Health management support system

Abstract

The goal is to provide support to help ordinary users who want to improve their health, or who want someone close to them to become healthier, to take appropriate steps towards achieving good health, even without specialized knowledge. [Solution] A health management diagnostic support device 21 is constructed which has a trained model 22 created using a group of training data that includes self-examination data of a subject as input data and diagnostic results by a practitioner for the subject as output data, and which uses the trained model 22 to output the estimated diagnostic results from the self-examination data of one of the subject individuals.

Description

Technical Field

[0001] This invention relates to a system for supporting health management by AI.

Background Art

[0002] When a general user tries to purchase a product, there are many cases where it is not determined what to buy. In such a situation, systems that promote purchases by proposing and allowing selection of products based on the user's purchase history, personal information, etc. have been introduced in various e-commerce sites.

[0003] For example, Patent Document 1 proposes a system that proposes products to purchasers who intend to give a gift, based on the attributes of the purchaser and the recipient. This creates a learned model based on various information regarding the purchaser and the recipient, and proposes products that are output as suitable by this learned model.

[0004] On the other hand, in the medical field as well, there has been consideration of constructing a learned model using examination data and interview data of a diagnostic subject in order for a doctor to make the best diagnosis in a hospital and supporting the determination of diseases (for example, Patent Document 2).

[0005] Further, Patent Document 3 describes an automatic diagnosis support device that can derive a diagnostic result similar to the comprehensive judgment by an experienced and skilled Kampo specialist from information on the patient's symptoms obtained from interview items. A learned model is created by machine learning based on the information on the patient's symptoms and the diagnostic result by a Kampo specialist, and it is described that a diagnostic result as an analysis result is output using this learned model based on the patient's self-interview data.

[0006] Furthermore, Patent Document 4 describes a system in which a server at an acupuncture and osteopathic clinic transmits pre-examination data to a patient's terminal, including questions for collecting Oriental medicine-based questionnaire items for health management and Western medicine-based questionnaire items for treatment, and collects patient symptoms by having the patient answer the questions via the terminal. Among these, it is proposed to include information necessary for predicting diseases within the Western medicine-based questionnaire items.

[0007] Furthermore, Patent Document 5 describes a function of a medical questionnaire creation support device that acquires personal health record data from the patient's mobile device. This personal health record data is described to include biometric data such as height, weight, blood pressure, heart rate, blood glucose level, and body temperature, as well as lifestyle data such as activity level, sleep duration, meal menu (calorie amount), and steps taken. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] Patent No. 7064738 [Patent Document 2] Japanese Patent Publication No. 2023-033182 [Patent Document 3] Japanese Patent Publication No. 2021-047504 [Patent Document 4] Japanese Patent Publication No. 2024-090940 [Patent Document 5] Japanese Patent Publication No. 2020-013245 [Overview of the project] [Problems that the invention aims to solve]

[0009] However, the technology described in Patent Document 2 was intended solely to support doctors' diagnoses within hospitals, and its content was highly specialized. Therefore, it could not be used by judo therapists, physical therapists, instructors, or trainers to support the daily health management of their clients, nor could it be used by general users for their own everyday health management.

[0010] Furthermore, while it is conceivable that the technology for deriving diagnostic results from self-examination data described in Patent Document 3 could be modified to include disease prediction information, such as that mentioned in Patent Document 4, instead of the examination data, or to combine it with biometric data, as described in Patent Document 5, the output of these methods would be directed to specialists such as doctors within hospitals and other facilities, resulting in diagnoses for disease treatment by specialists. These were not intended to be shared externally like medical records. Therefore, it was not envisioned that external practitioners such as instructors or trainers would use them, nor would patients themselves use them externally, making it difficult to directly translate them into improvements in daily life.

[0011] Therefore, this invention aims to improve the health environment for the nation by providing ordinary users who want to be healthy, or who wish for the health of their close relatives or loved ones, with ongoing and future support, such as raising awareness about prevention and early detection and treatment of pre-disease states, providing support for those who wish to improve their health, and assisting in prompt social reintegration and recurrence prevention after surgery or illness. [Means for solving the problem]

[0012] This invention is The system has a pre-trained model created using personal data, including self-assessment data of the subject, as input data, and a set of training data, including diagnostic results from the practitioner for the same subject, as output data. A health management diagnostic support device that uses the aforementioned trained model to output the estimated diagnostic result from the personal data of one of the aforementioned target individuals, The above problem was solved by the first solution, which involves using the aforementioned predicted diagnostic results as a health management diagnostic support device that includes advice for recommended activities.

[0013] Furthermore, in the first solution, this invention provides Have a subject database with identifiable personal information for each individual subject, The personal information includes the usage history of the health management diagnosis assistance device and the target values included in the recommended advice, The usage history of the subject and the content for achieving the target values can be output, and a second solution means can be adopted.

[0014] Furthermore, in the second solution means of this invention, A third solution means can be adopted in which the content includes images, videos, or both that show the activities included in the advice.

[0015] Furthermore, in the second or third solution means of this invention, A fourth solution means can be adopted in which the content includes content that interactively displays the progress towards the target value on the terminal.

[0016] Furthermore, in the first to fourth solution means of this invention, Have a user database that records the subject in an identifiable manner, and a different requester than the subject can input the self-questioning data of the subject, and a fifth solution means can be adopted to output the estimated diagnosis result for the subject.

[0017] Furthermore, this invention can adopt a sixth solution means to be a health management diagnosis assistance system having the health management diagnosis assistance device in the first to fifth solution means and A terminal that outputs the diagnosis result.

[0018] Furthermore, this invention Using a learned model created using a teacher data group including personal data including the self-questioning data of a subject as input data and the diagnosis result by a practitioner for the subject as output data, output the estimated diagnosis result from the personal data of an individual who is one of the subjects, As a seventh solution means, a health management diagnosis support method can be adopted in which the presumed diagnosis result includes advice for recommended activities.

Advantages of the Invention

[0019] According to the present invention, general users who want to be healthy or become healthy, and people who want general users to become healthy can continuously receive advice for recommended activities output by a learned model by submitting self-diagnosis data. Especially in the assistance of diagnosis, when the part where a person feels symptoms may be different from the actual affected part or the part with symptoms, there are practical problems with insufficient accuracy in a database classified only by simple cases. In the health management diagnosis support device according to this invention, a learned model that can find features by linking self-diagnosis data and the diagnosis result by a practitioner is created, and by using this, a highly probable diagnosis result can be submitted.

[0020] The advice for recommended activities included in this diagnosis result is a proposal and goal for activities that contribute to health by the subject continuously practicing. If the diagnosis result includes a health stage, the advice will be more suitable. When the diagnosis result is in the pre-disease stage before an injury, after surgery, etc. are confirmed, as the above advice, it includes self-improvement content such as maintaining or improving health, or strengthening muscle strength and physical strength so as to be primary prevention of the disease. When the diagnosis result is in the post-disease discovery stage where a precursor or initial stage of a disease is discovered, as the above advice, it includes treatment content so as to be secondary prevention of the disease. When the diagnosis result is in the post-disease and post-surgery stage after a prescription for a disease or surgery, etc. has been completed, it includes treatment or rehabilitation content. By setting an activity goal and outputting content that recommends the activity to the terminal of the subject so that it can be continued as continuous advice based on those diagnosis results, continuous advice can be continuously provided to contribute to health.

[0021] The content should include not only text outlining the proposal, but also images and videos that clearly demonstrate practical explanations and actual actions related to that text, making it easier for the target audience to implement the proposal. These images and videos can be automatically generated based on the text, or they can be pre-stocked images explained by experts corresponding to the anticipated text, or videos demonstrated by experts.

[0022] Furthermore, if the content includes not only the suggested exercises displayed on the user's device, but also interactive displays showing their progress towards their activity goals, the system can monitor the user and support their activities even without a physical therapist or trainer constantly present. This makes it easier to maintain health than simply continuing solitary exercise for health.

[0023] Furthermore, the social significance of the system is further enhanced if the progress toward the activity goals can be checked not only by the subject but also by the requester, such as a close relative who wishes for the subject's health, from their own device. By allowing the requester to check the progress toward the stated goals, which is displayed interactively on their device, the requester can feel at ease and provide support and guidance to the patient (subject) remotely.

[0024] Specifically, similar to personal advice provided in nursing homes, companies (employee benefits), and sports gyms, the diagnostic results output from the health management diagnostic support device can be used to provide highly accurate health support and advice. Furthermore, individual users can refer to the diagnostic results output by the health management diagnostic support device on their personal devices such as smartphones, and use this information to manage their health in their daily lives, including their diet and exercise.

[0025] For health management support and advice, practitioners who support necessary exercise in everyday environments (physical therapists, instructors, trainers, etc.) can use a health management diagnostic support device that operates as an ASP (Application Support System), and use the diagnostic results to help input the above activity goals for the subject. This allows for a combination of diagnosis using a trained model and assistance from a specialist. [Brief explanation of the drawing]

[0026] [Figure 1] Conceptual diagram of the generation stage of a trained model used in the health management and diagnostic support device according to this invention. [Figure 2] Block diagram of an example hardware configuration used in the generation stage of a trained model for the health management and diagnostic support device according to this invention. [Figure 3] Functional block diagram of an embodiment of the health management and diagnostic support device according to this invention. [Figure 4A] An illustrative diagram showing an example of how to display and select options in a medical questionnaire using a dropdown menu. [Figure 4B] An illustrative diagram showing an example of how to display and select options in a medical questionnaire, with some parts entered via chat. [Figure 4C] An image illustrating another example of displaying and selecting options in a health management questionnaire using a dropdown menu. [Figure 5] Table showing the correspondence between input and output data used for training. [Figure 6] Table showing examples of customized advice output for each target group. [Figure 7] An example diagram showing the progress toward activity goals on the device screen as content. [Modes for carrying out the invention]

[0027] The present invention will be described below with specific embodiments. This invention is a health management diagnostic support device that assists in the diagnosis results based on data related to the health of a subject, in order for a practitioner to manage the health of the subject, and a method for providing a health management diagnostic service performed thereby.

[0028] In this invention, the client is a person who applies for the service for the health of themselves or a close relative. The recipient submits their own self-assessment data to the health management diagnostic support device and receives the benefit of the diagnostic results being supplemented by the practitioner's diagnosis. The client and the recipient may be the same person, but the client may also register a family member or acquaintance as the recipient. If the client is the recipient, they are the person who "wants to be healthy." If the client and the recipient are different, the recipient is the person the client "wants to be healthy." The client can be an individual or a corporation. For example, a corporation can be the client and register its employees as recipients. If the client is an individual, close relatives can be the client's parents, children, grandchildren, relatives, boss, friends, etc., and there are no particular restrictions.

[0029] A practitioner is a person who possesses specialized knowledge and provides health advice, diagnosis, and treatment to a subject, such as a doctor, physical therapist, judo therapist, instructor, or trainer. In this invention, the practitioner is a person who uses the output of the diagnostic results predicted by the health management diagnostic support device as a reference to support the actual diagnosis. A user is an individual, corporation, or organization, including practitioners, who has the authority to use the health management diagnostic support device.

[0030] The health management and diagnostic support device according to this invention may be a physical device consisting of a single server or a group of servers located on a network, or it may be a virtual device on the cloud that implements a database and other means described later. The program according to this invention is a series of programs that cause these physical or virtual devices to run as the health management and diagnostic support device.

[0031] The health management and diagnostic assistance device according to this invention obtains output data as inference from input data using a trained model that has been trained using training data (*) (training stage) (utilization stage). (*: Here, the set of input data used for training and the correct output data for said input data is called training data.)

[0032] The health management and diagnostic support device according to the present invention is connected to a terminal via a network. During the learning phase, that is, when training a pre-trained model to be used in the health management and diagnostic support device, it is connected to a learning terminal that loads training data and sets parameters. During the utilization phase, that is, when providing the service according to the present invention, it receives personal information of the subject, including self-examination data, from various terminals (personal terminals and input terminals), makes inferences using the pre-trained model based on this data, outputs the inferred diagnostic result, and displays the diagnostic result on the practitioner's terminal (practitioner terminal).

[0033] First, let's explain the learning phase. In the learning phase, a trained model is created using supervised learning, with training data consisting of sets of input and output data. The model format can be any general method using a neural network, and is not particularly limited.

[0034] The input data will consist of personal data corresponding to the subject. Specifically, this includes self-assessment data such as age, gender, occupation, residential area, average sleep time, alcohol consumption, tobacco consumption, calorie intake, medical history, medications, family history, past allergy history, and travel history, as well as height, weight, chest circumference, waist circumference, and other measurable data. Measurement data can be obtained using general measuring instruments, smartphones, smartwatches, or other measuring devices. Smartwatches equipped with sensors such as accelerometers and gravity sensors can measure steps, walking distance, stride length, and walking balance; sleep quality and sleep duration using sleep sensors; calories burned through exercise measurement; stress levels, and blood oxygen levels. In addition, measurement data obtained by directly measuring the body can also be used. • Body composition analysis: Body components that make up body weight (body water, protein, minerals, body fat) • Muscle type: Standard, robust, obese, etc. • Muscle and Fat: Obesity, Lipids, Body Shape, and Obesity (The degree of obesity can be determined from both BMI and body fat percentage) • Muscle mass by body part: Confirm and understand the muscle mass and muscle balance in each body part: right arm, left arm, torso, right leg, and left leg. • Body fat percentage by body part: Body fat percentage in the right arm, left arm, torso, right leg, and left leg. • Recent measurement, progress, and target data (body weight, muscle mass, body fat percentage, extracellular water ratio) • Data obtained from blood tests (blood glucose levels, cholesterol levels, LDL, HDL, red blood cell and white blood cell counts, hemoglobin, ALT, AST, protein levels, etc.) These are some examples. These measurements are classified as follows: Physical measurements include height, weight, body fat percentage, and waist circumference, as well as measurements of metabolic syndrome risk factors such as height, weight, obesity level (body fat percentage), chest circumference, grip strength, and lung capacity. Physiological measurements include electroencephalogram (EEG), electrocardiogram (ECG), electromyogram (EMG), respiration, body temperature, blood pressure, and blood oxygen levels. Eye tracking is used for measuring facial expressions and other similar functions. Exercise measurement includes measuring the amount of exercise, such as the number of steps taken, and measuring before and after exercise. Morphological measurements include measuring body composition (fat mass, muscle mass, etc.) and skeletal structure. Biochemical measurements include blood tests, urine tests, and other tests that measure the chemical components in the body.

[0035] It is not necessary to have all of this data, either as training data or practical data; this invention can be used even with only some of the data. However, the more data there is, the higher the accuracy of the predictions. Furthermore, it is preferable to create a trained model and then add data as needed. It is preferable to be able to increase the types of input data that can be input as needed from the initial stage, as this can improve the accuracy of the predicted diagnostic results.

[0036] The output data is a diagnostic result for an individual corresponding to the subject with the input data. This diagnostic result should ideally include, or be able to link, advice for activities recommended for the subject afterward, in addition to information obtained through the practitioner's examination. The information obtained may include, for example, illnesses, injuries, symptoms, treatment methods, and response plans identified by a physician's diagnosis.

[0037] Advice for recommended activities should be appropriate to the content and stage of the activity. For example, a specific example would be defined as a diagnostic result based on the input data of the individual, such as, "Given this age, muscle type, and medical history, parallel bar exercises are preferable to jogging. Parallel bar exercises should ideally last about 10 minutes."

[0038] The stages can be divided into, for example, the pre-disease stage, the post-disease / post-surgery stage. The pre-disease stage is a situation where no specific symptoms have appeared, or where the condition has not yet been discovered or confirmed. This includes situations like "feeling vaguely unwell." Age-related decline is also included here. The post-disease / post-surgery stage is a situation where some kind of illness or disease has become clear as a result of diagnosis or other means. This can include cases where the name of the disease has been confirmed or not. It also includes cases where it has been confirmed that the person is deviating from a healthy state in some way. The post-disease / post-surgery stage is the situation after treatment or surgery has been completed. In the case of sprains or fractures, it is necessary to take care to prevent recurrence even after the initial recovery. In the case of surgical treatment, it is necessary to take care of post-operative recovery and prevent the recurrence of the disease that required the surgery. However, these stages are not strict, and a person may span multiple stages.

[0039] These input data are processed into statistically standardized sets of values ​​(x) (where x is a vector) so that they can be used as feature data for training and utilization. The set of input and output data used as training data during the training phase consists of information about one individual sample. In other words, the training data shows that an individual with such self-assessment data and measurement data (input data) would have received such a diagnosis result (output data). A large amount of such training data is prepared to train the trained model. When training, it is easier to treat the input and output data as vectors.

[0040] The input data includes, for example, self-assessment data such as an individual's age, gender, occupation, residential area, and medical history, as well as measurement data such as height, weight, and blood glucose levels. All of this data is standardized (mean: 0, variance: 1), and the resulting vector values ​​x = (x1, x2, ..., x n ) becomes the feature data. For each diagnostic result, the output data is φ(x)=w ~ x = w1x1 + w2x2 + ... + w n x n (w is the "weight" vector, w ~ is the transpose vector of w. In general notation, it is written in bold as shown in Figure 1, but for the sake of notation in the specification, it is written in regular letters. Adjust "w" so that the result classified by the function value given by ) calculates the output data that is considered correct. For example, when φ > 0, it is interpreted as corresponding to the diagnostic result. In general, w will have a different value for each φ (diagnostic result). This relationship is shown in Figure 1.

[0041] In the actual learning phase, it is preferable to operate the learning server 12, which has a learning model (which becomes a trained model after learning), from the learning terminal 11. The relationship during this learning phase is shown in Figure 2. Figure 2 shows an embodiment in which the learning terminal 11 and the learning server 12 are connected via a network, but is not limited to this. The learning terminal 11 can be a general-purpose personal computer, smartphone, etc. If the learning server 12 is a physical server, it may be integrated with the learning terminal 11, or it may be connected to the learning terminal 11 via a network as shown in Figure 2. Furthermore, the learning server 12 does not need to be a single unit, but may be a combination of multiple physical servers. Moreover, the learning server 12 may be a cloud server.

[0042] The learning server 12 receives a large number of training data sets (training data sets) from the learning terminal 11, databases and other servers connected via the network (collectively referred to as external servers 13 in Figure 2), and performs learning to adjust w in the relationship shown in Figure 1. As a result, the learning model w of the learning server 12 is trained according to the training data sets, and a trained model can be created that can return output data with a high probability as a diagnostic result for the input data.

[0043] As an example of learning to adjust w, suppose that feature data of a subject is used as input data, and the diagnostic results shown as output data include advice for recommended activities that include two exercises: "parallel bar training" and "jogging." On the other hand, if a professional such as a physical therapist or trainer actually recommends only "parallel bar training" for this subject and not "jogging," then by adjusting the value of w so that φ>0 (applicable) is maintained for "parallel bar training" and φ<0 (not applicable) for "jogging," the accuracy of the advice for recommended activities by the professional is improved.

[0044] As the learning process progresses, the weights of w are optimized. In particular, at the beginning of the learning process, self-assessment data, which tends to reflect expert knowledge, shows a stronger correlation with the output data, while measurement data shows a weaker correlation. The w (weight) for measurement data is set to 0 when the correlation is unknown. As the learning process continues, the w for each measurement data is adjusted, and the accuracy improves.

[0045] This learning process is more preferable if it incorporates feedback that reflects the subsequent results of participants who received advice on recommended activities, as this further improves its accuracy.

[0046] An embodiment of the health management and diagnostic support device 21 using this trained model will be described with reference to Figure 3. The health management and diagnostic support device 21 is a physical server or a cloud server and is accessible from the following terminals via a network. It may be on the internet or on a closed network. In the figure, it is shown as a single device, but it may be composed of multiple servers.

[0047] The health management and diagnostic support device 21 has, or at least has available, a trained model 22 obtained through the above-described learning process. It is preferable for operational speed if this model is stored within the health management and diagnostic support device 21's own memory. Alternatively, to distribute the processing load, it may be stored and made executable on another server connected via a network.

[0048] The health management and diagnostic support device 21 has an auxiliary estimation means 23 that inputs input data of a designated subject into the trained model 22 and outputs a diagnostic result corresponding to that subject. This auxiliary estimation means 23 may be executed by the operator of the health management and diagnostic support device 21 from the administrator terminal 45, or by practitioners such as doctors, physical therapists, judo therapists, instructors, and trainers from the practitioner terminal 44 described later. In addition, it may be made possible to execute it from terminals owned by the subject or client personally (subject terminal 41, client terminal 42) to a limited extent.

[0049] The health management and diagnostic support device 21 may have a user database 25 for registering individual users. However, the users in this health management and diagnostic support device 21 may be defined as subjects, requesters, and users, and each should be managed with separate settings or managed in separate databases. However, this does not prevent the same individual from being both a subject and a requester. For example, a requester who applies for a family member to be a subject because they want the family member to be healthy can also become a user of the health management and diagnostic support device 21 as a subject themselves. In the embodiment shown in the figure, an example is shown in which the system consists of a subject database 26 (subject DB 26) containing information about subjects, a requester database 27 (requester DB 27) containing information about requesters, and a user database 28 (user DB 28) containing information about users.

[0050] Of these, the subject database 26 contains input data that constitutes personal information, along with an identification code (subject ID) that can identify the subject. The input data is recorded by the data reception means 31, which will be described later. The input data does not need to include all items expected in the database; it may consist of partial measurement data or self-examination data, and it is desirable that data can be added or rewritten at any time even after a subject record has been created. Since the subject's condition naturally changes over time due to new physical measurements or the progression of their illness, it is desirable that the most up-to-date information is reflected.

[0051] The client database 27 should contain an identification code (client ID) that can identify the client, as well as identification information such as a password that the client uses to access the health management and diagnostic support device 21 from their terminal 40, and information for payment of service fees, etc. It should also contain information that links the client to the target person ID in the target person database 26, who is the person the client wishes to see become healthy. If the client is also a target person, it should contain information that links them to the target person ID so that their identity as the same person can be managed.

[0052] The user database 28 should ideally include information about users such as hospitals, chiropractic clinics, and sports gyms, as well as information about each practitioner among the users. Along with an identification number that can identify each practitioner, the database may also individually record their access rights for the target user IDs whose diagnostic results are available to them.

[0053] In cases where practitioners from a single hospital organization, such as within a large hospital complex or affiliated clinics, perform treatments at different locations or at the premises of a corporation (company), it is advisable to use a combined account system consisting of a user ID and password for the affiliated clinic or corporation, and a practitioner ID and password for each practitioner belonging to that corporation. By using the department name, individual name, and other classifications within the corporation as the practitioner ID, practitioners can be easily distinguished, and internal transfers within the corporation can be accommodated. Furthermore, passwords can be managed on a per-corporation basis, making the system easier to operate.

[0054] Examples of terminals 40 that utilize this health management diagnostic support device 21 include terminals installed in facilities related to direct treatment such as hospitals and clinics (hospitals, etc.), terminals installed in facilities related to health management such as sports gyms and public baths (gyms, etc.), terminals owned by individuals or clients, terminals installed by contractors who are commissioned by hospitals, gyms, etc. to perform work (including terminals owned by visitors when a person receiving treatment receives treatment at their home, such as in home care), and terminals installed by operators who operate the health management diagnostic support device 21 or the trained model 22 for management purposes (administrator terminals). Data can be input and output to these terminals by means described later.

[0055] The above-mentioned terminal 40 is not limited to any particular type, as long as it has network capabilities, such as a personal computer, tablet, smartphone, or game console.

[0056] Of the terminals 40, the terminals owned by individuals are the terminal owned by the subject (subject terminal 41) and the terminal owned by the requester (requester terminal 42). The subject and the requester access the health management and diagnostic support device 21 from these terminals and transmit information that will be used as input data for the subject. The data reception means 31 of the health management and diagnostic support device 21 records the transmitted information in the user database 25, which will be described later. In order to manage the transmitted data for each subject, the health management and diagnostic support device 21 may have a user management means 32 that authenticates and logs in the user count of the requester or subject. A general login function can be used as the user management means 32. The subject terminal 41 and the requester terminal 42 transmit their ID, password, and other authentication information to the health management and diagnostic support device 21 in order to be authenticated by the user management means 32. Specifically, they access the health management and diagnostic support device 21 using a general web browser or a dedicated application and transmit the subject ID or requester ID along with the password and other authentication information from the displayed screen.

[0057] It is preferable that facility terminals 43, which are installed in the aforementioned hospitals, gyms, and other facilities, also allow users to log in to the health management and diagnostic support device 21. The login screen for the health management and diagnostic support device 21 is displayed on the screen of the facility terminal 43, and the user enters the subject ID and requester ID using an input device such as a keyboard or pointing device on the facility terminal 43. The user is then authenticated by the user management means 32 of the health management and diagnostic support device 21, and login becomes possible. After logging in, the data measured for the subject is sent to the health management and diagnostic support device 21 as input data.

[0058] This facility terminal 43 may be not only a personal computer or tablet, but also training machines and other measuring instruments 46 that can be directly connected to the health management and diagnostic support device 21. In addition, the measuring instruments 46 are connected to the facility terminal 43 wirelessly or by wire, and the data measured by the measuring instruments 46 is transmitted to the data receiving means 31 via the facility terminal 43.

[0059] Among the terminals 40 installed in the above-mentioned hospitals, gyms, and other facilities, the practitioner terminal 44 operated by practitioners such as doctors, physical therapists, judo therapists, trainers, and instructors manages the practitioner's own account and allows the practitioner to input information about the account of the person they are in charge of. When making an inquiry, it is desirable that authentication is performed based on the information in the user database 28, and that the practitioner transmits the person ID of the person they intend to diagnose. Furthermore, when transmitting the person ID, it is preferable that the person or client be able to secretly input authentication information such as a password to authenticate the person ID, and that the practitioner be able to temporarily grant permission to access the person's information. Personal information is protected by limiting the acquisition of estimation results about the person to only when the person or client decides that they want the practitioner to use the estimation results.

[0060] Specific methods for authenticating permission to access a subject's information include, for example, the following procedure: The authentication means 35 of the health management and diagnostic support device 21 transmits identifiable information, such as a QR code, to the subject's terminal 41, such as a smartphone, owned by the subject. The practitioner's terminal 44 reads the identifiable information, such as a QR code, displayed on the screen of the subject's terminal 41, and sends a URL to the health management and diagnostic support device 21 indicating that it has been read, thereby confirming that the subject's consent has been obtained. Verification can also be achieved by sending a one-time password or other authentication information to the subject's terminal 41, which the practitioner's terminal 44 then enters. Other methods can be adopted as appropriate, and are not particularly limited.

[0061] The fact that permission has been granted to access the subject's information is registered and managed in the user database 25. For example, one method is to register the user ID of the practitioner who granted access as the person who granted access in the record of the subject database 26, and then allow access from there. Conversely, another method is to register the subject ID of the person who has been granted access to the practitioner's practitioner ID, and then the practitioner's terminal 44 is only permitted to access the information of other people for the person with the registered subject ID. The method is not limited to these examples, as long as the information of the subject that the practitioner can access is properly managed.

[0062] When the health management diagnostic support device 21 receives an inquiry about a target person from the practitioner terminal 44, the auxiliary estimation means 23 reads the input data for the specified target person stored in the target person database 26 of the user database 25, inputs it into the trained model 22, and obtains output data that results in a diagnosis with a high probability. At this time, it is preferable to specify parameters along with the read input data so that the diagnostic results in the field required by the practitioner can be obtained. These parameters are input from the practitioner terminal 44 and introduced into the trained model 22. Examples of these parameters include attributes related to the type of health equipment used by the practitioner during treatment. In other words, it can be used to address a wide range of applications and situations, such as judgment and response to general illnesses and injuries, rehabilitation, exercise for health management, and dietary therapy.

[0063] By referring to the output results estimated by the auxiliary estimation means 23, practitioners can increase the information they have to make decisions when performing diagnoses and treatments, enabling them to obtain more accurate diagnoses in a shorter time. Alternatively, the output results can also include recommendations or designations of practitioners or treatment facilities where the subject should seek diagnosis, depending on the treatment content desired by the subject (treatment, rehabilitation, exercise, dietary therapy, etc.). The output results can include not only categories but also reference values ​​and target values, which can be used by both practitioners and subjects. On the other hand, even if the subject has a desired treatment content, if a more serious illness is suspected, the output results may be configured to partially or completely invalidate that request and promptly encourage diagnosis and treatment.

[0064] Furthermore, if the device is used from the practitioner terminal 44, it is advisable to record in the subject database 26 as a history that the designated subject used the facility where the practitioner terminal 44 is installed. This is because the history of facility use and diagnosis will lead to improved accuracy in subsequent estimations.

[0065] Furthermore, the health management and diagnostic support device 21 may have a history confirmation means 33 that outputs the usage history recorded in the subject database 26 for a subject specified from the terminal 40. It is even more preferable that this history confirmation means 33 can refer to input data of the subject other than the usage history recorded in the subject database 26. If the changes in measurement data such as weight, body fat percentage, muscle mass, and BMI over time can be displayed in a graph along with the numerical values, it is suitable for practitioners, subjects, and clients to understand the subject's current condition.

[0066] Furthermore, the health management diagnostic support device 21 may have a self-examination means 34 that receives instructions about the subject from the subject terminal 41 or the requester terminal 42, inputs the subject's input data into a trained model 22, and outputs recommended medical departments and activity advice as diagnostic results. Estimating recommended medical departments can provide the subject or requester with information to decide which medical department to consult before consulting a practitioner. In addition, activity advice can provide guidance on activities that can be spontaneously performed at home or in everyday environments before a specific medical consultation with a practitioner. These are activities with a high probability of improvement, estimated based on information such as menus to be performed that are advised in the relevant situation during the learning stage. Examples include suggestions for exercises that lead to muscle strengthening in specific areas, suggestions for complex exercises such as Pilates or yoga, general dietary therapy and nutritional guidance for weight loss, suggestions for exercise, explanations and suggestions for preventive rehabilitation, suggestions for relaxation, and guidance on sleep duration. To enable high-accuracy estimation of these, it is preferable to use educational data containing information on these consultation menus in the output data during the training stage of the trained model.

[0067] Furthermore, if the health management and diagnostic support device 21 can output usage history output by the history confirmation means 33 and estimated diagnostic results output by the self-examination means 34 for a single subject, in accordance with the output on the terminal 40, this information can be viewed in a unified manner and used as information for health management. "Integrated output" includes not only outputting everything at once, but also outputting sequentially in accordance with screen switching, such as on an app or web screen. It is also preferable that input data other than the usage history of a single subject can be output along with the information, as this would further enrich the information provided.

[0068] The output results may include target values ​​for rehabilitation, exercise, and dietary therapy, as well as reference values ​​for those target values. The values ​​output by the health management diagnostic support device 21 may be displayed and set as they are, or the practitioner, subject, and client may agree to set the final target values ​​based on the output values. The subject and client can continue their daily health activities by performing rehabilitation, exercise, and dietary therapy while checking these target values. These target values ​​may also be included as data linked to the subject ID in the subject database 26. Furthermore, it is preferable that the effective values ​​be transmitted from the subject terminal 41 or facility terminal 43 (including the measuring instrument 46) to the subject database 26, linked to the subject ID, and displayed and analyzed in comparison with the target values.

[0069] The health management diagnostic support device 21, after outputting the diagnostic results, preferably receives feedback at a later date, such as the final diagnostic results from the practitioner and the effective values ​​achieved by the subject, to further retrain the trained model 22. The diagnostic results are values ​​with a high probability based on the trained model 22 learned from the previous training data, and are not necessarily correct. The accuracy can be further improved by correcting them with new training data. In addition, the accuracy of the diagnostic results can be improved by including the subject's subsequent effective values ​​and actual diagnostic results in the training data.

[0070] The health management and diagnostic support device 21 may offer a service to the target person or client through a self-examination means 34, and may also offer a package 29 with selected or customized functions as a purchasable service. For example, a "health management package" may be offered with parameters specified to allow customization of the output diagnostic results according to age, gender, health check results, medical history, region, season, etc. The target person can purchase a health management package that suits their situation or apply for the service, and appropriate treatment and therapy goals will be output.

[0071] For example, in a health management package, the following optimization of diagnostic results can be expected: If the hobby is golf, the diagnostic results might indicate that the self-image of rotating the hips during golf swings, etc., can lead to lower back pain when combined with raising the arms, or that bending forward can cause inflammation of the lumbar facet joints. Furthermore, for shoulder pain, it might suggest that what seems like frozen shoulder is more likely to be a rotator cuff tear, leading to suggestions such as an MRI. However, if the person is a woman in her 20s working in sales and has a long commute by train, the estimated cause of the pain and the recommended course of action will change significantly. In addition, based on the subject ID information and the symptoms of the lower back pain, it predicts that the cause of the lower back pain is the back muscles rather than the abdominal muscles, calculates muscle strengthening exercises to strengthen the back muscles, and the health management diagnostic support device suggests the optimal load based on the results of the movements that cause pain. If the practitioner judges from this suggestion that the pain may be of internal organ origin rather than muscle pain, it may be possible to output a recommendation or referral to an orthopedic or internal medicine specialist for X-rays or CT scans.

[0072] In such packages 29, parameters suitable for the target user's situation can be defined to optimize the output results. Such settings can be registered as a configuration file in storage, separate from the user database 25, making it easier to adjust the input information of the trained model 22.

[0073] When a user receives the service themselves, they purchase a package 29 tailored to their needs and receive continuous diagnostic results as a service on their user terminal 41. Alternatively, a requester who wishes for the user's health can apply for the service separately from the user, selecting a package 29 that is appropriate for the user. Package 29 may include not only parameters but also service fees. Furthermore, package 29 may include access rights to facilities that partner with the health management diagnostic support device 21 service. In this case, it is preferable if the facility has a facility terminal 43 installed or if the facility, as a corporation, has a user ID, as this allows for centralized management of health information, improving diagnostic accuracy and facilitating feedback. The application procedure for package 29 can be, for example, as follows (1) to (4).

[0074] (1) The client or the person to be treated pays the fee through a payment service such as a credit card or convenience store payment, and the health management and diagnostic support device 21 registers the client ID or person to be treated as a paid account in the user database 25. At this time, it is also advisable to specify the package 29 that will be used after payment. If contact information such as the address, phone number, or SNS ID of the person to be treated terminal 41 is registered at the time of application, the health management and diagnostic support device 21 sends the account information to that contact information. If the client and the person to be treated are different people, it is advisable to also notify the person to be treated as a gift from the client that the service is available to them.

[0075] (2) The subject sends further necessary information from the subject terminal 41 to the data reception means 31. This becomes the input data for the medical interview. Figure 4A shows an example of input for a medical interview when there is some kind of pain. Information that can be used as material for diagnosis is entered sequentially using a pull-down menu. In the example in the figure, the selected and entered parts are enclosed in thick lines. The website screen displayed on the subject terminal 41 can be switched for each medical interview, or the input can be done sequentially on a scrolling screen. Alternatively, a dedicated app can be registered on the subject terminal 41 and the input can be done through that. The health management diagnostic support device 21 executes the auxiliary estimation means 23 and sends the relevant diseases as a diagnostic result to the subject terminal 41. Then, it shows a list of facilities that can provide the service and presents information for making a reservation.

[0076] Alternatively, the questionnaire up to a certain point in Figure 4A can be entered via chat. An example of this input method is shown in Figure 4B. In this case, the data reception means 31 of the health management and diagnostic support device 21, or a means assisting it, basically repeatedly asks questions to the subject terminal 41 to fill in the necessary questionnaire items. What is needed here is the wording in the answers to the initial questions asked using the pull-down method in Figure 4A. In the example in Figure 4B, after completing the questions corresponding to questionnaire 3 in Figure 4A, questions are sent to the subject terminal 41 from questionnaire 4 onwards, similar to the questions in Figure 4A, to input the situation in detail.

[0077] As another format, Figure 4C shows an example of input when the health management diagnostic support device 21 conducts a medical interview with a subject who is trying to change their lifestyle for beauty and health reasons. The subject inputs their current weight and target weight, as well as their lifestyle and eating habits. The health management diagnostic support device 21 executes the auxiliary estimation means 23 and the self-examination means 34 and transmits improvement suggestions to the subject's terminal 41.

[0078] (3) Subsequently, the right to use the facility is issued to the subject, and data from the time of facility use is transmitted from the measuring device 46 to the data receiving means 31 via the facility terminal 43 and registered in the subject database 26. The subject can view the data output by the auxiliary estimation means 23 based on the above input data, along with the information registered in the subject database 26 via the subject terminal 41. If the requester is different from the subject, the data may also be viewed from the requester's terminal 42 to the extent permitted by the subject. The requester can obtain information on the health status of the subject, for whom they wish for good health, as well as information on anticipated diseases, and can also recommend exercises that the subject should engage in.

[0079] (4) In the case of diagnosis based on information such as pain, as illustrated in Figures 4A and 4B, the auxiliary estimation means 23 outputs possible diseases, etc., which can be referenced on the subject terminal 41. At the same time, the health management diagnostic support device 21 can recommend a facility among the facilities registered as users that can handle that disease, etc., and encourage the subject to get a diagnosis. Based on the steps shown in Figures 4A and 4B as history, the health management diagnostic support device 21 executes the auxiliary estimation means 23 and transmits the diagnosis result to the subject terminal 41, or to the facility terminal 43 or practitioner terminal 44. It is preferable that the diagnosis result here be set to selectively output content menus such as training that can be performed at that facility. Furthermore, based on the diagnosis result, the health management diagnostic support device 21 may transmit a questionnaire tailored to the diagnosis result to the subject terminal 41 and accept input of further information (Step 2~). This enables a more accurate diagnosis and the provision of a suitable menu. These records are stored in the subject database 26. Therefore, in cases of follow-up visits after a gap of about five years, it is possible to conduct a medical interview that incorporates the latest information while utilizing the information collected in previous steps.

[0080] Figure 5 shows an example of a training dataset to be used to train the pre-trained model described above. Column 1 is the broad classification of disease names, and Column 2 is the specific disease name and symptoms. However, in the pre-disease stage, inferences are also included. Column 3 consists of characteristic items corresponding to the medical interview and is personal data used as input data. Column 4 is advice for recommended activities and is the suggested content included in the diagnostic results, which are the output data. In other words, this training data is a list of cases in which a specialist would give advice like that in Column 4 if there is an item corresponding to Column 3. Also, "post-surgery" in Column 3 indicates the post-illness and post-surgery stage of tertiary prevention.

[0081] The accuracy of the recommended activities presented in the diagnostic results can be improved by categorizing them according to the stage of the diagnosis. These stages can be divided into the pre-disease stage, which involves primary prevention; the post-disease stage, which involves secondary prevention; and the post-illness / post-surgery stage, which involves tertiary prevention. Primary prevention involves efforts to prevent disease from occurring in the first place (such as health promotion), secondary prevention involves efforts to prevent the disease from becoming severe through early detection and treatment, and tertiary prevention aims at functional recovery and return to society and daily life through treatment and rehabilitation after the onset of the disease.

[0082] For input data indicating a pre-disease stage, it would be beneficial to output advice focusing on primary prevention, such as improving lifestyle and living environment, and providing health education. Broadly speaking, this aims to improve conditions that could potentially harm health, such as being easily fatigued or having shallow sleep. Examples of specific content are shown below. • Exercise instruction: Instruction on exercises that can be easily continued, such as walking, jogging, and stretching. • Dietary guidance: Balanced diet (maintaining nutritional balance and good health) • Sleep guidance: Promoting high-quality sleep and supporting fatigue recovery. • Stress Management: We provide relaxation techniques to reduce stress and advice to improve stress tolerance. • Posture guidance: Improve poor posture, such as slouching or poor standing posture, and prevent a decline in physical function. • Guidance on improving living environment: Supporting healthy living by improving the environment at work and home.

[0083] Whether or not a person is in the pre-disease stage can be determined by either medical interview data or measurement data. Suitable measurement data for the pre-disease stage includes, for example, basic patient data such as SMI (Skeletal Muscle Index), protein content, mineral content, body fat content (by body part), and body fat percentage, which can be obtained from diagnostic equipment, or data such as sleep quality, stress level, and daily step count, which can be obtained from smartwatches.

[0084] Advice for activities recommended in the pre-disease stage is viewed by both the practitioner and the patient via a device, and the patient is instructed to put it into practice. The effects of this include: • Extending healthy life expectancy: Supporting people in extending their healthy life expectancy so they can live healthier lives with less risk of illness. • Reducing healthcare costs: Reducing healthcare costs by preventing the onset of illness. • Improving Quality of Life (QOL): Leading a healthy life and improving quality of life through support. These are some of the things that can be expected.

[0085] For input data corresponding to the post-disease detection stage, it would be beneficial to output advice for secondary prevention, such as early detection and intervention of high-risk conditions for the disease, even if symptoms are not yet present, to prevent disease progression and worsening. The content of the response will partially overlap with the advice for primary prevention, but for people who are showing signs or symptoms, the output will be more accurate and tailored to those symptoms. Examples of symptoms that can be detected at the sign stage include high cholesterol, high blood pressure, and locomotive syndrome.

[0086] For input data corresponding to the post-illness and post-surgery stages, it would be desirable to output advice on tertiary prevention, such as improving the reduced physical strength, muscle strength, and range of motion of people who have already developed the disease and recovered through surgery or other treatment, as well as providing rehabilitation and lifestyle improvements to prevent recurrence. The content should aim at functional recovery, return to society and daily life, and prevention of recurrence. Examples include recovery from lifestyle-related diseases and orthopedic injuries, and post-operative rehabilitation for conditions such as disabilities (diabetes, heart disease, cancer, etc.).

[0087] It should be noted that primary, secondary, and tertiary prevention are not strictly distinguished, and multiple stages may overlap. Furthermore, the distinction between primary, secondary, and tertiary prevention does not prescribe a specific order. In practice, it is common for a disease to be discovered and treated, followed by tertiary prevention, and then primary prevention after recovery. The advice presented in this invention may also be given in accordance with such a progression.

[0088] The recommended advice presented as output data should preferably include not only the name of the relevant exercise, but also a well-written, easy-to-understand piece of advice. It is also acceptable to use a text generation function to construct sentences combining relevant words. For example, if the exercise measurement diagnosis, which is a concern at the pre-disease stage, indicates "chronic lower back pain / myofascial lower back pain," the output text should be something like, "If you continue like this, you may develop chronic lower back pain, myofascial lower back pain, etc., which may interfere with your work," to increase persuasiveness. If the recommended activity advice is "abdominal and back exercises, 5000 steps a day," the output text should be something like, "Plank, abdominal and back exercises, and walking about one train station's distance," to make it easier to understand. Furthermore, in the case of biochemical measurements, a format such as, "Similar exercises, along with attention to snacking and eating habits, are effective when uric acid or blood sugar levels are high. Specific goals should be decided in consultation with a physical therapist," would increase persuasiveness and sustainability through a cautionary warning. Alternatively, you can retrieve content that provides specific advice by referring to a separate text database based on the word.

[0089] An example of such advice is shown in Figure 6. Even with the same advice item, the trained model outputs more accurate advice based on the subject's measurement data and other personal data.

[0090] This advice should include specific activity values. The values ​​used as targets can be the self-assessment biometric data and lifestyle data mentioned above. Furthermore, if it is easy for the subject to visit a facility, data obtained from blood tests can be set as targets. Suitable measurable data to set as targets include the body composition, muscle mass, fat mass, and obesity values ​​mentioned above. The health management diagnostic support device 21 can acquire data from measuring devices 46, such as body fat scales and smart scales, via the subject's terminal 41. If the values ​​are easily measurable by the subject in their daily life, it reduces the burden and allows for individual verification. Additionally, if it is easy for the subject to visit a facility, setting data obtained from blood tests as targets enables highly accurate evaluation. Other options include setting specific exercise values ​​as targets, such as daily steps, distance traveled, or the number of repetitions of exercises like squats and push-ups.

[0091] Furthermore, the health management and diagnostic support device 21 should be able to output content that helps the subject achieve the above target values. This content includes not only the text itself, which is the advice, but also images, videos, and audio that clearly illustrate the content of the text. Even if the text in Figures 5 and 6 indicates "correct sitting posture," "bodyweight exercises," and "yoga," it may be difficult to understand exactly what to do from the text alone. By showing the subject content that visually and audibly illustrates in detail what kind of exercise to actually perform, the actual exercise becomes easier and more efficient. For this purpose, the health management and diagnostic support device 21 should have a content database 51 that records and allows retrieval of the above-mentioned images, videos, audio, etc. In addition to recording the content itself, it may also be recorded in the form of link addresses to the content or prompts for generating the content.

[0092] The aforementioned content is preferably not only individual text and images, but also content that interactively displays the progress of the subject's latest data toward the target value to terminals 40, such as the subject's terminal 41 and the client's terminal 42. In other words, it is preferable that the health management diagnostic support device 21 has a content distribution means 37 that interactively displays the content to terminals 40 for this purpose.

[0093] An example of such an interactive display is shown in Figure 7. Data that can be measured by the measuring device 46, such as weight and body fat percentage, are set as target values ​​and displayed. The measurement data measured by the measuring device 46 is recorded in the subject database 26, and the items for which a target value has been set from this measurement data are displayed as a graph over time. Furthermore, a panel is displayed that visually shows how much progress has been made in the process from the initial value to the target value. For example, suppose the target value item is weight, the initial weight is 85 kg, and the target weight (target value) is 70 kg. If the current weight has decreased to 77 kg, then 50% of the process to reach the target has been achieved. This is shown not only numerically, but also visually by half of the panel opening.

[0094] When the data acquired by the measuring instrument 46 is registered in the subject database 26 by the health management and diagnostic support device 21, and the content distribution means 37 is instructed to quickly update the display on the subject terminal 41, the subject can immediately feel a sense of accomplishment during the measurement.

[0095] Furthermore, the content distribution means 37 of the health management and diagnostic support device 21 should transmit the above-mentioned content not only to the subject terminal 41 but also to the requester terminal 42. This allows the requester, who wishes for their family member to live a healthy life, to know to what extent the subject is actually exercising and engaging in activities to improve their health in order to reach their target values, and to feel a sense of real reassurance.

[0096] Furthermore, for each individual, it is desirable that advice on recommended activities progresses along with the improvement of the diagnostic results. For example, for an individual who drinks alcohol and whose physical strength has deteriorated to the point where they cannot go up and down stairs, the output data for secondary prevention advice would be: "Avoid drinking alcohol after training and maintain a moderate amount of alcohol. Strengthen your lower limb muscles with squats to prevent the progression of symptoms. If symptoms progress, you may need artificial joint surgery. Continue step-up exercises with a goal of 10 repetitions. Start with twice a day." If, after three months, the individual stops drinking alcohol, various measurement data improves, and they become able to go up and down stairs, they move on to the post-illness / post-surgery stage. The trained model would then output data for tertiary prevention advice such as: "Switch to a high-protein diet to promote muscle synthesis. Aim to walk about 4,000 steps a day. Stretch your calves to make your knees more flexible," and the advice would change with each stage. [Explanation of symbols]

[0097] 11. Learning devices 12 Learning Servers 13 External Servers 21. Health management and diagnostic support device 22 Pre-trained models 23. Auxiliary Estimation Means 25 User Databases 26 Target Database 27. Client Database 28. User Database 29 packages 31. Data reception means 32 User Management Methods 33. Means of checking history 34 Self-examination tool 35 Authentication methods 41 Target user's device 42 Client terminal 43 Facility Terminals 44. Practitioner terminal 45 Administrator terminal 46 Measuring Instruments

Claims

1. The system has a pre-trained model created using personal data, including self-assessment data of the subject, as input data, and a set of training data, including diagnostic results from the practitioner for the same subject, as output data. A health management diagnostic support device that uses the aforementioned trained model to output the estimated diagnostic result from the personal data of one of the aforementioned target individuals, A health management diagnostic support device that includes advice for recommended activities based on the aforementioned predicted diagnostic results.

2. We have a database of subjects that contains identifiable personal information about each subject, The aforementioned personal information includes the usage history of the health management and diagnostic support device and the target values ​​included in the aforementioned recommended advice. The health management diagnostic support device according to claim 1, which is capable of outputting the user history of the subject and content for achieving the target value.

3. The health management diagnostic support device according to claim 2, wherein the content includes images, videos, or both, showing the activities included in the advice.

4. The health management diagnostic support device according to claim 2, wherein the content includes content that interactively displays the progress toward the target value on the terminal.

5. The system has a user database that records the aforementioned individuals in an identifiable manner, The self-assessment data of the aforementioned subject can be entered by a client other than the subject. A health management diagnostic support device according to any one of claims 1 to 4, which outputs the estimated diagnostic results for the subject.

6. A health management and diagnostic support device according to any one of claims 1 to 4, A health management and diagnostic support system comprising a terminal that outputs the aforementioned diagnostic results.

7. Using a trained model created with personal data including self-examination data of the subject as input data and training data including diagnostic results by a practitioner for the subject as output data, the model outputs the predicted diagnostic result from the personal data of one of the aforementioned subjects. A health management diagnostic support method that includes advice for recommended activities based on the aforementioned inferred diagnostic results.

Images