Method and system for predicting thyroid dysfunction of subject

Abstract

To provide a method for analyzing a patient's heart rate and predicting thyroid dysfunction on the basis of a result of the analysis, an apparatus, and a system.SOLUTION: The method includes the steps of: determining target data on the basis of a trigger signal; acquiring interval heart rates corresponding to a determined target date; acquiring, for a subject, a first pre-processing result for the acquired interval heart rates corresponding to the determined target date; acquiring, for the subject, at least one of concentration of hormone related to a thyroid corresponding to a reference date; acquiring, for the subject, a second pre-processing result for the interval heart rates corresponding to the reference date; acquiring a difference of the first pre-processing result with respect to the second pre-processing result; and acquiring a prediction result for thyroid dysfunction acquired on the basis of values including the at least one of concentration of hormone related to the thyroid corresponding to the reference date and the difference.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present application relates to a method for analyzing the heart rate of a patient and for predicting thyroid dysfunction based on the results of the analysis, and to an apparatus and a system for carrying out the method. [Background technology]

[0002] The thyroid gland is an endocrine gland located in the front of the neck that produces thyroid hormones in response to signals from the thyroid-stimulating hormone secreted by the pituitary gland in the brain.

[0003] Thyroid hormones are produced by the thyroid gland, which controls the metabolic rate of the human body. When more than normal levels of thyroid hormones are secreted, the body's metabolism proceeds abnormally fast, resulting in symptoms such as feeling hotter than normal or having a racing heart. Conversely, when less than normal levels of thyroid hormones are secreted, symptoms such as feeling very cold, feeling tired and depressed, or having a slow pulse may occur.

[0004] When the thyroid gland is functioning normally, the secretion of thyroid hormones occurs without any problems. However, when the thyroid gland is functioning abnormally, the secretion of thyroid hormones decreases or increases. The set of diseases caused by decreased secretion of thyroid hormones is called hypothyroidism, and the set of diseases caused by increased secretion of thyroid hormones is called hyperthyroidism. Furthermore, these two diseases are called dysthyroidism.

[0005] Currently, the only diagnostic method for testing for thyroid dysfunction is to test the concentration of thyroid hormone in the patient's blood and determine whether the concentration of thyroid hormone is within the normal range, at high levels, or at low levels from a clinical pathology standpoint.

[0006] Meanwhile, there are several methods for treating hyperthyroidism and hypothyroidism, but the most widely used treatment method is drug treatment method.Most patients with hyperthyroidism and most patients with hypothyroidism are administered drugs for a predetermined period of time to restore normal levels of thyroid hormone production and secretion function.However, hyperthyroidism and hypothyroidism are diseases with low cure rate and very high recurrence rate, so it is necessary to continuously monitor whether hyperthyroidism or hypothyroidism recurs.

[0007] However, as described above, the only diagnostic method for testing thyroid function is to test the hormone concentration through blood sampling, so there is a problem that patients need to visit the hospital regularly to monitor whether their thyroid gland is functioning abnormally.

[0008] Therefore, there is a need to develop technology to more accurately determine whether the thyroid gland is malfunctioning other than by testing hormone levels through a blood draw.

[0009] That is, it is desirable to develop a method that allows patients to monitor thyroid dysfunction more simply and quickly without having to visit a clinic in person, and that encourages patients to visit a clinic when necessary. Summary of the Invention [Problem to be solved by the invention]

[0010] The problem solved by the present disclosure is to provide a predictive model capable of predicting thyroid dysfunction based on heart rate information obtained using personal electronic devices available to the general public, rather than specialized medical diagnostic devices.

[0011] Another problem solved by the subject matter disclosed herein is to provide a method for training the above-mentioned predictive model capable of predicting thyroid dysfunction based on heart rate information.

[0012] Another problem solved by the present disclosure is to provide a method and system that allows the general public to continuously monitor for thyroid dysfunction by using the above described predictive model that is capable of predicting thyroid dysfunction based on heart rate information without the help of a doctor and without a direct hospital visit.

[0013] The technical problems to be solved by the present application are not limited to the above-mentioned technical problems, and other technical problems not mentioned will be clearly understood by those skilled in the art from this specification and the accompanying drawings. [Means for solving the problem]

[0014] According to an embodiment of the present application, a method for predicting thyroid dysfunction in a subject is disclosed. A method for predicting thyroid dysfunction in a subject comprises acquiring a trigger signal; determining a target date based on the acquired trigger signal; acquiring interval heart rates corresponding to the determined target date; acquiring a first pre-processing result of the acquired interval heart rates corresponding to the determined target date for the subject, where the first pre-processing result includes at least one parameter related to a mean and a variance of the interval heart rates corresponding to the target date; acquiring at least one of a concentration of a thyroid related hormone corresponding to a reference date for the subject; acquiring a second pre-processing result of the interval heart rates corresponding to the reference date for the subject, where the second pre-processing result includes at least one parameter related to a mean and a variance of the interval heart rates corresponding to the reference date; acquiring a difference of the first pre-processing result relative to the second pre-processing result; and acquiring a prediction result for thyroid dysfunction based on a value including the difference and the at least one of the concentrations of the thyroid related hormone corresponding to the reference date, where the prediction result for thyroid dysfunction may be a result of processing an input value by a thyroid dysfunction prediction model, where the input value includes at least one of the difference and the concentration of the thyroid related hormone corresponding to the reference date.

[0015] In some embodiments disclosed herein, the at least one parameter related to the variance of the interval heart rates corresponding to the target day may include a standard deviation, a skewness, and a kurtosis of the interval heart rates corresponding to the target day, and wherein the at least one parameter related to the variance of the interval heart rates corresponding to the reference day may include a standard deviation, a skewness, and a kurtosis of the interval heart rates corresponding to the reference day.

[0016] In some embodiments disclosed by the present application, the difference of the first pre-processing result from the second pre-processing result may be one or a combination selected from the group consisting of: 1) the change in the average interval heart rate for the target day from the reference date; 2) the rate of change in the average interval heart rate for the target day from the reference date; 3) the change in the standard deviation for the target day from the reference date; 4) the change in the relative standard deviation for the target day from the reference date; 5) the change in skewness for the target day from the reference date; 6) the change in kurtosis for the target day from the reference date; and 7) the JS divergence between the interval heart rate corresponding to the reference date and the interval heart rate corresponding to the target day.

[0017] In some embodiments disclosed herein, at least one of the hormone concentrations corresponding to the reference day may be one or a combination selected from the group consisting of: 1) thyroid stimulating hormone (TSH) concentration, 2) tetraiodothyronine (T4) concentration, 3) serum free T4 concentration, 4) triiodothyronine (T3) concentration, 5) serum free T3 concentration, and 6) thyrotropin releasing hormone (TRH) concentration.

[0018] In some embodiments disclosed herein, the method further comprises: (1) a change in the average interval heart rate for the subject day relative to the reference date; 2) a rate of change in the average interval heart rate for the subject day relative to the reference date; 3) a change in the standard deviation for the subject day relative to the reference date; 4) a change in the relative standard deviation for the subject day relative to the reference date; 5) a change in skewness for the subject day relative to the reference date; 6) a change in kurtosis for the subject day relative to the reference date; 7) a JS divergence between the interval heart rate corresponding to the reference date and the interval heart rate corresponding to the subject day; 8) a thyroid stimulating hormone (TSH) concentration for the subject corresponding to the reference date; 9) a tetraiodothyronine (T4) concentration for the subject corresponding to the reference date; 10) a free T4 concentration in the subject's serum corresponding to the reference date; 11) a triiodothyronine (T3) concentration for the subject corresponding to the reference date; 12) a free T3 concentration in the subject's serum corresponding to the reference date; T3) concentration of thyrotropin releasing hormone (TRH) of the subject corresponding to the reference date; and obtaining a first value and a second value from the group consisting of: i) a value obtained by multiplying the first value by the second value; ii) a value obtained by dividing the first value by the second value; and iii) a value obtained by dividing the second value by the first value.

[0019] Here, the step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date and a difference may be a step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date, the difference, and any one of the obtained values, wherein the prediction result for thyroid dysfunction may be a result of processing input values ​​by a thyroid dysfunction prediction model, wherein the input values ​​include at least one of the difference, any one of the obtained values, and the hormone concentration corresponding to the reference date.

[0020] In some embodiments disclosed herein, the method may further comprise obtaining a day interval between the target date and the reference date.

[0021] Here, the step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date and a difference may be a step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date, the difference, and a number of days between the target date and the reference date, where the prediction result for thyroid dysfunction may be a result of processing input values ​​by a thyroid dysfunction prediction model, where the input values ​​include at least one of the difference, the number of days between the target date and the reference date, and the hormone concentration corresponding to the reference date.

[0022] In some embodiments disclosed herein, the interval heart rates corresponding to the target day may be all resting heart rates corresponding to a predetermined period based on the target day, and the interval heart rates corresponding to the reference day may be all resting heart rates corresponding to a predetermined period based on the reference day.

[0023] Here, the predetermined period may be any one selected from 8, 9, 10, 11, 12, 13, 14, 15, 16 and 17 days.

[0024] In some embodiments disclosed herein, the thyroid dysfunction prediction model may include a hyperthyroidism prediction model and a hypothyroidism prediction model.

[0025] Here, a prediction result for thyroid dysfunction can be determined by considering a first prediction result in which input values ​​including at least one hormone concentration and a difference are processed by a hyperthyroidism prediction model, and a second prediction result in which input values ​​including at least one hormone concentration and a difference are processed by a hypothyroidism prediction model. Beneficial Effects

[0026] The disclosure in this application provides a predictive model capable of predicting thyroid dysfunction based on heart rate information obtained using personal electronic devices that can be used by the general public rather than specialized medical diagnostic devices.

[0027] According to the present disclosure in the present application, a method for training the above-mentioned prediction model capable of predicting thyroid dysfunction based on heart rate information is provided.

[0028] According to the present disclosure in this application, a method and system are provided that allows the general public to continuously monitor for thyroid dysfunction by using the above-mentioned predictive model that can predict thyroid dysfunction based on heart rate information without the help of a doctor and without an in-person hospital visit. [Brief description of the drawings]

[0029] [Figure 1] FIG. 1 illustrates a system for predicting thyroid dysfunction according to an embodiment described herein.

[0030] [Diagram 2] FIG. 1 is a block diagram illustrating a heart rate measuring device as described herein.

[0031] [Diagram 3] FIG. 2 is a block diagram illustrating a user terminal as described herein.

[0032] [Figure 4] FIG. 2 is a block diagram illustrating a server as described herein.

[0033] [Diagram 5] 1 is a flow chart illustrating the method for predicting thyroid dysfunction described herein.

[0034] [Figure 6] 1 is a flow chart illustrating the method for predicting thyroid dysfunction described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0035] The above-mentioned objects, features, and advantages of the present application will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Furthermore, various modifications may be made to the present application, and various embodiments of the present application may be practiced. Accordingly, the following detailed description of specific embodiments will be given with reference to the accompanying drawings.

[0036] Throughout the specification, the same reference numerals refer to the same elements in principle. Furthermore, elements having the same functions within the same scope shown in the drawings of the embodiment are described using the same reference numerals, and redundant descriptions are omitted.

[0037] Detailed descriptions of well-known functions or configurations of the present application will be omitted if they are deemed to obscure the nature and spirit of the present application. Furthermore, throughout this specification, the terms first, second, etc. are used only to distinguish one element from another.

[0038] Furthermore, the terms "module" and "section" used to list elements in the following description are used solely for ease of writing the specification, and are not intended to have different special meanings or functions, and therefore may be used individually or interchangeably.

[0039] In the following embodiments, expressions used in the singular form also include expressions in the plural form unless it has a clearly different meaning in the context.

[0040] In the following embodiments, it should be understood that terms such as "comprise", "have", and the like are intended to indicate the presence of features or elements disclosed in the specification, and are not intended to exclude the possibility that one or more other features or elements may be added.

[0041] The size of elements in the drawings may be exaggerated or reduced for convenience of explanation. For example, any size and thickness of each element shown in the drawings is shown for convenience of explanation, and the present disclosure is not limited thereto.

[0042] In cases where particular embodiments are otherwise implemented, certain processes may be performed out of the order described. For example, two processes described as successive may be performed substantially simultaneously or may proceed in the reverse order from that described.

[0043] In the following embodiments, when elements are referred to as being connected to each other, the elements are directly connected to each other, or the elements are indirectly connected to each other with an intervening element between them. For example, in this specification, when elements are referred to as being electrically connected to each other, the elements are directly electrically connected to each other, or the elements are indirectly electrically connected to each other with an intervening element between them.

[0044] According to an embodiment of the present application, a method for predicting thyroid dysfunction in a subject is disclosed. A method for predicting thyroid dysfunction in a subject comprises acquiring a trigger signal; determining a target date based on the acquired trigger signal; acquiring interval heart rates corresponding to the determined target date; acquiring a first pre-processing result of the acquired interval heart rates corresponding to the determined target date for the subject, where the first pre-processing result includes at least one parameter related to a mean and a variance of the interval heart rates corresponding to the target date; acquiring at least one of a concentration of a thyroid related hormone corresponding to a reference date for the subject; acquiring a second pre-processing result of the interval heart rates corresponding to the reference date for the subject, where the second pre-processing result includes at least one parameter related to a mean and a variance of the interval heart rates corresponding to the reference date; acquiring a difference of the first pre-processing result relative to the second pre-processing result; and acquiring a prediction result for thyroid dysfunction based on a value including the difference and the at least one of the concentrations of the thyroid related hormone corresponding to the reference date, where the prediction result for thyroid dysfunction may be a result of processing an input value by a thyroid dysfunction prediction model, where the input value includes at least one of the difference and the concentration of the thyroid related hormone corresponding to the reference date.

[0045] In some embodiments disclosed herein, the at least one parameter related to the variance of the interval heart rates corresponding to the target day may include a standard deviation, a skewness, and a kurtosis of the interval heart rates corresponding to the target day, and the at least one parameter related to the variance of the interval heart rates corresponding to the reference day may include a standard deviation, a skewness, and a kurtosis of the interval heart rates corresponding to the reference day.

[0046] In some embodiments disclosed by the present application, the difference of the first pre-processing result from the second pre-processing result may be one or a combination selected from the group consisting of: 1) the change in the average interval heart rate for the target day from the reference date; 2) the rate of change in the average interval heart rate for the target day from the reference date; 3) the change in the standard deviation for the target day from the reference date; 4) the change in the relative standard deviation for the target day from the reference date; 5) the change in skewness for the target day from the reference date; 6) the change in kurtosis for the target day from the reference date; and 7) the JS divergence between the interval heart rate corresponding to the reference date and the interval heart rate corresponding to the target day.

[0047] In some embodiments disclosed herein, at least one of the hormone concentrations corresponding to the reference day may be one or a combination selected from the group consisting of: 1) thyroid stimulating hormone (TSH) concentration, 2) tetraiodothyronine (T4) concentration, 3) serum free T4 concentration, 4) triiodothyronine (T3) concentration, 5) serum free T3 concentration, and 6) thyrotropin releasing hormone (TRH) concentration.

[0048] In some embodiments disclosed herein, the method further comprises: (1) a change in the average interval heart rate for the subject day relative to the reference date; 2) a rate of change in the average interval heart rate for the subject day relative to the reference date; 3) a change in the standard deviation for the subject day relative to the reference date; 4) a change in the relative standard deviation for the subject day relative to the reference date; 5) a change in skewness for the subject day relative to the reference date; 6) a change in kurtosis for the subject day relative to the reference date; 7) a JS divergence between the interval heart rate corresponding to the reference date and the interval heart rate corresponding to the subject day; 8) a thyroid stimulating hormone (TSH) concentration for the subject corresponding to the reference date; 9) a tetraiodothyronine (T4) concentration for the subject corresponding to the reference date; 10) a free T4 concentration in the subject's serum corresponding to the reference date; 11) a triiodothyronine (T3) concentration for the subject corresponding to the reference date; 12) a free T3 concentration in the subject's serum corresponding to the reference date; T3) concentration of thyrotropin releasing hormone (TRH) of the subject corresponding to the reference date; and obtaining a first value and a second value from the group consisting of: i) a value obtained by multiplying the first value by the second value; ii) a value obtained by dividing the first value by the second value; and iii) a value obtained by dividing the second value by the first value.

[0049] In this specification, the step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date and a difference may be a step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date, the difference, and any one of the obtained values, and the prediction result for thyroid dysfunction may be a result of processing input values ​​by a thyroid dysfunction prediction model, where the input value includes at least one of the difference, any one of the obtained values, and the hormone concentration corresponding to the reference date.

[0050] In some embodiments disclosed herein, the method may further comprise obtaining a day interval between the target date and the reference date.

[0051] Here, the step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date and a difference may be a step of obtaining a prediction result for thyroid dysfunction based on a value including at least one of the hormone concentrations corresponding to the reference date, the difference, and a number of days between the target date and the reference date, and the prediction result for thyroid dysfunction may be a result of processing input values ​​by a thyroid dysfunction prediction model, where the input values ​​include at least one of the difference, the number of days between the target date and the reference date, and the hormone concentration corresponding to the reference date.

[0052] In some embodiments disclosed herein, the interval heart rates corresponding to the target day may be all resting heart rates corresponding to a predetermined period based on the target day, and the interval heart rates corresponding to the reference day may be all resting heart rates corresponding to a predetermined period based on the reference day.

[0053] Here, the predetermined period may be any one selected from 8, 9, 10, 11, 12, 13, 14, 15, 16 and 17 days.

[0054] In some embodiments disclosed herein, the thyroid dysfunction prediction model may include a hyperthyroidism prediction model and a hypothyroidism prediction model.

[0055] Here, a prediction result for thyroid dysfunction can be determined by considering a first prediction result in which input values ​​including at least one hormone concentration and a difference are processed by a hyperthyroidism prediction model, and a second prediction result in which input values ​​including at least one hormone concentration and a difference are processed by a hypothyroidism prediction model.

[0056] 1. Definitions

[0057] Heart rate

[0058] As used herein, the term "heart rate" refers to the number of beats per unit of time. For example, heart rate may be measured as beats per minute (bpm). However, in this specification, unless otherwise specified or the term is used for purposes of general scientific description, "heart rate" is to be interpreted as corresponding to or relating to the time at which the heart rate is measured (estimated time).

[0059] Rest period

[0060] As used herein, the term "resting period" refers to a period during which a person's physical movement is below a particular level. Generally, a "resting period" refers to a period during which a subject is awake and motionless, but the term "resting period" as used herein also refers to a period during which a subject is asleep.

[0061] Resting heart rate

[0062] As used herein, the term "resting heart rate" refers to a heart rate where the person is within a resting period at the time the heart rate is measured.

[0063] In particular, the expression "resting heart rate corresponding to day X" in this specification may refer to all heart rates corresponding to the resting period included in the period from 12:00:00 AM on day X to 11:59:59 PM on day X.

[0064] However, considering that i) a resting period generally includes a person's sleeping time, and ii) taking into account a person's sleeping habits, many people start sleeping before 12:00:00 a.m., the "resting heart rate corresponding to day X" in this specification may include a resting heart rate corresponding to a period from a time when a person starts sleeping on a day immediately before day X (hereinafter referred to as day X-1) to 11:59:59 p.m. on day X-1. Similarly, the "resting heart rate corresponding to day X" may not include a resting heart rate corresponding to a period from a time when a person starts sleeping on day X to 11:59:59 p.m. on day X. That is, a resting heart rate corresponding to a period from a time when a person starts sleeping on day X to 11:59:59 p.m. on day X may be included in the "resting heart rate corresponding to a day after day X (hereinafter referred to as day X+1)" described above.

[0065] Interval heart rate (interval heart rate)

[0066] In this specification, the term "interval heart rate (interval heart rate)" refers to all resting heart rates corresponding to a period of two or more days. In this specification, the terms "interval resting heart rate (interval resting heart rate)" and "interval heart rate" can be used interchangeably.

[0067] Average heart rate over a period

[0068] As used herein, the term "interval average heart rate (period average heart rate)" refers to the average value of resting heart rates (i.e., interval heart rate) measured over a period of two or more days. For example, a 5-day period average heart rate refers to the average value of the resting heart rate corresponding to the first day, the resting heart rate corresponding to the second day, the resting heart rate corresponding to the third day, the resting heart rate corresponding to the fourth day, and the resting heart rate corresponding to the fifth day. As used herein, the terms "average interval heart rate (average period heart rate)" and "period average heart rate" may be used interchangeably.

[0069] Thyroid hormones

[0070] As used herein, the term "thyroid hormone" is intended to include thyroxine (T4) or tetraiodothyronine, which are hormones secreted by the thyroid gland; serum free T4, which is free T4 in serum; triiodothyronine (T3), which is the result of conversion of T4 to remove iodine; serum free T3, which is free T3 in serum; thyroid stimulating hormone (TSH), which is secreted by the pituitary gland; and thyrotropin releasing hormone (TRH).

[0071] However, in this specification, the terms "hormone" and "thyroid hormone" may be used interchangeably.

[0072] hyperthyroidism

[0073] As used herein, the term "hyperthyroidism" refers to a medical condition in which the concentration of thyroid hormones in the blood is excessively increased due to excessive production or secretion of hormones secreted by the thyroid gland, or due to other factors, and all resulting clinical pathological symptoms.

[0074] hypothyroidism

[0075] As used herein, the term "hypothyroidism" refers to a medical condition in which the concentration of thyroid hormone in the blood is excessively decreased due to too little production or secretion of hormone by the thyroid gland or due to other factors, and all the resulting clinical pathological symptoms.

[0076] Thyroid dysfunction

[0077] As used herein, the term "thyroid dysfunction" is intended to include both "hypothyroidism" and "hyperthyroidism" as described above.

[0078] 2. Training method

[0079] overview

[0080] In 2017, in order to solve the problems in the related art described above and to develop a technology that meets the needs of the market, the applicant of the present application completed the development of a method and system technology and filed a related patent application. The method and system are for determining whether a patient's current state is a euthyroid state or a dysthyroid state (i.e., hyperthyroidism or hypothyroidism) by using a heart rate corresponding to a time when the patient's thyroid function was determined to be "normal" and a current heart rate as main input variables based on the correlation between heart rate and hyperthyroidism.

[0081] However, the technology developed by the applicant of the present application had several problems.

[0082] First, because the heart rate associated with normal thyroid function (average heart rate over a predetermined period of time) varies from person to person, it has been difficult to derive a function that outputs whether the thyroid is malfunctioning by simply using the heart rate in a normal state and the heart rate in a current state as factors.

[0083] Second, because heart rate in the "normal" state was used as a factor, the techniques described above could not be applied when the hormone levels were determined to be "hyperthyroid" or "hypothyroid" rather than "normal" on the day of testing.

[0084] Due to these challenges described above, previously developed technologies have been inaccurate, inconvenient to use, and therefore inadequate to meet the needs of a market that demands continuous monitoring.

[0085] To address these challenges, the applicant proposes a machine learning-based thyroid dysfunction prediction model, which is described herein.

[0086] In addition, in training a prediction model for predicting thyroid dysfunction, there are several challenges in creating a personalized prediction model, so the applicant analyzes the attributes of the data for training the prediction model, and trains the model in various ways for the purpose of increasing the accuracy of prediction, and thus finds several meaningful training methods for increasing the accuracy of the prediction model. The training methods for increasing accuracy are disclosed herein.

[0087] Clinical data collection

[0088] First, we describe how to collect the basic clinical data for training the predictive models described in this application.

[0089] (1) Select patients suffering from hyperthyroidism or hypothyroidism.

[0090] (2) Obtain personal information of the selected patient. For example, personal information of the selected patient including the name, sex, age, weight, height, etc. of the patient is obtained.

[0091] (3) Providing each of the selected patients with a wearable device that can be worn on their wrist and has the capability of measuring the patient's heart rate.

[0092] (4) For each of the selected patients, when the selected patient visits the hospital to undergo a thyroid hormone test, obtain the interval heart rate (period heart rate) corresponding to the test day among the heart rates measured by the wearable device provided to the patient.

[0093] (5) For each of the selected patients, obtain thyroid hormone concentration values ​​on the day of the study.

[0094] (6) For each of the selected patients, obtain a physician's diagnosis for the patient based on the thyroid hormone test results, i.e., obtain the thyroid hormone test results and information on whether the patient is in a hyperthyroid, hypothyroid, or normal state.

[0095] As a result, if one patient is tested once for thyroid hormone testing, one data set may be obtained that includes: i) the patient's personal information, ii) the test date, iii) the interval heart rate corresponding to the test date, iv) the thyroid hormone concentration from the test performed on the test date, and v) the doctor's diagnosis of thyroid dysfunction from the result of the test performed on the test date.

[0096] That is, if a dataset with an average of four hormone test results per patient is obtained from 100 patients, a total of 400 datasets will be obtained, and if a dataset with an average of three hormone test results per patient is obtained from 300 patients, a total of 900 datasets will be obtained.

[0097] Patients suffering from thyroid dysfunction are recommended to visit the hospital once every three months to have blood drawn for hormone level testing. Therefore, in order to obtain a sufficient data set in the manner described above, it is basically necessary to increase the number of patients targeted for data set collection or to extend the period during which data sets per patient are obtained to several years or longer. That is, in the current situation, obtaining a large amount of clinical data related to thyroid dysfunction requires a huge investment of money or a huge investment of time.

[0098] Training Method Embodiment #1. Difference Values ​​from Heart Rate on Reference Day and Hormone Values ​​on Reference Day

[0099] According to a first exemplary embodiment described in the present application, to train the learning model, a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day and the hormone concentration on the second test day may be used as input values, and the diagnostic result corresponding to the first test day may be used as a labeling value.

[0100] Here, as the learning model, a classification model including a light gradient boosting machine, a support vector machine, a random forest, an extra tree, an adaboost, an extreme gradient boosting, a catboost, etc. may be used.

[0101] The learning models include a hyperthyroidism prediction model for predicting hyperthyroidism and a hypothyroidism prediction model for predicting hypothyroidism.

[0102] According to a first exemplary embodiment, when the hyperthyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed at least once with hyperthyroidism as a diagnostic result can be used as training data.

[0103] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hyperthyroidism as a diagnostic result may be referred to as a hyperthyroidism patient.

[0104] On the other hand, in calculating the period average heart rate, a resting heart rate over a predetermined period of time that is predetermined based on a particular date may be used, where the predetermined period of time may be one of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 days.

[0105] According to a first exemplary embodiment, to train a hyperthyroidism prediction model, clinical data of hyperthyroid patients may be used to generate a training dataset in a format such as [(period average heart rate corresponding to the first test date - period average heart rate corresponding to the second test date), thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hyperthyroidism or non-hyperthyroidism)].

[0106] For example, assuming that a total of three datasets are obtained from a first hyperthyroid patient, and the diagnosis for the first test is "hyperthyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that may be obtained from the clinical data of the first hyperthyroid patient may be generated as follows:

[0107] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), second concentration, hyperthyroidism]

[0108] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), third concentration, hyperthyroidism]

[0109] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), first concentration, non-hyperthyroidism]

[0110] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), third concentration, non-hyperthyroidism]

[0111] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), first concentration, non-hyperthyroidism], and

[0112] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), second concentration, non-hyperthyroidism]

[0113] As another example, assuming a total of four datasets are obtained from a second hyperthyroid patient, and the diagnosis for the first test is "hyperthyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hyperthyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that may be obtained from the clinical data of the second hyperthyroid patient may be generated as follows:

[0114] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), second concentration, hyperthyroidism]

[0115] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), third concentration, hyperthyroidism]

[0116] [(average heart rate for the period corresponding to the 1st test day - average heart rate for the period corresponding to the 4th test day), 4th concentration, hyperthyroidism]

[0117] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), first concentration, non-hyperthyroidism]

[0118] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), third concentration, non-hyperthyroidism]

[0119] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), fourth concentration, non-hyperthyroidism]

[0120] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), first concentration, hyperthyroidism]

[0121] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), second concentration, hyperthyroidism]

[0122] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the fourth test day), fourth concentration, hyperthyroidism]

[0123] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), 1st concentration, non-hyperthyroidism]

[0124] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), 2nd concentration, non-hyperthyroidism], and

[0125] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), 3rd concentration, non-hyperthyroidism]

[0126] According to a first exemplary embodiment, when the hypothyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed at least once with hypothyroidism as a diagnostic result can be used as training data.

[0127] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hypothyroidism as a diagnostic result may be referred to as a "hypothyroid patient."

[0128] According to a first exemplary embodiment, to train a hypothyroidism prediction model, clinical data of hypothyroid patients may be used to generate a training dataset in a format such as [(period average heart rate corresponding to the first test date - period average heart rate corresponding to the second test date), thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hypothyroidism or non-hypothyroidism)].

[0129] For example, assuming that a total of three datasets are obtained from a first hypothyroid patient, and the diagnosis for the first test is "hypothyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that may be obtained from the clinical data of the first hypothyroid patient may be generated as follows:

[0130] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), second concentration, hypothyroidism]

[0131] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), third concentration, hypothyroidism]

[0132] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), first concentration, non-hypothyroidism]

[0133] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), third concentration, non-hypothyroidism]

[0134] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), first concentration, non-hypothyroid], and

[0135] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), second concentration, non-hypothyroidism]

[0136] As another example, assuming a total of four datasets are obtained from a second hypothyroid patient, and the diagnosis for the first test is "hypothyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hypothyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that may be obtained from the clinical data of the second hypothyroid patient may be generated as follows:

[0137] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), second concentration, hypothyroidism]

[0138] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), third concentration, hypothyroidism]

[0139] [(average heart rate for the period corresponding to the 1st test day - average heart rate for the period corresponding to the 4th test day), 4th concentration, hypothyroidism]

[0140] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), first concentration, non-hypothyroidism]

[0141] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), third concentration, non-hypothyroidism]

[0142] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), fourth concentration, non-hypothyroidism]

[0143] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the first test day), first concentration, hypothyroidism]

[0144] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), second concentration, hypothyroidism]

[0145] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the fourth test day), fourth concentration, hypothyroidism]

[0146] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), 1st concentration, non-hypothyroidism]

[0147] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), 2nd concentration, non-hypothyroid], and

[0148] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), 3rd concentration, non-hypothyroidism]

[0149] In a first exemplary embodiment, the hormone concentration may be one or a combination selected from the group consisting of T4 concentration, free T4 concentration, T3 concentration, free T3 concentration, TSH concentration, and thyrotropin releasing hormone (TRH) concentration.

[0150] Training method embodiment #2. Difference value from the heart rate on the reference day, average heart rate for the period of the target day, and hormone value on the reference day

[0151] According to a second exemplary embodiment described in the present application, to train the learning model, a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day, the hormone concentration on the second test day, and the period average heart rate corresponding to the first test day may be used as input values, and the diagnostic result corresponding to the first test day may be used as a labeling value.

[0152] Here, as the learning model, a classification model including a light gradient boosting machine, a support vector machine, a random forest, an extra tree, an adaboost, an extreme gradient boosting, a catboost, etc. may be used.

[0153] The learning models include a hyperthyroidism prediction model for predicting hyperthyroidism and a hypothyroidism prediction model for predicting hypothyroidism.

[0154] According to a second exemplary embodiment, when the hyperthyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed at least once with hyperthyroidism as a diagnostic result can be used as training data.

[0155] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hyperthyroidism as a diagnostic result may be referred to as a hyperthyroidism patient.

[0156] On the other hand, in calculating the period average heart rate, a resting heart rate over a predetermined period of time that is predetermined based on a particular date may be used, where the predetermined period of time may be one of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 days.

[0157] According to a second exemplary embodiment, to train a hyperthyroidism prediction model, clinical data of hyperthyroid patients may be used to generate a training dataset in a format such as [(period average heart rate corresponding to the first test date - period average heart rate corresponding to the second test date), period average heart rate corresponding to the first test date, thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hyperthyroidism or non-hyperthyroidism)].

[0158] For example, assuming that a total of three datasets are obtained from a first hyperthyroid patient, and the diagnosis for the first test is "hyperthyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that may be obtained from the clinical data of the first hyperthyroid patient may be generated as follows:

[0159] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the first test day, second concentration, hyperthyroidism]

[0160] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the first test day, third concentration, hyperthyroidism]

[0161] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the second test day, first concentration, non-hyperthyroidism]

[0162] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the second test day, third concentration, non-hyperthyroidism]

[0163] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the third test day, first concentration, non-hyperthyroidism], and

[0164] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the third test day, second concentration, non-hyperthyroidism]

[0165] As another example, assuming a total of four datasets are obtained from a second hyperthyroid patient, and the diagnosis for the first test is "hyperthyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hyperthyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that may be obtained from the clinical data of the second hyperthyroid patient may be generated as follows:

[0166] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the first test day, second concentration, hyperthyroidism]

[0167] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the first test day, third concentration, hyperthyroidism]

[0168] [(average heart rate for the period corresponding to the 1st test day - average heart rate for the period corresponding to the 4th test day), average heart rate for the period corresponding to the 1st test day, 4th concentration, hyperthyroidism]

[0169] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the second test day, first concentration, non-hyperthyroidism]

[0170] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the second test day, third concentration, non-hyperthyroidism]

[0171] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), average heart rate for the period corresponding to the second test day, fourth concentration, non-hyperthyroidism]

[0172] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the third test day, first concentration, hyperthyroidism]

[0173] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), average heart rate during the period corresponding to the third test day, second concentration, hyperthyroidism]

[0174] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the fourth test day), average heart rate for the period corresponding to the third test day, fourth concentration, hyperthyroidism]

[0175] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), average heart rate for the period corresponding to the 4th test day, 1st concentration, non-hyperthyroidism]

[0176] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), average heart rate for the period corresponding to the 4th test day, 2nd concentration, non-hyperthyroidism], and

[0177] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), average heart rate for the period corresponding to the 4th test day, 3rd concentration, non-hyperthyroidism]

[0178] According to a second exemplary embodiment, when the hypothyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed at least once with hypothyroidism as a diagnostic result can be used as training data.

[0179] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hypothyroidism as a diagnostic result may be referred to as a "hypothyroid patient."

[0180] According to a second exemplary embodiment, to train a hypothyroidism prediction model, clinical data of hypothyroid patients may be used to generate a training dataset in a format such as [(period average heart rate corresponding to the first test date - period average heart rate corresponding to the second test date), period average heart rate corresponding to the first test date, thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hypothyroidism or non-hypothyroidism)].

[0181] For example, assuming that a total of three datasets are obtained from a first hypothyroid patient, and the diagnosis for the first test is "hypothyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that can be secured from the clinical data of the first hypothyroid patient can be generated as follows:

[0182] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the first test day, second concentration, hypothyroidism]

[0183] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the first test day, third concentration, hypothyroidism]

[0184] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the second test day, first concentration, non-hypothyroidism]

[0185] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the second test day, third concentration, non-hypothyroidism]

[0186] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the third test day, first concentration, non-hypothyroidism], and

[0187] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the third test day, second concentration, non-hypothyroidism]

[0188] As another example, assuming a total of four datasets are obtained from a second hypothyroid patient, and the diagnosis for the first test is "hypothyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hypothyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that can be secured from the clinical data of the second hypothyroid patient can be generated as follows:

[0189] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), average heart rate for the period corresponding to the first test day, second concentration, hypothyroidism]

[0190] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the first test day, third concentration, hypothyroidism]

[0191] [(average heart rate for the period corresponding to the 1st test day - average heart rate for the period corresponding to the 4th test day), average heart rate for the period corresponding to the 1st test day, 4th concentration, hypothyroidism]

[0192] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the second test day, first concentration, non-hypothyroidism]

[0193] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), average heart rate for the period corresponding to the second test day, third concentration, non-hypothyroidism]

[0194] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), average heart rate for the period corresponding to the second test day, fourth concentration, non-hypothyroidism]

[0195] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), average heart rate for the period corresponding to the third test day, first concentration, hypothyroidism]

[0196] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), average heart rate during the period corresponding to the third test day, second concentration, hypothyroidism]

[0197] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the fourth test day), average heart rate during the period corresponding to the third test day, fourth concentration, hypothyroidism]

[0198] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), average heart rate for the period corresponding to the 4th test day, 1st concentration, non-hypothyroidism]

[0199] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), average heart rate for the period corresponding to the 4th test day, 2nd concentration, non-hypothyroidism], and

[0200] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), average heart rate for the period corresponding to the 4th test day, 3rd concentration, non-hypothyroidism]

[0201] In a second exemplary embodiment, the hormone concentration may be one or a combination selected from the group consisting of T4 concentration, free T4 concentration, T3 concentration, free T3 concentration, TSH concentration, and thyrotropin releasing hormone (TRH) concentration.

[0202] Training Method Embodiment #3. Changes in Heart Rate Difference Values, Hormones Values, and Variances from Reference Days

[0203] According to a third exemplary embodiment described in the present application, to train a learning model, a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day, a value obtained by subtracting the standard deviation of the period heart rate corresponding to the second test day from the standard deviation of the period heart rate corresponding to the first test day, and the hormone concentration on the second test day may be used as input values, and the diagnostic result corresponding to the first test day may be used as a labeling value.

[0204] Here, as the learning model, a classification model including a light gradient boosting machine, a support vector machine, a random forest, an extra tree, an adaboost, an extreme gradient boosting, a catboost, etc. may be used.

[0205] The learning models include a hyperthyroidism prediction model for predicting hyperthyroidism and a hypothyroidism prediction model for predicting hypothyroidism.

[0206] According to a third exemplary embodiment, when the hyperthyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed with hyperthyroidism at least once as a diagnostic result can be used as training data.

[0207] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hyperthyroidism as a diagnostic result may be referred to as a hyperthyroidism patient.

[0208] On the other hand, in calculating the period average heart rate, a resting heart rate over a predetermined period of time that is predetermined based on a particular date may be used, where the predetermined period of time may be one of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 days.

[0209] According to a third exemplary embodiment, to train a hyperthyroidism prediction model, clinical data of hyperthyroid patients may be used to generate a training dataset in a format such as [(average period heart rate corresponding to the first test date - average period heart rate corresponding to the second test date), (standard deviation of period heart rate corresponding to the first test date - standard deviation of period heart rate corresponding to the second test date), thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hyperthyroidism or non-hyperthyroidism)].

[0210] For example, assuming that a total of three datasets are obtained from a first hyperthyroid patient, and the diagnosis for the first test is "hyperthyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that may be obtained from the clinical data of the first hyperthyroid patient may be generated as follows:

[0211] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), (standard deviation of heart rate for the period corresponding to the first test day - standard deviation of heart rate for the period corresponding to the second test day), second concentration, hyperthyroidism]

[0212] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the third test day), third concentration, hyperthyroidism]

[0213] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hyperthyroidism]

[0214] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the third test day), third concentration, non-hyperthyroidism]

[0215] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hyperthyroidism], and

[0216] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the second test day), second concentration, non-hyperthyroidism]

[0217] As another example, assuming a total of four datasets are obtained from a second hyperthyroid patient, and the diagnosis for the first test is "hyperthyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hyperthyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that may be obtained from the clinical data of the second hyperthyroid patient may be generated as follows:

[0218] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), (standard deviation of heart rate for the period corresponding to the first test day - standard deviation of heart rate for the period corresponding to the second test day), second concentration, hyperthyroidism]

[0219] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the third test day), third concentration, hyperthyroidism]

[0220] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, hyperthyroidism]

[0221] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hyperthyroidism]

[0222] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the third test day), third concentration, non-hyperthyroidism]

[0223] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, non-hyperthyroidism]

[0224] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the first test day), first concentration, hyperthyroidism]

[0225] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the second test day), second concentration, hyperthyroidism]

[0226] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, hyperthyroidism]

[0227] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 1st test day), 1st concentration, non-hyperthyroidism]

[0228] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 2nd test day), 2nd concentration, non-hyperthyroidism], and

[0229] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 3rd test day), 3rd concentration, non-hyperthyroidism]

[0230] According to a third exemplary embodiment, when the hypothyroidism prediction model is trained, not all acquired clinical data can be used, and only clinical data of patients who have been diagnosed at least once with hypothyroidism as a diagnostic result can be used as training data.

[0231] Hereinafter, for ease of explanation, a patient who has been diagnosed at least once with hypothyroidism as a diagnostic result may be referred to as a "hypothyroid patient."

[0232] According to a third exemplary embodiment, to train a hypothyroidism prediction model, clinical data of hypothyroid patients may be used to generate a training dataset in a format such as [(average heart rate over time corresponding to the first test date - average heart rate over time corresponding to the second test date), (standard deviation of heart rate over time corresponding to the first test date - standard deviation of heart rate over time corresponding to the second test date), thyroid hormone concentration corresponding to the second test date, diagnosis result corresponding to the first test date (hypothyroidism or non-hypothyroidism)].

[0233] For example, assuming that a total of three datasets are obtained from a first hypothyroid patient, and the diagnosis for the first test is "hypothyroid" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, and the diagnosis for the third test is "normal" with a third hormone concentration, a training dataset that may be obtained from the clinical data of the first hypothyroid patient may be generated as follows:

[0234] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), (standard deviation of heart rate for the period corresponding to the first test day - standard deviation of heart rate for the period corresponding to the second test day), second concentration, hypothyroidism]

[0235] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the third test day), third concentration, hypothyroidism]

[0236] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hypothyroidism]

[0237] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the third test day), third concentration, non-hypothyroidism]

[0238] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hypothyroidism], and

[0239] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the second test day), second concentration, non-hypothyroidism]

[0240] As another example, assuming a total of four datasets are obtained from a second hypothyroid patient, and the diagnosis for the first test is "hypothyroidism" with a first hormone concentration, the diagnosis for the second test is "normal" with a second hormone concentration, the diagnosis for the third test is "hypothyroidism" with a third hormone concentration, and the diagnosis for the fourth test is "normal" with a fourth hormone concentration, a training dataset that may be obtained from the clinical data of the second hypothyroid patient may be generated as follows:

[0241] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the second test day), (standard deviation of heart rate for the period corresponding to the first test day - standard deviation of heart rate for the period corresponding to the second test day), second concentration, hypothyroidism]

[0242] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the third test day), third concentration, hypothyroidism]

[0243] [(average heart rate for the period corresponding to the first test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the first test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, hypothyroidism]

[0244] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the first test day), first concentration, non-hypothyroidism]

[0245] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the third test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the third test day), third concentration, non-hypothyroidism]

[0246] [(average heart rate for the period corresponding to the second test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the second test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, non-hypothyroidism]

[0247] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the first test day), first concentration, hypothyroidism]

[0248] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the second test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the second test day), second concentration, hypothyroidism]

[0249] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the fourth test day), (standard deviation of the interval heart rate for the third test day - standard deviation of the interval heart rate for the fourth test day), fourth concentration, hypothyroidism]

[0250] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 1st test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 1st test day), 1st concentration, non-hypothyroidism]

[0251] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 2nd test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 2nd test day), 2nd concentration, non-hypothyroidism], and

[0252] [(average heart rate for the period corresponding to the 4th test day - average heart rate for the period corresponding to the 3rd test day), (standard deviation of the interval heart rate for the 4th test day - standard deviation of the interval heart rate for the 3rd test day), 3rd concentration, non-hypothyroidism]

[0253] In the above description of the third exemplary embodiment, the standard deviation difference value is used as the input value. However, any one or combination of values ​​selected from the group consisting of the standard deviation difference value of periodic heart rate; the relative standard deviation difference value of periodic heart rate; the skewness difference value of periodic heart rate; the kurtosis difference value of periodic heart rate; and the JS divergence indicating the difference between the variance of the periodic heart rate corresponding to the Xth test date and the variance of the periodic heart rate corresponding to the Yth test date may be used as the input value. These in the group are calculated with respect to the relationship between the periodic heart rate corresponding to the Xth test date and the periodic heart rate corresponding to the Yth test date.

[0254] Training method embodiment #4. Difference value from the heart rate on the reference day, average heart rate during the target day, hormone value on the reference day, and change in variance

[0255] According to a fourth exemplary embodiment described in the present application, to train the learning model, a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day, a value obtained by subtracting the standard deviation of the period heart rate corresponding to the second test day from the standard deviation of the period heart rate corresponding to the first test day, the period average heart rate corresponding to the first test day, and the hormone concentration on the second test day may be used as input values, and the diagnostic result corresponding to the first test day may be used as a labeling value.

[0256] Compared to the third exemplary embodiment, except that the "average heart rate over the period of the first test day" is further used as an input value, the fourth exemplary embodiment is almost identical to the third exemplary embodiment in other respects, so a detailed description of the fourth exemplary embodiment will be omitted.

[0257] In the above description of the fourth exemplary embodiment, the standard deviation difference value is used as the input value. However, any one or combination of values ​​selected from the group consisting of the standard deviation difference value of periodic heart rate; the relative standard deviation difference value of periodic heart rate; the skewness difference value of periodic heart rate; the kurtosis difference value of periodic heart rate; and the JS divergence indicating the difference between the variance of the periodic heart rate corresponding to the Xth test date and the variance of the periodic heart rate corresponding to the Yth test date may be used as the input value. These in the group are calculated with respect to the relationship between the periodic heart rate corresponding to the Xth test date and the periodic heart rate corresponding to the Yth test date.

[0258] Training Method Embodiment #5. Percentage Change in Heart Rate on Reference Days and Hormone Levels on Reference Days

[0259] According to a fifth exemplary embodiment described in the present application, to train a learning model, the rate of change of the period average heart rate on the first test day relative to the second test day (i.e., (period average heart rate corresponding to the first test day - period average heart rate corresponding to the second test day) / (period average heart rate corresponding to the second test day)) and the hormone concentration on the second test day may be used as input values, and the diagnosis result corresponding to the first test day may be used as a labeling value.

[0260] Compared to the first exemplary embodiment, except that "rate of change in period average heart rate" is used instead of "amount of change in period average heart rate", the fifth exemplary embodiment is almost identical to the first exemplary embodiment in other respects, so a detailed description of the fifth exemplary embodiment will be omitted.

[0261] Training Method Embodiment #6. Percentage change in heart rate on a reference day, average heart rate over a period of a subject day, and hormone values ​​on a reference day

[0262] According to a sixth exemplary embodiment described in the present application, to train a learning model, the rate of change of the period average heart rate on the first test day relative to the second test day (i.e., (period average heart rate corresponding to the first test day - period average heart rate corresponding to the second test day) / (period average heart rate corresponding to the second test day)), the period average heart rate corresponding to the first test day, and the hormone concentration on the second test day may be used as input values, and the diagnosis result corresponding to the first test day may be used as a labeling value.

[0263] Compared to the fifth exemplary embodiment, except that the "average heart rate over the period of the first test day" is further used as an input value, the sixth exemplary embodiment is almost identical to the fifth exemplary embodiment in other respects, so a detailed description of the sixth exemplary embodiment will be omitted.

[0264] Training Method Embodiment #7. Percentage change in heart rate, hormone levels, and variance on baseline

[0265] According to a seventh exemplary embodiment described in the present application, in order to train the learning model, the rate of change of the periodic average heart rate on the first test day relative to the second test day (i.e., (periodic average heart rate corresponding to the first test day-periodic average heart rate corresponding to the second test day) / periodic average heart rate corresponding to the second test day), the change in the relative standard deviation of the periodic heart rate on the first test day relative to the second test day (i.e., (standard deviation of the periodic heart rate corresponding to the first test day / periodic average heart rate corresponding to the first test day)-(standard deviation of the periodic heart rate corresponding to the second test day / periodic average heart rate corresponding to the second test day)), and the hormone concentration on the second test day may be used as input values, and the diagnosis result corresponding to the first test day may be used as a labeling value.

[0266] Compared to the fifth exemplary embodiment, except that the "amount of change in the relative standard deviation of the interval heart rate" is further used as an input value, the seventh exemplary embodiment is almost identical to the fifth exemplary embodiment in other respects, so a detailed description of the seventh exemplary embodiment will be omitted.

[0267] In the above description of the seventh exemplary embodiment, the relative standard deviation difference value is used as the input value. However, any one or combination of values ​​selected from the group consisting of the standard deviation difference value of period heart rate; the relative standard deviation difference value of period heart rate; the skewness difference value of period heart rate; the kurtosis difference value of period heart rate; and the JS divergence indicating the difference between the variance of the period heart rate corresponding to the Xth test date and the variance of the period heart rate corresponding to the Yth test date may be used as the input value. These in the group are calculated with respect to the relationship between the period heart rate corresponding to the Xth test date and the period heart rate corresponding to the Yth test date.

[0268] Training Method Embodiment #8. Percentage change in heart rate on the reference day, average heart rate over the period on the subject day, hormone levels on the reference day, and change in variance

[0269] According to an eighth exemplary embodiment described in the present application, in order to train the learning model, the rate of change of the period average heart rate on the first test day relative to the second test day (i.e., (period average heart rate corresponding to the first test day-period average heart rate corresponding to the second test day) / period average heart rate corresponding to the second test day), the change in the relative standard deviation of the interval heart rate on the first test day relative to the second test day (i.e., (standard deviation of the period heart rate corresponding to the first test day / period average heart rate corresponding to the first test day)-(standard deviation of the period heart rate corresponding to the second test day / period average heart rate corresponding to the second test day)), the period average heart rate corresponding to the first test day, and the hormone concentration on the second test day may be used as input values, and the diagnosis result corresponding to the first test day may be used as a labeling value.

[0270] Compared to the seventh exemplary embodiment, except that the "average heart rate over the period of the first reference day" is further used as an input value, the eighth exemplary embodiment is almost identical to the seventh exemplary embodiment in other respects, so a detailed description of the eighth exemplary embodiment will be omitted.

[0271] In the above description of the eighth exemplary embodiment, the difference value of the relative standard deviation is used as the input value. However, any one or combination of values ​​selected from the group consisting of the difference value of the standard deviation of the periodic heart rate; the difference value of the relative standard deviation of the periodic heart rate; the difference value of the skewness of the periodic heart rate; the difference value of the kurtosis of the periodic heart rate; and the JS divergence indicating the difference between the variance of the periodic heart rate corresponding to the Xth test date and the variance of the periodic heart rate corresponding to the Yth test date may be used as the input value. These in the group are calculated with respect to the relationship between the periodic heart rate corresponding to the Xth test date and the periodic heart rate corresponding to the Yth test date.

[0272] Training Method Embodiment #9. Percentage of change (or amount of change) in heart rate on baseline, hormone levels on baseline, change in variance, and number of days between test days

[0273] According to a ninth exemplary embodiment described in the present application, in order to train the learning model, the rate of change of the periodic average heart rate of the first test day relative to the second test day (i.e., (periodic average heart rate corresponding to the first test day-periodic average heart rate corresponding to the second test day) / periodic average heart rate corresponding to the second test day), the change in the relative standard deviation of the periodic heart rate of the first test day relative to the second test day (i.e., (standard deviation of the periodic heart rate corresponding to the first test day / periodic average heart rate corresponding to the first test day)-(standard deviation of the periodic heart rate corresponding to the second test day / periodic average heart rate corresponding to the second test day)), the hormone concentration of the second test day, and the number of days between the first test day and the second test day (days) may be used as input values, and the diagnosis result corresponding to the first test day may be used as a labeling value.

[0274] Compared with the seventh exemplary embodiment, the ninth exemplary embodiment is almost identical to the seventh exemplary embodiment, except that the day interval between the first test date and the second test date is further used as an input value, so a detailed description of the ninth exemplary embodiment is omitted.

[0275] In the above description of the ninth exemplary embodiment, the difference value of the relative standard deviation is used as the input value. However, any one or combination of values ​​selected from the group consisting of the difference value of the standard deviation of the periodic heart rate; the difference value of the relative standard deviation of the periodic heart rate; the difference value of the skewness of the periodic heart rate; the difference value of the kurtosis of the periodic heart rate; and the JS divergence indicating the difference between the variance of the periodic heart rate corresponding to the Xth test date and the variance of the periodic heart rate corresponding to the Yth test date may be used as the input value. These in the group are calculated with respect to the relationship between the periodic heart rate corresponding to the Xth test date and the periodic heart rate corresponding to the Yth test date.

[0276] On the other hand, instead of using the rate of change of the heart rate and the change in the relative standard deviation of the reference day, the change in the heart rate and the change in the standard deviation of the reference day may be used. In this case, in addition to the change in the relative standard deviation, any one or combination of values ​​selected from the group consisting of the difference value of the standard deviation of the periodic heart rate; the difference value of the relative standard deviation of the periodic heart rate; the difference value of the skewness of the periodic heart rate; the difference value of the kurtosis of the periodic heart rate; and the JS divergence indicating the difference between the variance of the periodic heart rate corresponding to the Xth test date and the variance of the periodic heart rate corresponding to the Yth test date may be used as an input value. These in the group are calculated with respect to the relationship between the periodic heart rate corresponding to the Xth test date and the periodic heart rate corresponding to the Yth test date.

[0277] Training Method Embodiment #10. Combination of variables

[0278] According to a tenth exemplary embodiment described in the present application, the combined values ​​of variables shown in Table 1 below may further be used as input values ​​to train a learning model.

[0279] [Table 1] [Table 1]

[0280] When the first period average heart rate, the first standard deviation, the first skewness, the first kurtosis, the first TSH hormone concentration, the first free T4 concentration, the first T4 concentration, the first free T3 concentration, the first T3 concentration, the first TRH concentration, and the diagnosis result for thyroid dysfunction on the first reference date correspond to the first reference date, and the second period average heart rate, the second standard deviation, the second skewness, the second kurtosis, the second TSH hormone concentration, the second free T4 concentration, the second T4 concentration, the second free T3 concentration, the second T3 concentration, the second TRH concentration, and the diagnosis result for thyroid dysfunction on the second reference date correspond to the second reference date, the variables are calculated as follows:

[0281] (1) Change in average heart rate over a period = average heart rate over a period on the first reference day - average heart rate over a period on the second reference day

[0282] (2) Rate of change in average heart rate over a period = (average heart rate over a period on the first reference date - average heart rate over a period on the second reference date) / average heart rate over a period on the second reference date

[0283] (3) Change in standard deviation = Standard deviation of heart rate over a period on the first reference date - Standard deviation of heart rate over a period on the second reference date

[0284] (4) Change in relative standard deviation = (Standard deviation of period heart rate on the first reference day / Period average heart rate on the first reference day) - (Standard deviation of period heart rate on the second reference day / Period average heart rate on the second reference day)

[0285] (5) Change in skewness of period heart rate = skewness of period heart rate on the first reference date - skewness of period heart rate on the second reference date

[0286] (6) Change in kurtosis of period heart rate = kurtosis of period heart rate on the first reference date - skewness of period heart rate on the second reference date

[0287] (7) JS Divergence: JS divergence calculated between the period heart rate on the first reference date and the period heart rate on the second reference date.

[0288] Here, the combined value of the variables means one of the following: a value obtained by multiplying one value selected from the variables (hereinafter referred to as the first value) by another value selected from the variables (hereinafter referred to as the second value); a value obtained by dividing the first value by the second value; and a value obtained by dividing the second value by the first value. For example, the combined value may be a value obtained by dividing the free T4 concentration value on the second reference date by the JS divergence.

[0289] Theoretically, the number of combination values ​​can be 12×11=132.

[0290] Alternatively, combinations of two or more values ​​of the variables may be used.

[0291] When the learning model is trained according to the tenth exemplary embodiment, the input values ​​may be as follows:

[0292] [At least one of the changes in mean heart rate over the period, hormone concentrations on the second test day, and at least one of the combined values ​​listed above]

[0293] [At least one of the following: change in mean heart rate over the period, hormone concentration on the second test day, change in standard deviation of mean heart rate over the period, and at least one of the above combined values]

[0294] [At least one of the changes in average heart rate over the period, the hormone concentration on the second test day, the average heart rate over the period on the first reference day, and at least one of the combined values ​​mentioned above]

[0295] [At least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in standard deviation of average heart rate over the period, average heart rate over the period on the first reference day, and combinations of the above]

[0296] [at least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in standard deviation of average heart rate over the period, average heart rate over the period on the first reference day, interval between the first reference day and the second reference day, and combination of the above values]

[0297] [At least one of the following: percent change in mean heart rate over the period, hormone concentration on the second test day, and at least one of the combined values ​​listed above]

[0298] [At least one of the following: the rate of change in mean heart rate over the period, the hormone concentration on the second test day, the change in the relative standard deviation of mean heart rate over the period, and at least one of the above combined values]

[0299] [At least one of the following: the rate of change in mean heart rate over the period, the hormone concentration on the second test day, the mean heart rate over the period on the first reference day, and at least one of the above combined values]

[0300] [At least one of the following: the rate of change in mean heart rate over the period, the hormone concentration on the second test day, the change in the relative standard deviation of mean heart rate over the period, the mean heart rate over the period on the first reference day, and at least one of the above combined values]

[0301] [At least one of the following: the rate of change in the average heart rate over the period, the hormone concentration on the second test day, the change in the relative standard deviation of the average heart rate over the period, the average heart rate over the period on the first reference day, the number of days between the first reference day and the second reference day, and at least one of the above combinations]

[0302] [At least one of the following: change in mean heart rate over the period, hormone concentration on the second test day, change in mean heart rate over the period or skewness or kurtosis of JS divergence, or combination of the above values]

[0303] [At least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, and combinations of the above]

[0304] [At least one of the following: the change in average heart rate over the period, the hormone concentration on the second test day, the change in skewness or kurtosis of the average heart rate over the period or the JS divergence, the average heart rate over the period on the first reference day, the number of days between the first reference day and the second reference day, and at least one of the combinations of the above]

[0305] [At least one of the following: the rate of change in mean heart rate over the period, the hormone concentration on the second test day, the amount of change in mean heart rate over the period or the skewness or kurtosis of JS divergence, or at least one of the combinations listed above]

[0306] [At least one of the following: rate of change in average heart rate over the period, hormone concentration on the second test day, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, and combinations of the above]

[0307] [At least one of the following: rate of change in average heart rate over the period, hormone concentration on the second test day, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, interval of days between the first reference day and the second reference day, and combination of the above values]

[0308] [At least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in standard deviation of average heart rate over the period, change in skewness or kurtosis of average heart rate over the period, or JS divergence, or a combination of the above]

[0309] [At least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in standard deviation of average heart rate over the period, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, and combinations of the above]

[0310] [at least one of the following: change in average heart rate over the period, hormone concentration on the second test day, change in standard deviation of average heart rate over the period, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, interval between the first reference day and the second reference day, and combination of the above values]

[0311] [At least one of the following: rate of change in mean heart rate over the period, hormone concentration on the second test day, change in relative standard deviation of mean heart rate over the period, change in skewness or kurtosis of mean heart rate over the period, or JS divergence, or a combination of the above]

[0312] [At least one of the following: rate of change in average heart rate over the period, hormone concentration on the second test day, change in relative standard deviation of average heart rate over the period, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, and combinations of the above]

[0313] [At least one of the following: rate of change in average heart rate over the period, hormone concentration on the second test day, change in relative standard deviation of average heart rate over the period, change in skewness or kurtosis of average heart rate over the period or JS divergence, average heart rate over the period on the first reference day, interval of days between the first reference day and the second reference day, and combination of the above values]

[0314] On the other hand, when the learning model is trained according to the tenth exemplary embodiment, the labeling value may be the diagnosis result of the first test day. For example, when the hyperthyroidism prediction model is trained, when the diagnosis result of the first test day is "hyperthyroidism", the label value may be labeled as "hyperthyroidism", or when the diagnosis result of the first test day is "normal", the label value may be labeled as "non-hyperthyroidism". When the hypothyroidism prediction model is trained, when the diagnosis result of the first test day is "hypothyroidism", the label value may be labeled as "hypothyroidism", or when the diagnosis result of the first test day is "normal", the label value may be labeled as "non-hypothyroidism".

[0315] Training Methodology Implementation Example #11. Use of personal information such as gender, age, etc.

[0316] According to the eleventh exemplary embodiment described in the present application, in addition to the input values ​​described above in the first to tenth exemplary embodiments, personal information of a subject may be used as an input value to train a learning model. For example, the gender and / or age of a subject may be used as an input value.

[0317] For example, in addition to the input values ​​described in the sixth exemplary embodiment, when the subject's gender and age are used as input values, the rate of change of the period average heart rate on the first test day relative to the second test day (i.e., (period average heart rate corresponding to the first test day - period average heart rate corresponding to the second test day) / (period average heart rate corresponding to the second test day)), the period average heart rate corresponding to the first test day, the hormone concentration on the second test day, gender, and age may be used as input values, and the diagnosis result corresponding to the first test day may be used as the labeling value.

[0318] In addition to the input values ​​described in the first to tenth exemplary embodiments, detailed descriptions of further uses of the subject's personal information are omitted.

[0319] While it has been described that both gender and age are used as personal information of the subject, embodiments may be possible in which instead of using both gender and age, only gender or age is used, and other types of personal information besides gender and age may also be used.

[0320] Described below are methods and systems for predicting thyroid dysfunction in a subject by using the predictive models described in this application.

[0321] 3. Overall System

[0322] (1) System hardware construction

[0323] FIG. 1 illustrates a system for predicting thyroid dysfunction according to an embodiment described herein.

[0324] Referring to FIG. 1, a system 1 includes a plurality of heart rate measuring devices 10 , a plurality of user terminals 20 , and a server 30 .

[0325] In the following, the heart rate measuring devices 10, the user terminals 20 and the server 30 are described in detail.

[0326] (2) Functions of the heart rate measuring device

[0327] A plurality of heart rate measuring devices 10 may measure the heart rate of a subject (user). For example, the heart rate measuring device 10 may measure the subject's heart rate per unit time (beats per minute). That is, the heart rate measuring device 10 may measure the subject's heart rate per minute.

[0328] A plurality of heart rate measuring devices 10 may store the measured heart rates of the subject. Here, when storing a plurality of heart rates, the heart rate measuring device 10 stores the heart rate together with the time points at which the heart rate (heart rate per unit time) is measured.

[0329] Multiple heart rate measuring devices 10 may transmit the stored heart rates to multiple user terminals 20 and / or a server 30.

[0330] A plurality of heart rate measuring devices 10 may receive prediction results for thyroid dysfunction from the server 30 or the user terminal 20.

[0331] (3) Type of heart rate measurement device

[0332] The heart rate measuring device may be a wearable watch, a wearable band, or a patch type device.

[0333] (4) Functions of the user device

[0334] A plurality of user terminals 20 transmit information to the server 30 and receive information from the server 30 via various networks.

[0335] A number of user terminals 20 may receive the measured heart rates from the heart rate measuring device 10 and may store the received heart rates.

[0336] The user terminal 20 may pre-process the stored heart rates. As an example, the user terminal 20 may select the heart rate measured during the resting period from all the stored heart rates. As another example, the user terminal 20 may select the period heart rate described above based on a specific date from all the stored heart rates. In yet another example, the user terminal 20 may calculate the average of the period heart rates selected based on a specific date. In yet another example, the user terminal 20 may calculate the standard deviation, relative standard deviation, skewness, kurtosis, etc. of the period heart rate selected based on a specific date.

[0337] The multiple user terminals 20 may obtain test dates for the user's hormone concentrations and may obtain test results (e.g., hormone concentrations and / or doctor's diagnosis results) corresponding to the test dates. The multiple user terminals 20 may obtain the test dates and the corresponding test results by user input or by externally receiving the test dates and the corresponding test results via a network.

[0338] Multiple user terminals 20 may transmit stored heart rates and / or pre-processing results to the server 30.

[0339] Multiple user terminals 20 may transmit test dates and test results to the server 30 .

[0340] A plurality of user terminals 20 may receive from the server 30 the prediction results for thyroid dysfunction processed by the server 30.

[0341] (5) Type of user device

[0342] The multiple user terminals 20 may be at least one of a smart phone, a tablet PC, a laptop, and a desktop.

[0343] (6) Server functions

[0344] The server 30 transmits information to a plurality of user terminals 20 via various networks, and receives information from a plurality of user terminals 20 .

[0345] The server 30 may receive all or part of the measured heart rates from a plurality of user terminals 20. Here, the server 30 may pre-process the received heart rates.

[0346] Alternatively, the server 30 may receive results from multiple user terminals 20 that have been preprocessed by the user terminals 20 .

[0347] In addition, the server 30 may obtain test dates and corresponding test results for hormone concentrations from each user. The test dates and test results may be received from the user terminal 20 or other external device.

[0348] The server 30 may obtain a prediction result of thyroid dysfunction for the subject based on the pre-processing results.

[0349] The server 30 may transmit the prediction results for thyroid dysfunction to the heart rate measuring device 10 and / or to multiple user terminals 20.

[0350] (7) System software construction

[0351] In order for System 1 to operate, several software components are required.

[0352] In order to carry out communications between the user terminals 20 and the server 30, terminal software needs to be installed on the user terminals 20, and server software needs to be installed on the server 30.

[0353] Several predictive models can be used to predict thyroid dysfunction based on pretreatment results.

[0354] The prediction models may be executed by software installed on the server 30. Alternatively, the prediction models may be executed by terminal software installed on the user terminal 20. Alternatively, some of the learning models may be executed by the user terminal 20, and others may be executed by the server 30.

[0355] (8) Elements of a heart rate measuring device

[0356] FIG. 2 is a block diagram illustrating a heart rate measuring device described herein.

[0357] Referring to FIG. 2, the heart rate measuring device 10 described in the present application includes an output unit 110, a communication unit 120, a memory 130, a heart rate measuring unit 140, and a controller 150.

[0358] The output unit 110 outputs various types of information according to control commands of the controller 150. According to an embodiment, the output unit 110 may include a display 112 for visually outputting information to the user. Alternatively, although not shown in the drawings, the output unit 110 may include a speaker for audibly outputting information to the user and a vibration motor for tactilely outputting information to the user.

[0359] The communication unit 120 may include a wireless communication module and / or a wired communication module, where examples of the wireless communication module include a Wi-Fi (registered trademark) communication module, a cellular communication module, and the like.

[0360] The memory 130 stores executable code readable by the controller 150, processed result values, necessary data, etc. Examples of the memory 130 may include a hard disk drive (HDD), a solid state disk (SSD), a silicon disk drive (SDD), a ROM, a RAM, etc. The memory 130 may store the above-mentioned terminal software, and may store executable code for implementing the above-mentioned various pre-processing algorithms and / or learning models. Furthermore, the memory 130 may store therein the subject's heart rate acquired through the heart rate measurement unit 140 and the time point at which the heart rate was measured.

[0361] The heart rate measuring unit 140 can measure the heart rate per minute of the subject (user). A method for measuring the heart rate per minute of the subject is widely known, and therefore a detailed description thereof will be omitted.

[0362] The controller 150 may include at least one processor, where each of the processors may perform a predetermined operation by executing at least one instruction stored in the memory 130. In particular, the controller 150 may process information according to terminal software, pre-processing algorithms, and / or learning models running on the heart rate measuring device 10. Meanwhile, the controller 150 controls the overall operation of the heart rate measuring device 10.

[0363] Although not shown in the drawings, the heart rate measuring device 10 may include a user input unit, through which the heart rate measuring device 10 may receive from the user various types of information required for the operation of the heart rate measuring device 10.

[0364] (9) User terminal elements

[0365] FIG. 3 is a block diagram illustrating a user terminal described herein.

[0366] Referring to FIG. 3, the user terminal 20 described in the present application includes an output unit 210, a communication unit 220, a memory 230, and a controller 250.

[0367] The output unit 210 outputs various types of information according to control commands from the controller 250. According to an embodiment, the output unit 210 may include a display 212 for visually outputting information to the user. Alternatively, although not shown in the drawings, the output unit 210 may include a speaker for audibly outputting information to the user and a vibration motor for tactilely outputting information to the user.

[0368] The communication unit 220 may include a wireless communication module and / or a wired communication module, where examples of the wireless communication module include a Wi-Fi communication module, a cellular communication module, and the like.

[0369] The memory 230 stores executable codes readable by the controller 250, processed result values, necessary data, etc. Examples of the memory 230 may include a hard disk drive (HDD), a solid state disk (SSD), a silicon disk drive (SDD), a ROM, a RAM, etc. The memory 230 may store the above-mentioned terminal software, and may store executable codes for implementing the above-mentioned various pre-processing algorithms and / or learning models.

[0370] The controller 240 may include at least one processor, where each of the processors may perform a predetermined operation by executing at least one instruction stored in the memory 230. In particular, the controller 240 may process information according to terminal software, pre-processing algorithms, and / or learning models running on the user terminal 20. Meanwhile, the controller 240 controls the overall operation of the user terminal 20.

[0371] Although not shown in the drawings, the user terminal 20 may include a user input unit through which the user terminal 20 may receive various types of information required for the operation of the user terminal 20 from a user.

[0372] (10) Server elements

[0373] FIG. 4 is a block diagram illustrating a server according to the present application.

[0374] Referring to FIG. 4, the server 30 described herein includes a communication unit 310, a memory 320, and a controller 330.

[0375] The communication unit 310 may include a wireless communication module and / or a wired communication module, where examples of the wireless communication module include a Wi-Fi communication module, a cellular communication module, and the like.

[0376] The memory 320 stores executable codes readable by the controller 330, processed result values, necessary data, etc. Examples of the memory 320 may include a hard disk drive (HDD), a solid state disk (SSD), a silicon disk drive (SDD), a ROM, a RAM, etc. The memory 320 may store the above-mentioned server software, and may store executable codes for implementing the above-mentioned various pre-processing algorithms and / or learning models. Furthermore, the memory 320 may store therein the heart rate and / or the corresponding pre-processing results for each user from the user terminal 20.

[0377] The controller 330 may include at least one processor, where each of the processors may perform a predetermined operation by executing at least one instruction stored in the memory 320. In particular, the controller 330 may process information according to server software, pre-processing algorithms, and / or learning models running on the server 30. Meanwhile, the controller 330 controls the overall operation of the server 30.

[0378] 4. Methods for predicting thyroid dysfunction

[0379] The method for predicting thyroid dysfunction described in the present application can be performed by the system 1 described above.

[0380] FIG. 5 is a flow chart illustrating the method for predicting thyroid dysfunction described herein.

[0381] Referring to FIG. 5, the method for predicting thyroid dysfunction in a subject described in the present application includes: acquiring a trigger signal in step S100; acquiring interval heart rates corresponding to a target day determined by the trigger signal in step S110; pre-processing the interval heart rates corresponding to the target day in step S120; acquiring hormone concentrations corresponding to a reference day in step S130; acquiring a pre-processed result for the interval heart rates corresponding to the reference day in step S140; acquiring a difference between the pre-processed result for the interval heart rates corresponding to the target day and the pre-processed result for the interval heart rates corresponding to the reference day in step S150; and processing the acquired difference and the hormone concentrations corresponding to the reference day using a thyroid dysfunction prediction model in step S160 to obtain a thyroid dysfunction result for the subject.

[0382] Acquiring a trigger signal in step S100

[0383] A trigger signal may be obtained in step S100.

[0384] Here, the trigger signal refers to a signal for requesting acquisition of a prediction result of thyroid dysfunction for a subject.

[0385] The trigger signal may be generated by a user's input. For example, the user may make a request through a user input unit of the user terminal 20 for obtaining a prediction result for thyroid dysfunction, and in response to the user's request, the user terminal 20 may generate a trigger signal. As another example, the user may make a request through a user input unit of the heart rate measuring device 10 communicating with the user terminal 20 for obtaining a prediction result for thyroid dysfunction, and in response to the user's request, the heart rate measuring device 10 or the user terminal 20 may generate a trigger signal.

[0386] Alternatively, the trigger signal may be generated automatically by the system 1 regardless of a user request. For example, software installed in the heart rate measuring device 10, the user terminal 20 or the server 30 may generate the trigger signal according to a predetermined rule. When the trigger signal is generated by the heart rate measuring device 10 or the server 30, the generated trigger signal may be transmitted to the user terminal 20.

[0387] Determining the target date based on the trigger signal in step S110

[0388] The target date may be determined by the trigger signal. For example, the date on which the trigger signal is generated may be determined as the target date. As another example, when the trigger signal is generated, a specific date may be specified by a user input or by the system 1. In this case, the specified specific date may be determined as the target date.

[0389] Acquisition of section heart rate corresponding to the target day in step S110

[0390] In step S110, the interval heart rate corresponding to the target day may be obtained according to a predetermined period.

[0391] The predetermined period of time may be one of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 days.

[0392] When the predetermined period is 10 days and the target date is July 21st, interval heart rates including the resting heart rate corresponding to July 21st, the resting heart rate corresponding to July 20th, the resting heart rate corresponding to July 19th, the resting heart rate corresponding to July 18th, the resting heart rate corresponding to July 17th, the resting heart rate corresponding to July 16th, the resting heart rate corresponding to July 15th, the resting heart rate corresponding to July 14th, the resting heart rate corresponding to July 13th, and the resting heart rate corresponding to July 12th are obtained.

[0393] The interval heart rates are selected from all stored heart rates corresponding to the determined date of interest and the date determined by the predetermined time period. Based on the measurement time points corresponding to the heart rates, the heart rate whose measurement time point is within the subject's (i.e., user's) resting period is selected as the resting heart rate.

[0394] The rest period may be determined in a variety of ways.

[0395] For example, a resting period may be determined, via a motion sensor integrated into the heart rate measuring device 10 or the user terminal 20, to be a period during which the user's movement is equal to or lower than a reference value.

[0396] As another example, rest periods may be determined through various sensors integrated into the heart rate measuring device 10 or the user terminal 20 to be periods during which the user is considered to be at rest or asleep.

[0397] As yet another example, the resting period may be determined by a sleep period (sleep start time and sleep end time) directly input by the user via the heart rate measuring device 10 or the user terminal 20.

[0398] Pre-processing of section heart rates corresponding to target days in step S120

[0399] In step S120, segment heart rates corresponding to the target days may be pre-processed.

[0400] The average of the segment heart rates corresponding to the days of interest may be calculated.

[0401] Various parameters related to the dispersion of the interval heart rates corresponding to the target day may be calculated. As an example, the standard deviation of the interval heart rates corresponding to the target day may be calculated. As another example, the skewness of the interval heart rates corresponding to the target day may be calculated. As yet another example, the kurtosis of the interval heart rates corresponding to the target day may be calculated.

[0402] Among various pre-processing results of the interval heart rate corresponding to the target day, only the pre-processing results related to the input values ​​used to train the prediction model described later may be obtained. For example, when the rate of change of the average value of the interval heart rate on the target day and the reference day and the change amount of the relative standard deviation are used to train the prediction model, only the average and standard deviation of the interval heart rate corresponding to the target day may be obtained from the interval heart rate corresponding to the target day. However, all or part of the pre-processing results described above may be obtained regardless of the input values ​​used to train the prediction model.

[0403] Determination of the reference date in step S130

[0404] A baseline date for the subject may be determined in step S130.

[0405] The reference date may be selected from one or more test dates on which the subject underwent hormone testing, i.e., the reference date may be selected from a previous hormone test date of the subject.

[0406] The reference date may be selected from one or more stored dates corresponding to the subject's hormone levels, i.e., the reference date may be selected from dates for which the subject's past hormone levels are stored.

[0407] Only one reference date may be selected, but no limitation is imposed thereto, i.e., multiple reference dates may be selected.

[0408] When only one reference date is selected from multiple test dates or multiple dates for which hormone concentrations were stored, the reference date may be determined to be the date closest to the subject date.

[0409] Obtaining hormone concentrations corresponding to the reference date in step S130

[0410] Hormone concentrations corresponding to the determined reference date may be obtained.

[0411] The hormone concentration may be one selected from the group consisting of T4 concentration, free T4 concentration, T3 concentration, free T3 concentration, TSH concentration, and TRH concentration, or may be a combination thereof.

[0412] Among the hormone concentrations corresponding to the determined reference date, only the hormone concentration values ​​used for training the described prediction model can be selectively selected.For example, if the hormone concentrations corresponding to the determined reference date include T4 concentration, free T4 concentration, T3 concentration, free T3 concentration, TSH concentration, and TRH concentration, when only free T4 and TSH concentration are used for training the prediction model, only free T3 and TSH concentration can be obtained.

[0413] When multiple reference dates are determined, hormone concentrations corresponding to each of the multiple reference dates may be obtained.

[0414] In step S140, the pre-processing result of the section heart rate corresponding to the reference date is obtained.

[0415] When pre-processing of the section heart rates corresponding to the reference date has already been performed and the pre-processing results are stored, the stored pre-processing results can be obtained.

[0416] When pre-processing of the interval heart rates corresponding to the reference date has not yet been performed, pre-processing of the interval heart rates corresponding to the reference date may be performed.

[0417] The acquisition and pre-processing of the section heart rate corresponding to the reference date is similar to the acquisition and pre-processing of the section heart rate corresponding to the target date described above, and therefore a detailed description thereof will be omitted.

[0418] As a result, the pre-processing results for the interval heart rates for the reference days may include at least one of the parameters related to the average interval heart rate corresponding to the reference days and the variance of the interval heart rate corresponding to the reference days, where the parameters related to the variance of the interval heart rate corresponding to the reference days may include at least one of the standard deviation of the interval heart rate corresponding to the target days, the skewness of the interval heart rate corresponding to the target days, and the kurtosis of the interval heart rate corresponding to the target days.

[0419] When a plurality of reference dates are determined, pre-processing results for the section heart rates corresponding to each of the plurality of reference dates may be obtained.

[0420] On the other hand, among various pre-processing results of the interval heart rate corresponding to the reference date, only the pre-processing results related to the input values ​​used for training the prediction model described later may be obtained. For example, when the change rate of the average value of the interval heart rate on the target date and the reference date and the change amount of the relative standard deviation are used for training the prediction model, only the average value and the standard deviation of the interval heart rate corresponding to the reference date may be obtained from the interval heart rate corresponding to the reference date. However, all or part of the pre-processing results described above may be obtained regardless of the input values ​​used for training the prediction model.

[0421] Obtaining the difference between the pre-processed result of the section heart rate of the target day and the pre-processed result of the section heart rate of the reference day in step S150

[0422] The difference between the pre-processed result for the interval heart rate of the target day and the pre-processed result for the interval heart rate of the reference day may be one or a combination of the following:

[0423] 1) The amount of change in the average section heart rate for the target day (average heart rate for the target day) from the average section heart rate for the reference day (average heart rate for the reference day) (average heart rate for the target day - average heart rate for the reference day)

[0424] 2) The rate of change in the average section heart rate corresponding to the target day (average heart rate for the target day) relative to the average section heart rate corresponding to the base day (average heart rate for the base day) ((average heart rate for the target day - average heart rate for the base day) / average heart rate for the base day)

[0425] 3) The change in the standard deviation of the interval heart rate corresponding to the target day relative to the standard deviation of the interval heart rate corresponding to the base day (standard deviation of the interval heart rate corresponding to the target day - standard deviation of the interval heart rate corresponding to the base day)

[0426] 4) The change in the relative standard deviation of the interval heart rate corresponding to the target day compared to the relative standard deviation of the interval heart rate corresponding to the base day ((standard deviation of the interval heart rate corresponding to the target day / average heart rate for the period on the target day) - (standard deviation of the interval heart rate corresponding to the base day / average heart rate for the period on the base day))

[0427] 5) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0428] 6) The amount of change in kurtosis of the section heart rate corresponding to the target day relative to the kurtosis of the section heart rate corresponding to the reference day (kurtosis of the section heart rate corresponding to the target day - kurtosis of the section heart rate corresponding to the reference day)

[0429] 7) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0430] When multiple reference dates are determined, a difference between the pre-processing results for each of the multiple reference dates may be obtained. For example, a difference between a first pre-processing result determined between a first reference date and a target date, and a difference between a second pre-processing result determined between a second reference date and a target date may be obtained.

[0431] On the other hand, among the differences between the preprocessing results, only the differences between the preprocessing results related to the input values ​​used in the training of the prediction model described later can be obtained. For example, when the change rate of the average value of the section heart rate for the target day and the reference day and the change amount of the relative standard deviation are used in the training of the prediction model, only the change rate of the average section heart rate (the period average heart rate of the target day) corresponding to the target day relative to the average section heart rate (the period average heart rate of the reference day) corresponding to the reference day and the change amount of the standard deviation of the section heart rate corresponding to the target day relative to the standard deviation of the section heart rate corresponding to the reference day are obtained. However, all or part of the differences between the preprocessing results described above can be obtained regardless of the input values ​​used in the training of the prediction model.

[0432] Calculating combination values

[0433] According to some embodiments described herein, a combination value of the variables may be calculated.

[0434] When the first period average heart rate, the first standard deviation, the first skewness, the first kurtosis, the first TSH hormone concentration, the first free T4 concentration, the first T4 concentration, the first free T3 concentration, the first T3 concentration, the first TRH concentration, and the diagnosis result for thyroid dysfunction on the first reference date correspond to the first reference date, and the second period average heart rate, the second standard deviation, the second skewness, the second kurtosis, the second TSH hormone concentration, the second free T4 concentration, the second T4 concentration, the second free T3 concentration, the second T3 concentration, the second TRH concentration, and the diagnosis result for thyroid dysfunction on the second reference date correspond to the second reference date, the variables are calculated as follows:

[0435] (1) Change in average heart rate over a period = average heart rate over a period on the first reference day - average heart rate over a period on the second reference day

[0436] (2) Rate of change in average heart rate over a period = (average heart rate over a period on the first reference date - average heart rate over a period on the second reference date) / average heart rate over a period on the second reference date

[0437] (3) Change in standard deviation = Standard deviation of interval heart rate on the first reference day - Standard deviation of interval heart rate on the second reference day

[0438] (4) Change in relative standard deviation = (Standard deviation of interval heart rate on the first reference day / Average heart rate for the period on the first reference day) - (Standard deviation of interval heart rate on the second reference day / Average heart rate for the period on the second reference day)

[0439] (5) Change in skewness of interval heart rate = skewness of interval heart rate on the first reference day - skewness of interval heart rate on the second reference day

[0440] (6) Change in kurtosis of interval heart rate = kurtosis of interval heart rate on the first reference day - skewness of interval heart rate on the second reference day

[0441] (7) JS Divergence: JS Divergence calculated between the interval heart rate on the first reference day and the interval heart rate on the second reference day

[0442] (8) TSH hormone concentration

[0443] (9) First free T4 concentration

[0444] (10) 1st T4 concentration

[0445] (11) First free T3 concentration

[0446] (12) 1st T3 concentration

[0447] (13) 1st TRH concentration

[0448] (14) Secondary TSH hormone concentration

[0449] (15)Second free T4 concentration

[0450] (16)Second T4 concentration

[0451] (17)Second free T3 concentration

[0452] (18) second T3 concentration, and

[0453] (19)Second TRH concentration

[0454] Here, the combined value of the variables means one of the following: a value obtained by multiplying one value selected from the variables (hereinafter referred to as the first value) by another value selected from the variables (hereinafter referred to as the second value); a value obtained by dividing the first value by the second value; and a value obtained by dividing the second value by the first value. For example, the combined value may be a value obtained by dividing the free T4 concentration value on the second reference date by the JS divergence.

[0455] Calculating the number of days between the target date and the base date

[0456] According to some embodiments described herein, the interval of days between the target date and the reference date may be obtained.

[0457] Acquisition of user's personal information

[0458] According to some embodiments described herein, personal information of the subject, i.e., personal information of the user, may be obtained.

[0459] Obtaining a prediction result of thyroid dysfunction through the thyroid dysfunction prediction model in step S160

[0460] Obtaining a prediction result of hyperthyroidism through a hyperthyroidism prediction model

[0461] The hyperthyroidism prediction model trained as described in embodiment #1 of the training method may be input with the acquired hormone concentration on the reference day, the acquired change amount of the average of the interval heart rate corresponding to the target day (the average interval heart rate on the target day) relative to the average (the interval heart rate corresponding to the reference day and the average interval heart rate on the reference day). Thus, the result of whether the subject has or does not have hyperthyroidism may be obtained. That is, the prediction result for hyperthyroidism may be obtained.

[0462] The hyperthyroidism prediction model trained as described in embodiment #2 of the training method may be input with the acquired hormone concentration on the reference day, the acquired change in the average section heart rate (average period heart rate on the reference day) on the target day relative to the average section heart rate (average period heart rate on the reference day) on the reference day, and the average period heart rate on the target day. Thus, the result of whether the subject has or does not have hyperthyroidism may be obtained. That is, the prediction result for hyperthyroidism may be obtained.

[0463] For the hyperthyroidism prediction model trained as described in embodiment #3 of the training method, the acquired hormone concentration of the reference day, the acquired change amount of the average of the interval heart rate (the average of the interval heart rate of the target day) corresponding to the reference day relative to the average of the interval heart rate (the average of the interval heart rate of the reference day), and the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be input. Thus, the result of whether the subject has or does not have hyperthyroidism can be obtained. That is, the prediction result for hyperthyroidism can be obtained. Here, the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be one or a combination selected from the group consisting of:

[0464] 1) The change in the standard deviation of the interval heart rate on the target day compared to the standard deviation of the interval heart rate on the base day (standard deviation of the interval heart rate on the target day - standard deviation of the interval heart rate on the base day)

[0465] 2) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0466] 3) The amount of change in kurtosis of the interval heart rate corresponding to the target day relative to the kurtosis of the interval heart rate corresponding to the reference day (kurtosis of the interval heart rate corresponding to the target day - kurtosis of the interval heart rate corresponding to the reference day)

[0467] 4) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0468] The hyperthyroidism prediction model trained as described in embodiment #4 of the training method may be input with the acquired hormone concentration of the reference day, the acquired change of the average of the interval heart rate (average interval heart rate of the reference day) corresponding to the target day with respect to the average of the interval heart rate (average interval heart rate of the reference day), the average of the interval heart rate corresponding to the target day, and the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day.Therefore, the result of whether the subject has or does not have hyperthyroidism may be obtained. That is, the prediction result of hyperthyroidism may be obtained.

[0469] The acquired hormone concentration on the reference day, the acquired rate of change of the average interval heart rate (average interval heart rate on the reference day) corresponding to the target day relative to the average interval heart rate (average interval heart rate on the reference day) corresponding to the reference day may be inputted into the hyperthyroidism prediction model trained as described in embodiment #5 of the training method. Thus, a result of whether the subject has or does not have hyperthyroidism may be obtained. That is, a prediction result regarding hyperthyroidism may be obtained.

[0470] The hyperthyroidism prediction model trained as described in embodiment #6 of the training method can be input with the acquired hormone concentration of the reference day, the acquired rate of change of the average of the interval heart rate (the average of the interval heart rate of the reference day) corresponding to the target day with respect to the average of the interval heart rate (the average of the interval heart rate of the reference day) corresponding to the reference day, and the average of the interval heart rate corresponding to the target day. Thus, the result of whether the subject has or does not have hyperthyroidism can be obtained. That is, the prediction result of hyperthyroidism can be obtained.

[0471] For the hyperthyroidism prediction model trained as described in embodiment #7 of the training method, the acquired hormone concentration of the reference day, the acquired rate of change of the average of the interval heart rate (the average of the interval heart rate of the target day) corresponding to the reference day relative to the average of the interval heart rate (the average of the interval heart rate of the reference day), and the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be input. Thus, the result of whether the subject has or does not have hyperthyroidism can be obtained. That is, the prediction result for hyperthyroidism can be obtained. Here, the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be one or a combination selected from the group consisting of:

[0472] 1) The change in the relative standard deviation of the interval heart rate corresponding to the target day compared to the relative standard deviation of the interval heart rate corresponding to the base day ((standard deviation of the interval heart rate corresponding to the target day / average heart rate for the period on the target day) - (standard deviation of the interval heart rate corresponding to the base day / average heart rate for the period on the base day))

[0473] 2) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0474] 3) The amount of change in kurtosis of the interval heart rate corresponding to the target day relative to the kurtosis of the interval heart rate corresponding to the reference day (kurtosis of the interval heart rate corresponding to the target day - kurtosis of the interval heart rate corresponding to the reference day)

[0475] 4) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0476] For the hyperthyroidism prediction model trained as described in embodiment #8 of the training method, the acquired hormone concentration of the reference day, the acquired rate of change of the average of the interval heart rate (the average of the interval heart rate of the reference day) corresponding to the target day with respect to the average of the interval heart rate (the average of the interval heart rate of the reference day), the average of the interval heart rate corresponding to the target day, and the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be input. Thus, the result of whether the subject has or does not have hyperthyroidism can be obtained. That is, the prediction result for hyperthyroidism can be obtained. Here, the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be one or a combination selected from the group consisting of:

[0477] 1) The change in the relative standard deviation of the interval heart rate corresponding to the target day compared to the relative standard deviation of the interval heart rate corresponding to the base day ((standard deviation of the interval heart rate corresponding to the target day / average heart rate for the period on the target day) - (standard deviation of the interval heart rate corresponding to the base day / average heart rate for the period on the base day))

[0478] 2) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0479] 3) The amount of change in kurtosis of the interval heart rate corresponding to the target day relative to the kurtosis of the interval heart rate corresponding to the reference day (kurtosis of the interval heart rate corresponding to the target day - kurtosis of the interval heart rate corresponding to the reference day)

[0480] 4) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0481] For the hyperthyroidism prediction model trained as described in embodiment #9 of the training method, the acquired hormone concentration of the reference day, the acquired change rate (or change amount) of the average of the interval heart rate (the average of the interval heart rate of the target day) corresponding to the reference day relative to the average of the interval heart rate (the average of the interval heart rate of the reference day), the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day, and the number of days between the target day and the reference day can be input. Thus, the result of whether the subject has or does not have hyperthyroidism can be obtained. That is, the prediction result for hyperthyroidism can be obtained. Here, the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day can be one or a combination selected from the group consisting of:

[0482] 1) The change in the standard deviation of the interval heart rate on the target day compared to the standard deviation of the interval heart rate on the base day (standard deviation of the interval heart rate on the target day - standard deviation of the interval heart rate on the base day)

[0483] 2) The change in the relative standard deviation of the interval heart rate corresponding to the target day compared to the relative standard deviation of the interval heart rate corresponding to the base day ((standard deviation of interval heart rate corresponding to the target day / average heart rate for the period on the target day) - (standard deviation of interval heart rate corresponding to the base day / average heart rate for the period on the base day))

[0484] 3) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0485] 4) The amount of change in kurtosis of the interval heart rate corresponding to the target day relative to the kurtosis of the interval heart rate corresponding to the reference day (kurtosis of the interval heart rate corresponding to the target day - kurtosis of the interval heart rate corresponding to the reference day)

[0486] 5) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0487] The hyperthyroidism prediction model trained as described in embodiment #10 of the training method may be input with the acquired hormone concentration on the reference day, the acquired rate of change (or amount of change) of the average of the interval heart rate (average heart rate on the reference day) corresponding to the target day with respect to the average of the interval heart rate (average heart rate on the reference day), the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day, and the combination value of the variables. Thus, the result of whether the subject has or does not have hyperthyroidism may be obtained. That is, the prediction result for hyperthyroidism may be obtained.

[0488] Here, the difference between the parameter related to the variance of the interval heart rate corresponding to the target day and the parameter related to the variance of the interval heart rate corresponding to the reference day may be one or a combination selected from the group consisting of:

[0489] 1) The change in the standard deviation of the interval heart rate on the target day compared to the standard deviation of the interval heart rate on the base day (standard deviation of the interval heart rate on the target day - standard deviation of the interval heart rate on the base day)

[0490] 2) The change in the relative standard deviation of the interval heart rate corresponding to the target day compared to the relative standard deviation of the interval heart rate corresponding to the base day ((standard deviation of interval heart rate corresponding to the target day / average heart rate for the period on the target day) - (standard deviation of interval heart rate corresponding to the base day / average heart rate for the period on the base day))

[0491] 3) The amount of change in the skewness of the interval heart rate corresponding to the target day relative to the skewness of the interval heart rate corresponding to the reference day (skewness of the interval heart rate corresponding to the target day - skewness of the interval heart rate corresponding to the reference day)

[0492] 4) The amount of change in kurtosis of the interval heart rate corresponding to the target day relative to the kurtosis of the interval heart rate corresponding to the reference day (kurtosis of the interval heart rate corresponding to the target day - kurtosis of the interval heart rate corresponding to the reference day)

[0493] 5) JS divergence between the variance of the interval heart rate corresponding to the target day and the variance of the interval heart rate corresponding to the reference day

[0494] Meanwhile, the combination values ​​of the variables are described in detail in embodiment #10 of the training method, so a detailed description will be omitted.

[0495] The following values ​​may be input to the hyperthyroidism prediction model trained as described in embodiment #11 of the training method. Thus, the result of whether the subject has hyperthyroidism or not may be obtained. That is, the prediction result for hyperthyroidism may be obtained.

[0496] (1) The hormone concentration obtained on the reference day, the change in the average interval heart rate corresponding to the reference day (average heart rate for the period on the reference day) from the average interval heart rate corresponding to the reference day (average heart rate for the period on the reference day), and the age and / or sex of the subject.

[0497] (2) The hormone concentration obtained on the reference day, the change in the average section heart rate corresponding to the target day (average heart rate for the period on the target day) from the average section heart rate corresponding to the reference day (average heart rate for the period on the reference day), the average heart rate for the period on the target day, and the age and / or sex of the subject.

[0498] (3) the hormone concentration obtained on the reference day, the change in the average interval heart rate corresponding to the target day (average interval heart rate on the target day) from the average interval heart rate corresponding to the reference day (average interval heart rate on the reference day), the difference between a parameter related to the variance of the interval heart rate corresponding to the target day and a parameter related to the variance of the interval heart rate corresponding to the reference day, and the age and / or sex of the subject.

[0499] (4) the hormone concentration obtained on the reference day, the change in the average interval heart rate corresponding to the target day (average interval heart rate on the target day) relative to the average interval heart rate corresponding to the reference day (average interval heart rate on the reference day), the average interval heart rate corresponding to the target day, a parameter related to the variance of the interval heart rate corresponding to the target day, and a difference between a parameter related to the variance of the interval heart rate corresponding to the reference day, and the age and / or sex of the subject.

[0500] (5) The hormone concentration obtained on the reference day, the rate of change obtained in the average interval heart rate corresponding to the reference day (average heart rate for the period on the reference day) relative to the average interval heart rate corresponding to the reference day (average heart rate for the period on the reference day), and the age and / or sex of the subject.

[0501] (6) The hormone concentration obtained on the reference day, the rate of change obtained of the average interval heart rate corresponding to the target day (average interval heart rate on the target day) relative to the average interval heart rate corresponding to the reference day (average interval heart rate on the reference day), the average interval heart rate corresponding to the target day, and the age and / or sex of the subject.

[0502] (7) the hormone concentration obtained on the reference day, the rate of change of the average interval heart rate corresponding to the target day (average interval heart rate on the target day) relative to the average interval heart rate corresponding to the reference day (average interval heart rate on the reference day), the difference between a parameter related to the variance of the interval heart rate corresponding to the target day and a parameter related to the variance of the interval heart rate corresponding to the reference day, and the age and / or sex of the subject.

[0503] (8) the hormone concentration obtained on the reference day, the rate of change of the average interval heart rate corresponding to the target day (average interval heart rate on the target day) relative to the average interval heart rate corresponding to the reference day (average interval heart rate on the reference day), the average interval heart rate corresponding to the target day, a parameter related to the variance of the interval heart rate corresponding to the target day, and a difference between a parameter related to the variance of the interval heart rate corresponding to the reference day, and the age and / or sex of the subject.

[0504] (9) the hormone concentration obtained on the reference day, the rate of change (or amount of change) of the average interval heart rate (average heart rate for the period on the reference day) corresponding to the target day relative to the average interval heart rate (average heart rate for the period on the reference day) corresponding to the reference day, the difference between a parameter related to the variance of the interval heart rate corresponding to the target day and a parameter related to the variance of the interval heart rate corresponding to the reference day, the number of days between the target day and the reference day, and the age and / or sex of the subject

[0505] (10) the hormone concentration obtained on the reference day, the rate of change (or amount of change) of the average interval heart rate (average interval heart rate on the reference day) corresponding to the target day relative to the average interval heart rate (average interval heart rate on the reference day) corresponding to the reference day, the difference between a parameter related to the variance of the interval heart rate corresponding to the target day and a parameter related to the variance of the interval heart rate corresponding to the reference day, a combination value of the variables, and the age and / or sex of the subject.

[0506] Obtaining a prediction result for hypothyroidism through a hypothyroidism prediction model

[0507] Similar to the description of obtaining the prediction result for hyperthyroidism through the hyperthyroidism prediction model, the prediction result for hypothyroidism can be obtained through the hypothyroidism prediction model, and therefore, detailed description is omitted.

[0508] Obtaining prediction results for thyroid dysfunction

[0509] The predicted outcome for hyperthyroidism and the predicted outcome for hypothyroidism are combined to obtain a predicted outcome of dysthyroidism for the subject.

[0510] A prediction result for thyroid dysfunction considering a prediction result A for hyperthyroidism and a prediction result B for hypothyroidism can be obtained as shown in Table 2 below.

[0511] [Table 2] [Table 2]

[0512] On the other hand, when the predicted result for hyperthyroidism is hyperthyroidism and the predicted result for hypothyroidism is hypothyroidism, the output value (probability value) of the hyperthyroidism prediction model is compared with the output value (probability value) of the hypothyroidism prediction model, and the result with the output of the higher probability value can be adopted as the final result for thyroid dysfunction.

[0513] The heart rate measuring device 10 may perform all of the above steps, including obtaining a trigger signal in step S100; obtaining interval heart rates corresponding to a target day determined by the trigger signal in step S110; pre-processing the interval heart rates corresponding to the target day in step S120; obtaining hormone concentrations corresponding to a reference day in step S130; obtaining a pre-processing result for the interval heart rates corresponding to the reference day in step S140; obtaining a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S150; and processing the obtained difference and the hormone concentrations corresponding to the reference day by using a thyroid dysfunction prediction model in step S160 to obtain a thyroid dysfunction result for the subject.

[0514] The user terminal 20 may perform all of the above steps, including acquiring a trigger signal in step S100; acquiring interval heart rates corresponding to a target day determined by the trigger signal in step S110; pre-processing the interval heart rates corresponding to the target day in step S120; acquiring hormone concentrations corresponding to a reference day in step S130; acquiring a pre-processing result for the interval heart rates corresponding to the reference day in step S140; acquiring a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S150; and processing the acquired difference and the hormone concentrations corresponding to the reference day by using a thyroid dysfunction prediction model in step S160 to acquire a thyroid dysfunction result for the subject.

[0515] The server 30 may perform all of the above steps, including obtaining a trigger signal in step S100; obtaining interval heart rates corresponding to a target day determined by the trigger signal in step S110; pre-processing the interval heart rates corresponding to the target day in step S120; obtaining hormone concentrations corresponding to a reference day in step S130; obtaining a pre-processing result for the interval heart rates corresponding to the reference day in step S140; obtaining a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S150; and processing the obtained difference and the hormone concentrations corresponding to the reference day by using a thyroid dysfunction prediction model to obtain a thyroid dysfunction result for the subject in step S160.

[0516] The heart rate measuring device 10, the user terminal 20 and the server 30 may perform the above-mentioned steps in a suitably distributed manner, including obtaining a trigger signal in step S100; obtaining interval heart rates corresponding to a target day determined by the trigger signal in step S110; pre-processing the interval heart rates corresponding to the target day in step S120; obtaining hormone concentrations corresponding to a reference day in step S130; obtaining a pre-processing result for the interval heart rates corresponding to the reference day in step S140; obtaining a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S150; and processing the obtained difference and the hormone concentrations corresponding to the reference day by using a thyroid dysfunction prediction model in step S160 to obtain a thyroid dysfunction result for the subject.

[0517] For example, the user terminal 20 may execute the following steps: acquiring a trigger signal in step S100; acquiring an interval heart rate corresponding to a target date determined by the trigger signal in step S110; and pre-processing the interval heart rate corresponding to the target date in step S120. The server 30 may execute the following steps: acquiring a hormone concentration corresponding to a reference date in step S130; acquiring a pre-processing result for the interval heart rate corresponding to the reference date in step S140; acquiring a difference between the pre-processing result for the interval heart rate corresponding to the target date and the pre-processing result for the interval heart rate corresponding to the reference date in step S150; and processing the acquired difference and the hormone concentration corresponding to the reference date by using a thyroid dysfunction prediction model to acquire a thyroid dysfunction result for the subject in step S160.

[0518] As another example, the user terminal 20 may execute acquiring a trigger signal in step S100. The server 30 may execute the following steps: acquiring an interval heart rate corresponding to a target date determined by the trigger signal in step S110; pre-processing the interval heart rate corresponding to the target date in step S120; acquiring a hormone concentration corresponding to a reference date in step S130; acquiring a pre-processing result for the interval heart rate corresponding to the reference date in step S140; acquiring a difference between the pre-processing result for the interval heart rate corresponding to the target date and the pre-processing result for the interval heart rate corresponding to the reference date in step S150; and processing the acquired difference and the hormone concentration corresponding to the reference date by using a thyroid dysfunction prediction model to acquire a thyroid dysfunction result for the subject in step S160.

[0519] However, the form of the dispersion performance described above is not limited thereto, and the dispersion performance can be realized in various forms.

[0520] 5. Experimental Example

[0521] Below, experimental examples in which the thyroid dysfunction prediction model described in this application was trained and the results of the accuracy of the analysis are described.

[0522] The clinical data used in the experimental examples and the comparative examples described below were collected by the method described above, and a total of 1,027 clinical datasets collected from a total of 297 patients were used. Of these, a total of 168 clinical datasets were classified as having a diagnosis of "hyperthyroidism" for hormone concentrations, a total of 801 clinical datasets were classified as having a diagnosis of "normal", and a total of 58 clinical datasets were classified as having a diagnosis of "hypothyroidism".

[0523] (1) Comparative Example #1 - Heart Rate Difference to Normal Baseline - Labeling

[0524] Preparing the training dataset

[0525] In order to train a predictive model according to a comparative example for comparison with the experimental results of various embodiments described in this application, a training dataset was prepared in the following manner.

[0526] For each patient, the value obtained by subtracting the period average heart rate corresponding to the test day with the diagnosis "normal" from the period average heart rate corresponding to the test day with the diagnosis "abnormal" was used as the input value (e.g., the amount of change in the period average heart rate), and the diagnosis (hyperthyroidism or hypothyroidism) corresponding to the test day with the diagnosis "abnormal" was used as the labeling value. In addition, when there were two or more test days with the diagnosis "normal", the value obtained by subtracting the period average heart rate corresponding to one test day (a test day with the diagnosis "normal") from the period average heart rate corresponding to another test day (another test day with the diagnosis "normal") was used as the input value, and non-hypothyroidism or non-hyperthyroidism was used as the labeling value.

[0527] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0528] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0529] Here, 495 pieces of data obtained from 157 patients were used as clinical data for training, and a total of 1,199 training datasets were generated from them.

[0530] Training a hyperthyroidism prediction model

[0531] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0532] Testing a hyperthyroidism prediction model

[0533] Test datasets used to determine the accuracy of the trained predictive models were generated from 437 clinical data obtained from a total of 121 patients, generating a total of 1,138 test datasets.

[0534] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Comparative Example #1 are shown in Table 3 below.

[0535] [Table 3] [Table 3]

[0536] (2) Experimental Example #1 - Heart Rate Difference + Hormone Levels (vs. Comparative Example #1)

[0537] Preparing the training dataset

[0538] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0539] The model was trained according to the method for training a hyperthyroidism prediction model described above in embodiment #1 of the training method, i.e., a training dataset was generated in a format such as [(average heart rate over period corresponding to the first test day-average heart rate over period corresponding to the second test day), thyroid hormone concentration over period corresponding to the second test day, diagnosis result over period corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0540] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0541] In addition, with regard to the hormone levels used, both a method using only free T4 concentrations and a method using both free T4 and TSH concentrations were used.

[0542] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0543] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0544] Training a hyperthyroidism prediction model

[0545] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0546] Testing a hyperthyroidism prediction model

[0547] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0548] According to Example #1, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0549] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #1 are shown in Table 4 below.

[0550] [Table 4] [Table 4]

[0551] The inventors of the present application have discovered that, without determining a date when hormone levels are "normal" as the reference date, as in Comparative Example #1, and without using the difference in the average heart rate over a period between the reference date and the date under analysis, it is fully possible to predict whether or not there is "hyperthyroidism" by determining reference data, regardless of whether there is "normal" or "hyperthyroidism", and by using the difference in the average heart rate over a period between the reference data and the date under analysis together with the hormone concentration corresponding to the reference date, and have discovered that the results, such as accuracy and sensitivity, are much higher.

[0552] Therefore, the more training data sets are obtained from the same level of clinical data, the more accurately a predictive model is obtained that can predict thyroid dysfunction.

[0553] (3) Experimental example #2 - Heart rate difference + current heart rate + hormone level

[0554] Preparing the training dataset

[0555] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0556] The model was trained according to the method for training a hyperthyroidism prediction model described above in embodiment #2 of the training method, i.e., a training dataset was generated in a format such as [(average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day), average period heart rate corresponding to the first test day, thyroid hormone concentration corresponding to the second test day, diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0557] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0558] In addition, with regard to the hormone levels used, both a method using only free T4 concentrations and a method using both free T4 and TSH concentrations were used.

[0559] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0560] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0561] Training a hyperthyroidism prediction model

[0562] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0563] Testing a hyperthyroidism prediction model

[0564] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0565] According to Experimental Example #2, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0566] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #2 are shown in Table 5 below.

[0567] [Table 5] [Table 5]

[0568] The inventors of the present application have discovered that when the change in the period average heart rate for the target day and the period average heart rate (current period average heart rate) are used as input values, there is a slight increase in accuracy in predicting hyperthyroidism.

[0569] (4) Experimental Example #3 - Heart Rate Difference + Hormone Level + Standard Deviation

[0570] Preparing the training dataset

[0571] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0572] The model was trained according to the method for training a hyperthyroidism prediction model described above in embodiment #3 of the training method. That is, a training dataset was generated in a format such as [(average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day), (standard deviation of interval heart rate corresponding to the first test day-standard deviation of interval heart rate corresponding to the second test day), thyroid hormone concentration corresponding to the second test day, diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0573] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0574] In addition, for the hormone levels used, both free T4 and TSH concentrations were used as input values.

[0575] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0576] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0577] Training a hyperthyroidism prediction model

[0578] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0579] Testing a hyperthyroidism prediction model

[0580] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0581] According to Example #3, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0582] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #3 are shown in Table 6 below.

[0583] [Table 6] [Table 6]

[0584] The inventors of the present application have discovered that when the change in the period average heart rate for the target day and the period average heart rate (current period average heart rate) are used as input values, there is a slight increase in accuracy in predicting hyperthyroidism.

[0585] (5) Experimental Example #4 - Heart Rate Difference + Hormone Level + Standard Deviation Change + Skewness

[0586] Preparing the training dataset

[0587] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0588] According to the method of training a hyperthyroidism prediction model described above in embodiment #4 of the training method, the model was trained. However, standard deviation and skewness were used as parameters related to dispersion. That is, a training data set was generated in the format of [(average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day), (standard deviation of interval heart rate corresponding to the first test day-standard deviation of interval heart rate corresponding to the second test day), (skewness of interval heart rate corresponding to the first test day-skewness of interval heart rate corresponding to the second test day), thyroid hormone concentration corresponding to the second test day, diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training data set was used to train the model.

[0589] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0590] In addition, for the hormone levels used, both free T4 and TSH concentrations were used as input values.

[0591] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0592] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0593] Training a hyperthyroidism prediction model

[0594] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0595] Testing a hyperthyroidism prediction model

[0596] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0597] According to Example #4, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0598] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #4 are shown in Table 7 below.

[0599] [Table 7] [Table 7]

[0600] (6) Experimental example #5 - Heart rate difference + hormone level + standard deviation change + kurtosis

[0601] Preparing the training dataset

[0602] Compared to Experimental Example #4, in Experimental Example #5, the training and test datasets were generated in the same manner, except that instead of using standard deviation and skewness as parameters related to dispersion, standard deviation and kurtosis were used.

[0603] Training a hyperthyroidism prediction model

[0604] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0605] Testing a hyperthyroidism prediction model

[0606] According to Experimental Example #5, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0607] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #5 are shown in Table 8 below.

[0608] [Table 8] [Table 8]

[0609] (7) Experimental example #6 - Heart rate difference + hormone level + standard deviation change + JS divergence

[0610] Preparing the training dataset

[0611] Compared to Example #4, in Example #6, the training and test datasets were generated in an identical manner, except that instead of using standard deviation difference and skewness difference as the difference of variance-related parameters, JS divergence was used.

[0612] Training a hyperthyroidism prediction model

[0613] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0614] Testing a hyperthyroidism prediction model

[0615] According to Experimental Example #6, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0616] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #6 are shown in Table 9 below.

[0617] [Table 9] [Table 9]

[0618] As can be seen through Experimental Examples #4, 5, and 6, the inventors of the present application have discovered that when additional variables to observe the change in standard deviation and the change in variance of the interval heart rate are used as input values, there is a slight increase in accuracy in predicting hyperthyroidism.

[0619] (8) Experimental Example #7 - Rate of change of heart rate + hormone level + relative standard deviation change

[0620] Preparing the training dataset

[0621] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0622] According to the method for training a hyperthyroidism prediction model described above in embodiment #5 of the training method, the model was trained. That is, a training dataset was generated in a format such as [((average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day) / average period heart rate corresponding to the second test day), (relative standard deviation of interval heart rate corresponding to the first test day-relative standard deviation of interval heart rate corresponding to the second test day), thyroid hormone concentration corresponding to the second test day, diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used in training the model.

[0623] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0624] In addition, with regard to the hormone levels used, a method using both free T4 and TSH concentrations was used.

[0625] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0626] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0627] Training a hyperthyroidism prediction model

[0628] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0629] Testing a hyperthyroidism prediction model

[0630] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0631] According to Experimental Example #5, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0632] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #4 are shown in Table 10 below.

[0633] [Table 10] [Table 10]

[0634] Compared to Experimental Example #3, Experimental Example #5 is an experiment carried out under the same conditions, except that instead of using the change in average heart rate over a period and the change in standard deviation, the rate of change in average heart rate over a period and the change in relative standard deviation are used.

[0635] The inventors of the present application have discovered that the rate of change in the average heart rate over a period and the amount of change in the relative standard deviation can be used as input values.

[0636] (9) Experimental Example #8 - Rate of change of heart rate + hormone level + relative standard deviation change + number of days interval

[0637] Preparing the training dataset

[0638] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0639] According to the method for training a hyperthyroidism prediction model described above in embodiment #9 of the training method, the model was trained. That is, a training dataset was generated in a format such as [((average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day) / average period heart rate corresponding to the second test day), (relative standard deviation of interval heart rate corresponding to the first test day-relative standard deviation of interval heart rate corresponding to the second test day), the number of days between the second test day and the first test day, the thyroid hormone concentration corresponding to the second test day, the diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0640] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0641] In addition, with regard to the hormone levels used, a method using both free T4 and TSH concentrations was used.

[0642] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0643] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0644] Training a hyperthyroidism prediction model

[0645] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0646] Testing a hyperthyroidism prediction model

[0647] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0648] According to Example #8, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0649] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #8 are shown in Table 11 below.

[0650] [Table 11] [Table 11]

[0651] (10) Experimental Example #9 - Heart Rate Change Rate + Hormone Level + Relative Standard Deviation Change + Combined Value

[0652] Preparing the training dataset

[0653] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0654] According to the method for training a hyperthyroidism prediction model described above in embodiment #10 of the training method, the model was trained. That is, a training dataset was generated in a format such as [((average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day) / average period heart rate corresponding to the second test day), (relative standard deviation of interval heart rate corresponding to the first test day-relative standard deviation of interval heart rate corresponding to the second test day), the interval of days between the second test day and the first test day, the thyroid hormone concentration corresponding to the second test day, the combination value of the variables, the diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0655] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0656] In addition, with regard to the hormone levels used, a method using both free T4 and TSH concentrations was used.

[0657] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0658] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0659] Training a hyperthyroidism prediction model

[0660] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0661] Testing a hyperthyroidism prediction model

[0662] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0663] According to Experimental Example #9, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0664] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #9 are shown in Tables 12 and 13 below.

[0665] [Table 12] [Table 12]

[0666] [Table 13] [Table 13]

[0667] (11) Experimental Example #10 - Use of Subjects' Personal Information

[0668] Preparing the training dataset

[0669] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0670] The model was trained according to the method for training a hyperthyroidism prediction model described above in training method embodiment #11.

[0671] In particular, training was performed using a training data set such as (1) [(average heart rate over the period corresponding to the first test day - average heart rate over the period corresponding to the second test day), free T4 concentration over the second test day, age of the subject, sex of the subject, and diagnosis over the first test day (hyperthyroidism or non-hyperthyroidism)]. In addition, training was performed using a training data set such as (2) [(average heart rate over the period corresponding to the first test day - average heart rate over the period corresponding to the second test day) / average heart rate over the period corresponding to the second test day), free T4 concentration over the second test day, age of the subject, sex of the subject, and diagnosis over the first test day (hyperthyroidism or non-hyperthyroidism)]. Further, training was performed using a training data set such as (3) [(average heart rate for the first test day-average heart rate for the second test day), free T4 concentration for the second test day, TSH concentration for the second test day, age of the subject, sex of the subject, and diagnosis result for the first test day (hyperthyroidism or non-hyperthyroidism)]. In addition, training was performed using a training data set such as (4) [(average heart rate for the first test day-average heart rate for the second test day) / average heart rate for the second test day), free T4 concentration for the second test day, TSH concentration for the second test day, age of the subject, sex of the subject, and diagnosis result for the first test day (hyperthyroidism or non-hyperthyroidism)]. Finally, training was performed using a training dataset such as (5) [((period average heart rate corresponding to the first test day - period average heart rate corresponding to the second test day) / period average heart rate corresponding to the second test day), (relative standard deviation of interval heart rate corresponding to the first test day - relative standard deviation of interval heart rate corresponding to the second test day), (skewness of interval heart rate corresponding to the first test day - skewness of interval heart rate corresponding to the second test day), (kurtosis of interval heart rate corresponding to the first test day - kurtosis of interval heart rate corresponding to the second test day), (JS divergence between interval heart rate corresponding to the first test day and interval heart rate corresponding to the second test day), free T4 concentration corresponding to the second test day, TSH concentration corresponding to the second test day, subject's age, subject's sex, diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)].

[0672] Here, for the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test day, and the average of the obtained resting heart rates was calculated.

[0673] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0674] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0675] Training a hyperthyroidism prediction model

[0676] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0677] Testing a hyperthyroidism prediction model

[0678] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0679] According to Experimental Example #10, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0680] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #10 are shown in Table 14 below.

[0681] [Table 14] [Table 14]

[0682] (12) Experimental Example #11 - Experiment with Predetermined Period Length

[0683] Preparing the training dataset

[0684] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0685] The model was trained according to the method for training a hyperthyroidism prediction model described above in training method embodiment #10.

[0686] In particular, the change in relative standard deviation, the change in skewness, the change in kurtosis, and the JS divergence were used as parameters to determine the change in variance. Additionally, (free T4 concentration on the second reference day / JS divergence) was used as the combined value of the variables. That is, a training dataset was generated in a format such as [((average period heart rate corresponding to the first test day-average period heart rate corresponding to the second test day) / average period heart rate corresponding to the second test day), (relative standard deviation of interval heart rate corresponding to the first test day-relative standard deviation of interval heart rate corresponding to the second test day), (skewness of interval heart rate corresponding to the first test day-skewness of interval heart rate corresponding to the second test day), (kurtosis of interval heart rate corresponding to the first test day-kurtosis of interval heart rate corresponding to the second test day), (JS divergence between interval heart rate corresponding to the first test day and the second test day), the number of days between the first test day and the second test day, thyroid hormone concentration corresponding to the second test day (free T4 concentration on the second test day / JS divergence between interval heart rate corresponding to the second test day and the first test day), and diagnosis result corresponding to the first test day (hyperthyroidism or non-hyperthyroidism)], and the training dataset was used to train the model.

[0687] In Experiment #11, the predetermined period for calculating the period average heart rate corresponding to the test day was varied to 1, 5, 10, 15, 25 and 30 days to determine the effect on accuracy.

[0688] In addition, with regard to the hormone levels used, a method using both free T4 and TSH concentrations was used.

[0689] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0690] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0691] Training a hyperthyroidism prediction model

[0692] Using the Wright Gradient Boosting Machine, a hyperthyroidism prediction model was trained on the training dataset described above.

[0693] Testing a hyperthyroidism prediction model

[0694] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0695] According to Experimental Example #11, the trained hyperthyroidism prediction model was tested for accuracy using the test set described above.

[0696] The accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the hyperthyroidism prediction model according to Experimental Example #11 are shown in Table 15 below.

[0697] [Table 15] [Table 15]

[0698] The inventors of the present application have discovered that the highest accuracy in predicting hyperthyroidism is achieved when 15 days is used as the predetermined period for calculating interval heart rates.

[0699] A method for training a thyroid dysfunction prediction model, a method for predicting thyroid dysfunction by using the model, and a system thereof have been described.

[0700] Below, a method for training a model for predicting thyroid hormone levels, a method for predicting thyroid dysfunction by using the model, and a system thereof are briefly described.

[0701] 6. How to train a thyroid hormone prediction model

[0702] Training Method Embodiment #12. Difference Values ​​from Heart Rate on Reference Day and Hormone Values ​​on Reference Day

[0703] According to a twelfth exemplary embodiment described in the present application, to train a learning model, a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day and the hormones on the first test day may be used as input values, and the hormone concentrations corresponding to the second test day may be used as labeling values.

[0704] Here, a machine learning model capable of regression may be used as the learning model. For example, a light gradient boosting machine, a support vector machine, a random forest, an extra tree, an adaboost, an extreme gradient boost, a CatBoost, or the like may be used as the learning model.

[0705] The thyroid hormone concentration prediction model includes at least one selected from the group consisting of a TRH concentration prediction model, a TSH concentration prediction model, a T4 concentration prediction model, a free T4 concentration prediction model, a T3 concentration prediction model, and a free T3 concentration prediction model for the target day.

[0706] In calculating the period average heart rate, a resting heart rate for a predetermined period based on a particular date may be used, where the predetermined period may be one of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 days.

[0707] According to the type of thyroid hormone concentration prediction model, the concentration value corresponding to the type among the multiple hormone concentrations corresponding to the first test date can be used as the labeling value.For example, when training a free T4 concentration prediction model, the free T4 concentration value corresponding to the first test date is used as the labeling value.As another example, when training a TSH concentration prediction model, the TSH concentration value corresponding to the first test date is used as the labeling value.

[0708] For convenience of explanation, the following description is given based on training a free T4 concentration prediction model.

[0709] According to a twelfth exemplary embodiment, to train a free T4 concentration prediction model, the secured clinical data may be used to generate a training data set in a format such as [(period average heart rate corresponding to the first test day-period average heart rate corresponding to the second test day), thyroid hormone concentration corresponding to the second test day, free T4 concentration corresponding to the first test day], where the thyroid hormone concentration corresponding to the second test day may be one selected from the group of TRH concentration, TSH concentration, T4 concentration, free T4 concentration, T3 concentration, and free T3 concentration, or may be a combination thereof.

[0710] For example, assuming that a total of three datasets are obtained from a first patient, and for a first test, the determined free T4 hormone concentration is a first free T4 concentration and the determined TSH concentration is a first TSH concentration, for a second test, the determined free T4 hormone concentration is a second free T4 concentration and the determined TSH concentration is a second TSH concentration, and for a third test, the determined free T4 hormone concentration is a third free T4 concentration and the determined TSH concentration is a third TSH concentration, the training datasets that may be obtained from the patient's clinical data may be as follows:

[0711] [(average heart rate during the period corresponding to the first test day - average heart rate during the period corresponding to the second test day), second free T4 concentration, second TSH concentration, first free T4 concentration]

[0712] [(average heart rate during the period corresponding to the first test day - average heart rate during the period corresponding to the third test day), third free T4 concentration, third TSH concentration, first free T4 concentration]

[0713] [(average heart rate during the period corresponding to the second test day - average heart rate during the period corresponding to the first test day), first free T4 concentration, first TSH concentration, second free T4 concentration]

[0714] [(average heart rate during the period corresponding to the second test day - average heart rate during the period corresponding to the third test day), third free T4 concentration, third TSH concentration, third free T4 concentration]

[0715] [(average heart rate for the period corresponding to the third test day - average heart rate for the period corresponding to the first test day), first free T4 concentration, first TSH concentration, third free T4 concentration]], and

[0716] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), second free T4 concentration, second TSH concentration, third free T4 concentration]

[0717] As another example, assuming that a total of four datasets are obtained from a second patient, and for a first test, the determined free T4 hormone concentration is a first free T4 concentration and the determined TSH concentration is a first TSH concentration, for a second test, the determined free T4 hormone concentration is a second free T4 concentration and the determined TSH concentration is a second TSH concentration, for a third test, the determined free T4 hormone concentration is a third free T4 concentration and the determined TSH concentration is a third TSH concentration, and for a fourth test, the determined free T4 hormone concentration is a fourth free T4 concentration and the determined TSH concentration is a fourth TSH concentration, training datasets that may be obtained from the patient's clinical data may be as follows:

[0718] [(average heart rate during the period corresponding to the first test day - average heart rate during the period corresponding to the second test day), second free T4 concentration, second TSH concentration, first free T4 concentration]

[0719] [(average heart rate during the period corresponding to the first test day - average heart rate during the period corresponding to the third test day), third free T4 concentration, third TSH concentration, first free T4 concentration]

[0720] [(average heart rate during the period corresponding to the first test day - average heart rate during the period corresponding to the fourth test day), fourth free T4 concentration, fourth TSH concentration, first free T4 concentration]

[0721] [(average heart rate during the period corresponding to the second test day - average heart rate during the period corresponding to the first test day), first free T4 concentration, first TSH concentration, second free T4 concentration]

[0722] [(average heart rate during the period corresponding to the second test day - average heart rate during the period corresponding to the third test day), third free T4 concentration, third TSH concentration, second free T4 concentration]

[0723] [(average heart rate during the period corresponding to the second test day - average heart rate during the period corresponding to the fourth test day), fourth free T4 concentration, fourth TSH concentration, second free T4 concentration]

[0724] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the first test day), first free T4 concentration, first TSH concentration, third free T4 concentration]]

[0725] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the second test day), second free T4 concentration, second TSH concentration, third free T4 concentration]

[0726] [(average heart rate during the period corresponding to the third test day - average heart rate during the period corresponding to the fourth test day), fourth free T4 concentration, fourth TSH concentration, third free T4 concentration]

[0727] [(average heart rate during the period corresponding to the 4th test day - average heart rate during the period corresponding to the 1st test day), 1st free T4 concentration, 1st TSH concentration, 4th free T4 concentration]

[0728] [(average heart rate for the period corresponding to the fourth test day - average heart rate for the period corresponding to the second test day), second free T4 concentration, second TSH concentration, fourth free T4 concentration], and

[0729] [(average heart rate during the period corresponding to the 4th test day - average heart rate during the period corresponding to the 3rd test day), 3rd free T4 concentration, 3rd TSH concentration, 4th free T4 concentration]

[0730] When it is intended to train a model for predicting TSH concentrations rather than a model for predicting free T4 concentrations, the input values ​​in the training data set described above are maintained and TSH concentrations are used as labeling values ​​instead of free T4 concentrations.

[0731] Training Method Embodiment #13. Percentage Change in Heart Rate on Reference Days and Hormone Levels on Reference Days

[0732] According to the thirteenth exemplary embodiment described in the present application, in comparison with the twelfth exemplary embodiment described above, instead of a value obtained by subtracting the period average heart rate corresponding to the second test day from the period average heart rate corresponding to the first test day, the rate of change of the period average heart rate of the second test day relative to the first test day is used as an input value, the rest being identical to the twelfth exemplary embodiment.

[0733] Training Method Embodiment #14. Percentage change in heart rate, relative standard deviation change, and hormone levels on baseline

[0734] According to the fourteenth exemplary embodiment described in the present application, in comparison with the thirteenth exemplary embodiment described above, a value obtained by subtracting the relative standard deviation of the interval heart rate corresponding to the second test day from the relative standard deviation of the interval heart rate corresponding to the first test day is further used as an input value, and the rest is identical to the thirteenth exemplary embodiment.

[0735] Training Method Embodiment #15. Percentage change in heart rate, change in relative standard deviation, change in skewness, change in kurtosis, JS divergence, and hormone values ​​on the reference day

[0736] According to the fifteenth exemplary embodiment described in the present application, compared to the fourteenth exemplary embodiment described above, a value related to the difference between a parameter related to the variance of the interval heart rate corresponding to the first test day and a parameter related to the variance of the interval heart rate corresponding to the second test day is further used as an input value, and the rest is identical to the fourteenth exemplary embodiment.

[0737] Training Method Embodiment #16. Percentage change in heart rate, change in relative standard deviation, change in skewness, change in kurtosis, JS divergence, and hormone values ​​+ days interval on the reference day

[0738] According to the sixteenth exemplary embodiment described in the present application, compared to the fifteenth exemplary embodiment described above, a value related to the number of days between the first test date and the second test date is further used as an input value, and the rest is identical to the fifteenth exemplary embodiment.

[0739] 7. Method for predicting thyroid dysfunction based on a thyroid hormone prediction model

[0740] The method for predicting thyroid dysfunction described in the present application can be performed by the system 1 described above.

[0741] FIG. 6 is a flow chart illustrating the method for predicting thyroid dysfunction described herein.

[0742] Referring to FIG. 6, a method described herein for predicting thyroid dysfunction for a subject includes: acquiring a trigger signal in step S200; acquiring interval heart rates corresponding to a target day determined by the trigger signal in step S210; pre-processing the interval heart rates corresponding to the target day in step S220; acquiring hormone concentrations corresponding to a reference day in step S230; acquiring a pre-processing result for the interval heart rates corresponding to the reference day in step S240; acquiring a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S250; processing the acquired difference and the hormone concentrations corresponding to the reference day by using a thyroid hormone concentration prediction model to obtain a prediction result of thyroid hormone concentration for the subject in step S260; and acquiring a thyroid dysfunction result for the subject based on the prediction result for hormone concentration in step S270.

[0743] Here, the steps including obtaining a trigger signal in step S200, obtaining an interval heart rate corresponding to a target day determined by the trigger signal in step S210, preprocessing the interval heart rate corresponding to the target day in step S220, obtaining a hormone concentration corresponding to a reference day in step S230, obtaining a preprocessing result for the interval heart rate corresponding to the reference day in step S240, and obtaining a difference between the preprocessing result for the interval heart rate corresponding to the target day and the preprocessing result for the interval heart rate corresponding to the reference day in step S250 are respectively set as steps S100 and S210. 5, including obtaining a trigger signal in step S110, obtaining interval heart rates corresponding to a target day determined by the trigger signal in step S110, pre-processing the interval heart rates corresponding to the target day in step S120, obtaining hormone concentrations corresponding to a reference day in step S130, obtaining a pre-processing result for the interval heart rates corresponding to the reference day in step S140, and obtaining a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S150. Therefore, detailed description is omitted.

[0744] Obtaining a prediction result for the thyroid hormone concentration through the thyroid hormone concentration prediction model in step S260.

[0745] Obtaining a prediction result for the first thyroid hormone concentration through a first thyroid hormone concentration prediction model

[0746] A first thyroid hormone concentration for the subject can be obtained through a first thyroid hormone concentration prediction model. For example, the first thyroid hormone can be TRH.

[0747] Obtaining prediction results for the second thyroid hormone concentration through a second thyroid hormone concentration prediction model

[0748] A second thyroid hormone concentration for the subject can be obtained through a second thyroid hormone concentration predictive model. For example, the second thyroid hormone can be TSH.

[0749] Obtaining prediction results for third thyroid hormone concentration through a third thyroid hormone concentration prediction model

[0750] A third thyroid hormone concentration for the subject can be obtained through a third thyroid hormone concentration prediction model. For example, the third thyroid hormone can be T4.

[0751] Obtaining prediction results for thyroid hormone concentration through a thyroid hormone concentration prediction model

[0752] A fourth thyroid hormone concentration for the subject can be obtained through a fourth thyroid hormone concentration prediction model. For example, the fourth thyroid hormone can be free T4.

[0753] Obtaining prediction results for thyroid hormone V concentration through a thyroid hormone V concentration prediction model

[0754] A fifth thyroid hormone concentration for the subject can be obtained through a fifth thyroid hormone concentration prediction model. For example, the fifth thyroid hormone can be T3.

[0755] Obtaining prediction results for thyroid hormone VI concentration through a thyroid hormone VI concentration prediction model

[0756] The sixth thyroid hormone concentration for the subject can be obtained through a sixth thyroid hormone concentration prediction model. For example, the sixth thyroid hormone can be free T3.

[0757] As described above, one or a combination selected from the group consisting of thyroid hormone concentrations 1 through 6 may be obtained.

[0758] Obtaining a thyroid dysfunction outcome for the subject based on the predicted outcome for the hormone concentration in step S270.

[0759] Based on the obtained predicted values ​​for hormone concentrations, it can be determined whether the subject is in a hyperthyroid state, a hypothyroid state, or a euthyroid state.

[0760] Since ranges corresponding to "normal", "hyperthyroidism" and "hypothyroidism" are predetermined for each thyroid hormone concentration, predetermined ranges may be used.

[0761] The heart rate measuring device 10, the user terminal 20 and the server 30 may perform the above-mentioned steps in a suitably distributed manner, including: acquiring a trigger signal in step S200; acquiring interval heart rates corresponding to a target day determined by the trigger signal in step S210; pre-processing the interval heart rates corresponding to the target day in step S220; acquiring hormone concentrations corresponding to a reference day in step S230; acquiring a pre-processing result for the interval heart rates corresponding to the reference day in step S240; acquiring a difference between the pre-processing result for the interval heart rates corresponding to the target day and the pre-processing result for the interval heart rates corresponding to the reference day in step S250; processing the acquired difference and the hormone concentrations corresponding to the reference day by using a thyroid hormone concentration prediction model in step S260 to obtain a prediction result of thyroid hormone concentration for the subject; and acquiring a result of thyroid dysfunction for the subject based on the prediction result for hormone concentrations in step S270.

[0762] 8. Experimental Example

[0763] Experimental Example #12

[0764] Preparing the training dataset

[0765] A total of 563 clinical datasets collected from a total of 171 patients were used.

[0766] According to the method for training a thyroid hormone concentration predictive model described above in embodiment #13 of the training method, a model was trained. In particular, a predictive model for predicting free T4 concentration was trained.

[0767] Meanwhile, the free T4 and TSH concentrations were used as the hormone concentrations corresponding to each test day, and with respect to the period average heart rate corresponding to the test day, the resting heart rates for 10 days were obtained based on the test days, and the average of the obtained resting heart rates was calculated.

[0768] Here, the training dataset was not generated using "differences" of clinical data between different patients, but the training dataset was generated using a clinical dataset obtained from one patient.

[0769] Here, 563 pieces of data from 171 patients were used as clinical data for training, and a total of 1,542 training datasets were generated from them.

[0770] Training a hyperthyroidism prediction model

[0771] Using ExtraTree, a free T4 concentration prediction model was trained on the training dataset described above.

[0772] Testing a hyperthyroidism prediction model

[0773] The test datasets used to determine the accuracy of the trained predictive models were generated from 464 clinical data sets secured from a total of 126 patients, generating a total of 1,416 test datasets.

[0774] According to Experimental Example #13, the trained free T4 concentration prediction model was tested for accuracy using the test set described above.

[0775] The mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the free T4 concentration prediction model according to Experimental Example #13 are shown in Table 16 below.

[0776] [Table 16] [Table 16]

[0777] Experimental Example #13

[0778] The training dataset, test dataset, and generation method used in Example #13 are identical to those described in Example #12.

[0779] However, unlike example #12, in example #13, the predictive model was trained according to training method embodiment #14 instead of training the predictive model according to training method embodiment #13.

[0780] Testing a hyperthyroidism prediction model

[0781] According to Experimental Example #13, the trained free T4 concentration prediction model was tested for accuracy using the test set described above.

[0782] The mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the free T4 concentration prediction model according to Experimental Example #13 are shown in Table 17 below.

[0783] [Table 17] [Table 17]

[0784] Experimental Example #14

[0785] The training dataset, test dataset, and generation method used in Example #14 are identical to those described in Example #12.

[0786] However, unlike example #12, in example #14 the predictive model was trained according to training method embodiment #15 instead of training the predictive model according to training method embodiment #13.

[0787] Testing a hyperthyroidism prediction model

[0788] According to Experimental Example #14, the trained free T4 concentration prediction model was tested for accuracy using the test set described above.

[0789] The mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the free T4 concentration prediction model according to Experimental Example #14 are shown in Table 18 below.

[0790] [Table 18] [Table 18]

[0791] Experimental Example #15

[0792] The training dataset, test dataset, and generation method used in Example #15 are identical to those described in Example #12.

[0793] However, unlike example #12, in example #15 the predictive model was trained according to training method embodiment #16 instead of training the predictive model according to training method embodiment #13.

[0794] Testing a hyperthyroidism prediction model

[0795] According to Experimental Example #15, the trained free T4 concentration prediction model was tested for accuracy using the test set described above.

[0796] The mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the free T4 concentration prediction model according to Experimental Example #15 are shown in Table 19 below.

[0797] [Table 19] [Table 19]

[0798] 1: System

[0799] 10: Heart rate measuring device

[0800] 20: User terminal

[0801] 30: Server

Claims

1. A method for predicting thyroid dysfunction in a subject, executed by one or more processors, comprising: Determining a target date; Obtaining the interval heart rate corresponding to the determined target date; For the subject, obtaining a first preprocessing result of the obtained interval heart rate corresponding to the determined target date, wherein the first preprocessing result includes at least one parameter related to the average of the interval heart rates corresponding to the target date; Determining a reference date, wherein the reference date is selected from one or more test dates on which the subject undergoes a hormone test, regardless of whether the thyroid function of the subject is normal or abnormal; For the subject, obtaining at least one of the concentrations of hormones related to the thyroid corresponding to the reference date; For the subject, obtaining a second preprocessing result of the interval heart rate corresponding to the reference date, wherein the second preprocessing result includes at least one parameter related to the average of the interval heart rates corresponding to the reference date; Obtaining a set of input values based on the first preprocessing result, the second preprocessing result, and the at least one of the concentrations of hormones related to the thyroid corresponding to the reference date; and Generating a prediction result for thyroid dysfunction based on the set of input values wherein the prediction result is generated by applying the set of input values to a trained thyroid dysfunction prediction model. Method.

2. The method according to claim 1, wherein the set of input values includes one selected from the group consisting of 1) the amount of change in the average of the interval heart rates of the target date with respect to the reference date, and 2) the rate of change in the average of the interval heart rates of the target date with respect to the reference date, or a combination thereof.

3. The first preprocessing result further includes at least one parameter related to the variance of the interval heart rate corresponding to the target date, wherein the at least one parameter related to the variance of the interval heart rate corresponding to the target date includes at least one of the standard deviation, skewness, kurtosis, and JS divergence of the interval heart rate corresponding to the target date, The second preprocessing result further includes at least one parameter related to the variance of the interval heart rate corresponding to the reference date, Here, the at least one parameter related to the variance of the interval heart rate corresponding to the reference date includes at least one of the standard deviation, skewness, kurtosis, and JS divergence of the interval heart rate corresponding to the reference date, according to the method of claim 1.

4. The set of input values includes 1) the amount of change in the average of the interval heart rate of the target date relative to the reference date, 2) the rate of change in the average of the interval heart rate of the target date relative to the reference date, or a combination thereof, and the set of input values further includes 3) the amount of change in the standard deviation of the interval heart rate of the target date relative to the reference date, 4) the amount of change in the relative standard deviation of the interval heart rate of the target date relative to the reference date, 5) the amount of change in the skewness of the interval heart rate of the target date relative to the reference date, 6) the amount of change in the kurtosis of the interval heart rate of the target date relative to the reference date, 7) one selected from the group consisting of the JS divergence between the interval heart rate corresponding to the reference date and the interval heart rate corresponding to the target date, or a combination thereof, according to the method of claim 3.

5. The at least one of the concentrations of the hormones related to the thyroid corresponding to the reference date is one selected from the group consisting of 1) the concentration of thyroid stimulating hormone (TSH), 2) the concentration of tetraiodothyronine (T4), 3) the concentration of free T4 in serum, 4) the concentration of triiodothyronine (T3), 5) the concentration of free T3 in serum, 6) the concentration of thyrotropin releasing hormone (TRH), or a combination thereof, according to the method of any one of claims 1 to 4.

6. The method further comprises: 1) The amount of change in the average of the interval heart rates on the target date with respect to the reference date; 2) The rate of change in the average of the interval heart rates on the target date with respect to the reference date; 3) The amount of change in the standard deviation of the interval heart rates on the target date with respect to the reference date; 4) The amount of change in the relative standard deviation of the interval heart rates on the target date with respect to the reference date; 5) The amount of change in the skewness of the interval heart rates on the target date with respect to the reference date; 6) The amount of change in the kurtosis of the interval heart rates on the target date with respect to the reference date; 7) The JS divergence between the interval heart rates corresponding to the reference date and the interval heart rates corresponding to the target date; 8) The concentration of thyroid-stimulating hormone (TSH) of the subject corresponding to the reference date; 9) The concentration of tetraiodothyronine (T4) of the subject corresponding to the reference date; 10) The concentration of free T4 in the serum of the subject corresponding to the reference date; 11) The concentration of triiodothyronine (T3) of the subject corresponding to the reference date; 12) The concentration of free T3 in the serum of the subject corresponding to the reference date; 13) The step of obtaining a first value and a second value from the group consisting of the concentration of thyrotropin-releasing hormone (TRH) of the subject corresponding to the reference date i) The value obtained by multiplying the first value by the second value; ii) The value obtained by dividing the first value by the second value; and iii) The step of obtaining any one of the values obtained by dividing the second value by the first value The method according to any one of claims 1 to 4, comprising the above.

7. The method according to claim 6, wherein the set of input values further includes any one of the obtained values.

8. The method further comprises: The step of obtaining the number of days interval between the target date and the reference date The method according to any one of claims 1 to 4, comprising the above.

9. The method according to claim 8, wherein the set of input values further includes the number of days interval between the target date and the reference date.

10. The interval heart rate corresponding to the target date is all the resting heart rates corresponding to a predetermined period based on the target date, wherein the interval heart rate corresponding to the reference date is all the resting heart rates corresponding to the predetermined period based on the reference date. The method according to any one of claims 1 to 4.

11. The method according to claim 10, wherein the predetermined period is any one selected from 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, and 17 days.

12. The method according to any one of claims 1 to 4, wherein the trained thyroid dysfunction prediction model includes a hyperthyroidism prediction model and a hypothyroidism prediction model.

13. The method according to claim 12, wherein the prediction result for thyroid dysfunction is determined by considering a first prediction result obtained by processing the set of input values by the hyperthyroidism prediction model and a second prediction result obtained by processing the set of input values by the hypothyroidism prediction model.

Images