, the inference result is acquired (S33). Here, the control unit 1 sets the inference output having the highest reliability value as the inference result when the inference is performed in step S31. It is possible to provide health advice by effectively using the information acquired in various life scenes.

In this way, in the flow of inference by combining the historical data shown in FIG. 6, time-series data is acquired (S21), and when time-series data for a specific time width can be acquired (S23Yes), each device is used. A correction calculation is performed on the time-series data so that the level is the same, and inference is performed using the history data to which the correction calculation has been performed (S29). Then, the reliability of the inference is determined while making corrections for each device, and the highly reliable one is used as the inference result (S31). Therefore, it is possible to obtain a highly accurate inference result by using the inspection data of a plurality of devices. Further, since the inference is performed using the time series data having a specific time width, it is possible to perform the inference with high accuracy.

Further, in the present embodiment, in the case of a numerical change that is contained in a specific change, the change pattern of the test data of the subject is cut out within a predetermined time width, and inference is performed according to the inference model learned together with the time information. ing. That is, when the specific criteria are not satisfied (when the data of the specific time width cannot be acquired (S23No)), the change pattern of the test data of the subject is cut out with the predetermined time width and learned together with the time information. I try not to make inferences according to the inference model.

Next, the operation of "creating an inference specification" in step S13 of FIG. 5 will be described using the flowchart shown in FIG. When it is determined that the specific information has been acquired in step S3, the subroutine for creating the inference specification is the learning unit 5 when the inference model for inference based on the specific information is not set in the inference engine 7. Create a specification for requesting the generation of an inference model.

When the flow for creating inference specifications starts, first, the related disease is determined based on the specific information (S41). Based on the specific information determined in step S3 (FIG. 5), the related disease is determined. For example, since the biometric information related to each disease is determined based on the user's urine test result and stool test result, a table or the like showing the relationship between the test item and the related disease is recorded in the DB section 8 or the like. Then, the related disease can be determined based on this recorded table.

Once the related disease is known, the patient with the related disease is determined (S43). The DB section 8 records and organizes the daily health (biological information / examination data) of a large number of patients. Therefore, the control unit 1 determines (searches) a patient other than the user suffering from the related disease determined in step S41.

Next, it is determined whether or not there is a history of patient health information (S45). Here, the control unit 1 determines whether or not the history of the patient's health information determined in step S43 is recorded in the DB unit 8. As a result of this determination, if sufficient data (history of health information) is not accumulated, no inference or inference request is made, and no advice information is given (S51). End this flow and return to the original flow.

On the other hand, if there is a history of health information as a result of the determination in step S45, the history data is input in the time width for each related disease and the disease is output (S47). Here, the control unit 1 searches the health information of the patient who has already been diagnosed with the disease, which is recorded in the DB unit 8, and determines whether or not there is a period or amount of the health information that can be used as the teacher data. do. When this determination is satisfied, historical data is extracted from the time-series data of this patient with a time width necessary for determining the determined related disease, and this time-series data is used as teacher data. This time-series data is information on patients who have already been diagnosed with a disease, and there is biometric data measured by an instructor with specialized knowledge using a relatively accurate measuring device. Therefore, the control unit 1 corrects the substitute data obtained by the household device or the mobile terminal obtained in daily life based on this biometric data (a series of data), and uses this time-series data as the teacher data. do. The control unit 1 inputs the teacher data to the inference engine 7 and acquires the disease information as an output.

The above-mentioned biometric data (data sequence) has a different pattern if there is no information such as which time point of the disease corresponds to. If the reasoning is for non-illness, the pattern may be traced back to this point based on the timing of the first visit or the timing of the first diagnosis of the disease. That is, the data of the patient who went to the hospital after the lapse of a predetermined period as shown in FIG. 2 (a) and the data of the person who did not go to the hospital as shown in FIG. 2 (c). A large number of data are collected, learning is performed so that these differences can be determined, and an inference model is created. By inferring using this inference model, it is possible to determine which pattern the obtained data resembles, and whether or not the patient is likely to get sick. In the data obtained after the start of the hospital visit as shown in FIG. 2 (b), the pattern may change depending on the treatment. However, there is also a need to infer the future, including the effects of the drug, so there is also a method of creating and learning teacher data that shows the time of the start of hospital visits.

When inferring using the inference model created as teacher data while correcting the time series data group so that the reliability of the inference is improved, the inference result with high reliability was input while correcting the time series data group. By adopting the method, highly accurate prediction can be made. In addition, if there is other reference information that changes with time even if it is not directly related to the disease, it will be used as teacher data as separate data.

In step S47, when the specification for creating the inference model is created, the specification is set so that the advice for the disease and the current non-illness level are also output (S49). Here, the inference engine 7 creates an inference model specification capable of outputting advice for diseases. At that time, the advice such as the stage of the disease may be output, or the user may be able to output information such as at what timing before going to the hospital. When the specifications for creating the inference model are created in steps S47 and S49, this flow ends and returns to the original flow.

Note that the learning for which specifications are created in the flow shown in FIG. 7 can be applied to fields such as the relationship between labor pain and the time of childbirth. By learning the frequency of labor pains retroactively based on childbirth, it is possible to generate an inference model that can give advice on how long later to go to the hospital or call a midwife.

Next, using the flowchart shown in FIG. 8, the learning operation performed by the learning unit 5 when the “inference model creation request” is made in step S15 of FIG. 5 will be described. Here, the history data of the related disease patient determined in FIG. 7 is used as the teacher data to generate an inference model that can obtain the set output. This subroutine for creating an inference model is mainly executed in the input / output modeling unit 5a in the learning unit 5.

Teacher data is generally created by adding specific annotations to specific data. In this flow, data (acquired information) acquired at multiple timings related to the same subject is annotated with that person's health-related information (test results, date and time of hospital visit, advice, etc.). Two teacher data. This teacher data was prepared for multiple subjects, and it was possible to infer what kind of health information the transition pattern of time-series data corresponds to. The original data for making this teacher data may be recorded in a file format, or necessary metadata groups may be associated and recorded. This metadata may be for annotation. In addition, when creating an inference model, it is selected as needed, so as a basis for the inference model, information such as an ID that identifies the created inference model can be recorded as metadata in the adopted file. It may be. By these treatments, it becomes possible to prevent the AI from becoming a black box.

When the inference model creation flow starts operating, first set the input / output (S61). Here, the learning unit 5 sets the input / output of the inference model based on the specifications transmitted from the control unit 1 through the learning request unit 6. That is, what (what kind of information) is input to the inference model, what (what kind of information) is inferred and output, and the like are set. Further, the number of intermediate layers of the neural network is set, and the weighting and the like in each intermediate layer are set as initial values. In this step, the so-called "requirement specifications" of the inference model to be created are set.

Next, input the teacher data and create a model (S63). Since the teacher data is created by the control unit 1 from the data recorded in the DB unit 8 and transmitted to the learning unit 5 (see S47 and S49 in FIG. 7), this teacher data is input / output modeling unit 5a. Sequentially input to the input section of. Further, since the teacher data is a set of input and output, an inference model is created by determining the weighting of each intermediate layer of the neural network so that the output corresponds to the input.

After inputting all the teacher data in step S63, it is next determined whether or not the model could be created with high reliability (S65). Here, it is determined whether or not the value indicating the reliability of the inference model generated in step S63 is higher than the predetermined value.

If a highly reliable inference model cannot be created as a result of the determination in step S65, re-learning is performed (S69). Since the reliability is low, the population of teacher data is changed, the process returns to step S63, and the inference model is recreated. If the reliability does not reach a predetermined value even after re-learning a predetermined number of times, the generation of the inference model is terminated and a notification to that effect is transmitted to the control unit 1. When re-learning, a metadata group or a file may be attached to record that the data group or file was not used as teacher data. It is possible to prevent poor quality data groups and files from being used for learning and becoming unsuccessful.

On the other hand, if a highly reliable model can be generated as a result of the determination in step S65, that model is used as an inference model (S67). The learning unit 5 transmits the inference model generated here to the control unit 1. When the control unit 1 acquires the specific information based on the inference specification created in step S13, the control unit 1 can set the received inference model in the inference engine 7 and perform inference. When the inference model is created, it ends this flow and returns to the original flow.

In this way, in the flow of creating an inference model, the learning unit 5 to which the teacher data is given performs learning and creates an inference model. In this flow, re-learning is repeated by selecting the teacher data or selecting the data included in the teacher data until highly reliable inference becomes possible (S65Yes) (see S69). .. Through this learning, we create an inference model that matches the characteristics of changes in the health numerical change pattern in the pre-illness stage with the pattern of getting sick.

In the process of selecting teacher data at the time of re-learning, if there is data unrelated to the specific disease, the control unit 1 may exclude the individual data or the data group, assuming that the reliability does not increase. If the control unit 1 tracks the specific biometric data excluded in such a process, the device that output the data, the environment in which the data was obtained, etc., the data or device that is not suitable for use as inference The environment can be specified, and the specified data, devices, etc. may be recorded and excluded when creating an inference model from the next time onward.

With this unusable data information, it is possible to narrow down the biometric information acquisition devices, data, and environment that should be used when there are signs of a specific disease. For example, in order to detect signs of colorectal cancer, stool test history is important, but treating heart rate, which is information on different organs, in the same way should be avoided due to the increase in data volume and complexity of calculation. Is preferable. However, if it is the same type of biological information, it is often preferable to increase the number of data or to track it even if the difference in equipment and the difference in measurement method are taken into consideration.

On the other hand, it is not very effective to combine the unchanged stool results measured with another device many years ago and the recent sudden changes in stool results into teacher data and learn using this teacher data. It will not be a learning process. In addition to the data of the stool viewing results performed using the recent specific device, the data obtained within the unit time (for example, see S3 in FIG. 4) among the data of the stool viewing results when the device is not used. If the number is insufficient, or if the data that has been thinned out is supplemented because it is inaccurate and cannot be used, the amount of data information of the viewing result will increase, so there is a high possibility that it will be effective teacher data. That is, if the amount of information input to the inference model is appropriate, the reliability of the inference result is also high.

Next, a modified example of the operation of inference (see FIG. 6) will be described with reference to FIGS. 9 and 10. In the example of FIG. 3 described above, it is assumed that the amount of data is insufficient with only one device and appropriate inference cannot be performed. By supplementing with the information of a plurality of devices, the inference result can be obtained. I'm trying to be correct. However, as shown in FIG. 9, when each device has a sufficient data history regarding a specific person, as shown in Graph 34 of FIG. 3, each device does not supplement the amount of information within a unit time. You may perform inference with the inference model for the device and make a comprehensive judgment of the result. In addition, not only the inspection timing but also the information from the device that outputs different information is added, and the judgment can be made more accurately by making a judgment based on this added information. By supplementing the information itself, the reliability of the inference result may be improved.

When inserting data groups, files, etc. into the inference model, the metadata indicating what kind of inference model the data group or file was created for inference is associated with this metadata. , The optimum inference model may be specified. For example, as health-related information to be inferred becomes specialized, an inference model for colorectal cancer and an inference model for hemorrhoids may be prepared separately. With such a device, it is possible to avoid wasting the output of information such as hemorrhoids to users who are exclusively concerned about colorectal cancer. In addition, if the data group or file used for inference is associated with the information obtained by converting the inference result into metadata, it is effective when searching for what kind of case becomes what kind of data group or file. Information.

At this time, there is a difference between the device that measured the teacher data used in learning and the device that obtained the data to be input to the inference model. For example, the graph 91 shown in FIG. 9 shows the change of the time-series inspection data acquired by the first device 2a, and the graph 92 shows the change of the time-series inspection data acquired by the second device 2b.

As described above, since there is a difference depending on the device, the correction calculation is performed on each inspection data while changing the correction value little by little by the correction inputs 91a and 92a, respectively. Then, the corrected inspection data of the first device 2a is input to the inference engine 7a for the first device, and the corrected inspection data of the second device 2b is input to the inference engine 7b for the second device. In this way, each corrected data is input to the corresponding inference model.

If the correction value is changed little by little, the reliability of the inference output also changes little by little, so the inference result when the reliability of the inference output becomes appropriate is adopted for each device. When the inference result is determined for each device, the final output is the result that comprehensively reflects the individual inference results. As a comprehensive judgment, the inference result with higher reliability may be selected, and if there is no big difference in reliability, it may be an intermediate judgment between the two results, or a judgment including both results. good. As described above, in this modification, the result of inference using a plurality of time series data corresponding to a plurality of inference models is comprehensively judged and used as the inference result. Therefore, a method of displaying advice with high accuracy. Can be provided.

Next, using the flowchart shown in FIG. 10, the operation of the modified example of inference will be described together with the historical data. Similar to the case of FIG. 6, this process is performed by the control unit 1 through the communication control unit 1a in cooperation with the inference engine 7, the DB unit 8, and the like.

When the operation of the flowchart shown in FIG. 10 starts, first, inference is made while correcting the numerical value of the history of the first device (S71). Here, as described using the graph 91 of FIG. 9 and the correction input 91a, the control unit 1 corrects the time-series inspection data acquired by the first device 2a by addition / subtraction, multiplication / division, or the like. To give. The control unit 1 inputs the corrected time-series inspection data to the inference engine 7a and causes the inference to be performed.

Subsequently, the result at the correction value that makes the reliability appropriate is adopted (S73). As described above, the control unit 1 calculates the reliability of the inference while gradually changing the correction value of the correction operation. In this step, the control unit 1 adopts the inference result when the reliability is the highest as the inference result when the historical data of the first device is used.

Next, infer while correcting the numerical value in the history of the second device (S75). Here, as described using the graph 92 of FIG. 9 and the correction input 92a, the control unit 1 corrects the time-series inspection data acquired by the second device 2b by addition / subtraction, multiplication / division, or the like. To give. The control unit 1 inputs the corrected time-series inspection data to the inference engine 7b to perform inference.

Subsequently, the result at the correction value that makes the reliability appropriate is adopted (S77). As described above, the control unit 1 calculates the reliability of the inference while gradually changing the correction value of the correction operation. In this step, the control unit 1 adopts the inference result when the reliability is the highest as the inference result when the historical data of the second device is used.

In steps S71 to S77, the inference results are determined for each of the first device and the second device, and then, if the adopted results are similar, they are adopted for advice (S79). Here, if the results adopted in steps S73 and S77 are similar, the inference result is adopted as advice. If the adopted results are dissimilar, it may be determined by inferring which one may be correct. As shown in FIG. 9, a comprehensive judgment may be made.

In this way, in the modified example of inference by combining historical data, the time-series inspection data output from the first device and the second device are used for the inference engine for the first device and the inference engine for the second device, respectively. Inference is performed by an inference engine. At the time of this inference, a correction operation is performed on each time-series inspection data, and the inference when the reliability is the highest is adopted as the inference result for each device. Finally, the inference results of the two devices are used to make a comprehensive judgment.

In this modification, two devices, the first device 2a and the second device 2b, are used, but the processing is not limited to this, and the processing may be performed using three or more devices.

As described above, the information transmission system according to the embodiment of the present invention is a first inspection data acquisition unit (ID determination) that acquires a time-series first inspection data group of a subject by the first device 2a. Part 1b), a second inspection data acquisition unit (ID determination unit 1b) that acquires a time-series second inspection data group of the subject by the second device 2b, a first inspection data group, and a first It has a transmission information determination unit (information provision unit 1c) that determines transmission information to be provided to the target person using the inspection data group of 2. The inspection data group is acquired from the first and second devices, and the information to be provided to the target person is generated based on this data. That is, since the inspection data is acquired from a plurality of devices, the number of data can be increased, and the data can be acquired in various situations, so that more accurate information can be generated. As described above, the information transmission system according to the embodiment of the present invention can grasp the accurate health condition by considering the situation of the subject and provide customized information such as advice according to the health condition. can.

Further, in one embodiment of the present invention, the transmitted information is determined according to the inference model learned according to the change pattern of the inspection data group acquired by a plurality of devices. That is, in the present embodiment, an inference model is generated using the inspection data group acquired in time series (see, for example, FIG. 8). Further, the inspection data group acquired from the subject is input to the inference engine to obtain the inference result (see, for example, S107 in FIG. 4 and S9 in FIG. 5). By using the time-series test data group, it is possible to identify the disease indicated by the change pattern of the test data, and to infer the disease that will develop in the future, the time of its onset, and the like.

Further, in one embodiment of the present invention, the first inspection data group acquired by the first apparatus and the second inspection data group acquired by the second apparatus are the first and second, respectively. The reliability is calculated for each inspection data group of the above, and the reliability when the corrected inspection data group is inferred as an input is calculated, and the transmission information is determined according to the reliability (see, for example, FIGS. 3 and 4). For this reason, when inspection data of multiple different devices is acquired, even if there is a difference in the output level of the device, this is corrected, so more accurate transmission information is determined with abundant data. can do.

Further, in one embodiment of the present invention, the first inspection data group acquired by the first apparatus and the second inspection data group acquired by the second apparatus are the first and second, respectively. It is corrected for each test data group of, and the plurality of corrected test data groups are combined into one test data group, and inference is performed by inputting the combined test data group into the inference model, and based on the inference result. The transmitted information is determined (see, for example, S9 in FIGS. 3, 4, and 5). Therefore, the inspection data group acquired by a plurality of devices can be treated as if it were one inspection data group, and the number of data increases, so that more accurate transmission information can be determined.

Further, in one embodiment of the present invention, the first inspection data group acquired by the first apparatus and the second inspection data group acquired by the second apparatus are the first and second, respectively. It is corrected for each inspection data group of, and each of the corrected multiple inspection data groups is input to the inference model, the inference result by each inference model is comprehensively judged, and the transmission information is determined based on this judgment result. (See, for example, FIGS. 9 and 10). Therefore, more accurate transmission information can be determined from the inspection data group acquired by a plurality of devices.

Further, in one embodiment of the present invention, the first and second inspection data acquisition units determine whether the inspection data group is from the subject or the inspection data group from a person other than the subject. (Refer to the ID determination unit 1b in FIG. 1), in the case of the inspection data group from the subject, the inspection data group is acquired as the first inspection data group or the second inspection data group. When the control unit of the information transmission system inputs the inspection data from the first and second devices 2a and 2b used by the target person and the inspection data from the third device 3 used by other than the target person, the target person And the inspection data of non-target persons can be distinguished. Therefore, the subject can obtain the inspection data by using a plurality of devices, can enrich the inspection data, and can obtain more accurate transmission information.

In the explanation of this embodiment, it has been explained many times that the results of stool test, stool collection, etc. performed by various sensors attached to the toilet are effectively used as the first device 2a or the second device 2b. It is not limited to. The first device 2a and the second device 2b may be any device for acquiring health-related information of the target person, for example, vital information, sample information, and the like. In the simplest example, it can be applied to face image information obtained from a mobile terminal such as a smartphone, heartbeat information based on the face image information, and the like, and these information may be utilized. In addition, it may be linked with a device such as a wearable terminal that is used in close contact with the user, and data to be noted such as arrhythmia can be easily acquired by these devices. It is also possible to detect health problems that affect the feet by the pattern of the acceleration sensor during walking. By analyzing using a history pattern that includes multiple data, instead of analyzing single-shot data that may contain errors depending on the equipment, physical condition, eating and drinking, and living scenes, the presence or absence of illness, possibility, recovery, and outpatient visits Information such as when to do it, advice information, etc. can be provided with high accuracy. If this information is of low accuracy, the user will be delayed in seeing the doctor and will be more worried than necessary.

Many of the proposals so far have insufficient measures against such accuracy, but in this embodiment, the subject is treated reasonably in consideration of the accuracy and the situation of the subject. It is possible to provide information that can be visited by institutions. Since it is possible to provide information on facilities where tests and treatments can be received to grasp the accurate health condition, users can receive treatment and improve their lifestyles with this health grasp. , You can lead a healthier life.

In one embodiment of the present invention, the control unit 1 has been described as an IT device composed of a CPU, a memory, an HDD, and the like. However, in addition to being configured in software by a CPU and a program, part or all of each part may be configured in a hardware circuit, and a gate generated based on the program language described by Verilog. A hardware configuration such as a circuit may be used, or a hardware configuration using software such as a DSP (Digital Signal Processor) may be used. Of course, these may be combined as appropriate.

Further, the control unit 1 is not limited to the CPU, and may be any element that functions as a controller, and the processing of each unit described above may be performed by one or more processors configured as hardware. For example, each part may be a processor each of which is configured as an electronic circuit, or may be each circuit part of a processor composed of an integrated circuit such as an FPGA (Field Programmable Gate Array). Alternatively, a processor composed of one or more CPUs may execute the functions of each unit by reading and executing the computer program recorded on the recording medium.

In addition, among the techniques described in this specification, the controls mainly described in the flowchart can often be set by a program, and may be stored in a recording medium or a recording unit. The recording method to the recording medium and the recording unit may be recorded at the time of product shipment, the distributed recording medium may be used, or may be downloaded via the Internet.

Further, in one embodiment of the present invention, the operation in the present embodiment has been described using a flowchart, but the order of the processing procedures may be changed, or any step may be omitted. Steps may be added, and specific processing contents in each step may be changed.

In addition, even if the scope of claims, the specification, and the operation flow in the drawings are explained using words expressing the order such as "first" and "next" for convenience, in the parts not particularly explained, It does not mean that it is essential to carry out in this order.

The present invention is not limited to the above embodiment as it is, and at the implementation stage, the components can be modified and embodied within a range that does not deviate from the gist thereof. In addition, various inventions can be formed by an appropriate combination of the plurality of components disclosed in the above-described embodiment. For example, some components of all the components shown in the embodiment may be deleted. In addition, components across different embodiments may be combined as appropriate.

1 ... Control unit, 1a ... Communication control unit, 1b ... ID determination unit, 1c ... Information provision unit, 1d ... Inference model specification determination unit, 1e ... Inference request unit, 1f ... Search unit, 2a ... 1st device, 2b ... 2nd device, 4 ... Terminal, 5 ... Learning unit, 5a ... Input / output modeling unit, 5b ... Specifications Collation unit, 6 ... Learning request unit, 6a ... Recording unit, 6b ... Object type A image group, 6c ... Teacher data, 6d ... Specification setting unit, 6e ... Communication unit , 6f ... Control unit, 7 ... Inference engine, 8 ... DB unit, 8a ... History list by ID

Claims

A first inspection data acquisition unit that acquires a time-series first inspection data group of a subject by a first device, and a first inspection data acquisition unit.
A second inspection data acquisition unit that acquires a time-series second inspection data group of the subject by a second device capable of performing an inspection capable of interpolating the first inspection data group, and a second inspection data acquisition unit.
A transmission information determination unit that determines transmission information to be provided to the target person using the first inspection data group and the second inspection data group.
Have,
An information transmission device characterized in that the first inspection data group and the second inspection data group complement each other in inspection timing or inspection items.
The information transmission device according to claim 1, wherein the transmission information determination unit determines the transmission information according to an inference model learned according to a change pattern of a test data group acquired by a plurality of devices.
The transmission information determination unit performs the first and second inspections of the first inspection data group acquired by the first device and the second inspection data group acquired by the second device. The information transmission device according to claim 1, wherein the information transmission device is corrected for each data group, the reliability when the corrected inspection data group is inferred as an input is calculated, and the transmission information is determined according to the reliability. ..
The third aspect of claim 3, wherein the transmission information determination unit performs four arithmetic operations on numerical values common to each of the data included in the inspection data group when correcting each inspection data group. Information transmission device.
The transmission information determination unit performs the first and second inspections of the first inspection data group acquired by the first device and the second inspection data group acquired by the second device. Correction is made for each data group, the corrected plurality of inspection data groups are combined into one inspection data group, inference is performed by inputting the combined inspection data group into the inference model, and the above transmission is performed based on the inference result. The information transmission device according to claim 1, wherein the information is determined.
The transmission information determination unit performs the first and second inspections of the first inspection data group acquired by the first device and the second inspection data group acquired by the second device. A claim characterized in that correction is made for each data group, each of the corrected plurality of inspection data groups is input to the inference model, and the inference result by each inference model is comprehensively judged to determine the above-mentioned transmission information. Item 1. The information transmission device according to item 1.
The first and second inspection data acquisition units determine whether the inspection data group is from the target person or the inspection data group from a person other than the target person, and the inspection data from the target person is determined. The information transmission device according to claim 1, wherein when the group is a group, the data is acquired as the first inspection data group or the second inspection data group.
The first and second inspection data groups include a color sensor for defecation, a shape sensor, a hardness sensor, an olfactory sensor (including reaction determination of nematodes and animals), a gas component sensor, and a color when a specific reagent is added. The information transmission device according to claim 1 to 7, wherein the data is obtained according to one of the output results of either the change detection sensor or the shape determination based on the magnified observation image.
Acquire the time-series first inspection data group of the subject by the first device,
The second inspection data group in time series of the subject is acquired by the second device capable of performing the inspection so as to interpolate the first inspection data group.
Using the first test data group and the second test data group, the transmission information to be provided to the subject is determined.
An information transmission method characterized in that the first inspection data group and the second inspection data group complement each other's inspection timings or inspection items.