KR20240076868A - System and method for estimating the possibility of obesity based on physical growth data using growth prediction ai model - Google Patents

Abstract

A method of predicting obesity based on infant growth and development data using a growth prediction AI model according to an embodiment of the present invention, wherein the method uses first data, which is physical information data including gender, height, and weight for infants and toddlers. Obtaining and storing in a database, Obtaining second data, which is physical activity data for infants and toddlers, and storing in the database, Normalizing and categorizing the first and second data, Normalizing the first and second data And predicting obesity after n months through the infant growth curve of the growth prediction AI model based on the second data, reacquiring the first and second data after the n months and storing them in a database, the prediction A step of training a growth prediction AI model based on the obtained obesity level and the re-acquired first data, predicting the obesity level after m months through the learned growth prediction AI model, and analyzing the predicted obesity level to determine the risk of obesity. It is characterized by including.

Description

Method and system for predicting obesity based on infant growth and development data using growth prediction AI model {SYSTEM AND METHOD FOR ESTIMATING THE POSSIBILITY OF OBESITY BASED ON PHYSICAL GROWTH DATA USING GROWTH PREDICTION AI MODEL}

The present invention generally relates to a system and method for predicting obesity based on growth and development data of infants and children (hereinafter referred to as 'infants'), and more specifically, to a system and method for predicting obesity, not to a medical diagnosis method, but rather to anthropometry data and physical activity. This relates to a system and method for predicting more accurate and standardized obesity risk through a growth prediction AI model using data.

Obesity in infants and young children refers to a state of being relatively obese compared to the same peer group, and body mass index (BMI), which is generally determined by height and weight, is known to be a common and easy way to determine the degree of obesity.

Obesity is divided into primary obesity, which is caused by genetic or environmental factors, and secondary obesity, which is caused by central nervous system abnormalities, endocrine diseases, drugs, etc., and more than 99% of cases are primary obesity. In fact, if both parents are obese, there is a 70-80% chance that the child will be obese, and if the mother is obese, the risk of obesity in the child is known to be doubled. Recently, the prevalence of obesity is increasing as the amount of activity among adolescents has drastically decreased due to the westernization of eating habits and excessive education.

In particular, obesity among children and adolescents, including infants and toddlers, appears to have increased rapidly due to a decrease in physical activity due to the COVID-19 incident that occurred in 2020. According to the National Health Insurance Corporation, the amount of obesity treatment for infants and children under 9 years old increased by 45.3% in the first half of 2021 compared to the first half of 2019 before COVID-19, and the number of teenagers also increased by 29.6%.

Obesity in children and adolescents, including infants and toddlers, leads to obesity in adulthood and is a factor in the early development of metabolic diseases such as cardiovascular disease and diabetes. Furthermore, it affects the psychosocial development of children and adolescents, and problems such as low self-esteem, depression, alienation from peers, and emotional anxiety due to obesity have been reported.

Despite these circumstances, society is still not properly aware of the seriousness of the problem of obesity in infants and young children. Moreover, to date, advice on the future risk of obesity in infants and young children can only be received through traditional medical services, where a doctor, a clinical expert, analyzes growth and development data such as height and weight and provides counseling to the child and guardian through a direct visit to the hospital. . According to this traditional service method, an expert must directly diagnose and explain obesity, which not only limits monitoring a large number of subjects, but also creates a gap depending on each expert's clinical ability in obesity screening.

Therefore, there is a need for a system that can prevent systemic diseases such as obesity and high blood pressure in adulthood by developing a standardized and systemized infant obesity model to identify infants who are currently obese or at high risk of obesity at an early stage and take appropriate measures.

The present invention is intended to solve the above problems, and uses physical information data such as height and weight of infants and physical activity data such as data on endurance, agility, balance, and agility to determine the infant's obesity level or risk of obesity at a certain point in time. The purpose is to provide a system and method for early identification of infants and young children at high risk of obesity by providing a method to predict.

Another object of the present invention is to provide a system that enables more accurate prediction of obesity by adjusting the weight of the predictor variable from the predicted value of the infant's obesity at a given point in time and the actual obesity derived from the infant's physical information data acquired at the given time. It provides a method.

Other objects, specific advantages and novel features of the present invention will become more apparent from the following detailed description and preferred embodiments taken in conjunction with the accompanying drawings.

A method of predicting obesity based on growth and development data of infants and toddlers using a growth prediction AI model according to an embodiment of the present invention to achieve the above object, the method includes body weight including gender, height and weight for infants and toddlers. Obtaining first data, which is information data, and storing it in a database; Obtaining second data, which is physical activity data for infants and toddlers, and storing it in the database; Normalizing and categorizing the first and second data; , predicting obesity after n months through the infant growth curve of the growth prediction AI model based on the normalized first and second data, reacquiring the first and second data after n months and storing them in the database. storing, training a growth prediction AI model based on the predicted obesity and the re-acquired first data, predicting obesity after m months through the learned growth prediction AI model, and predicting the predicted obesity. It is characterized by including the step of analyzing the risk of obesity as a basis.

According to another embodiment of the present invention, the predicted value of obesity is a value obtained by calculating a predicted value based on the first data and a predicted value based on the second data, and adding the respective predicted values according to their respective weights. , The step of learning the growth prediction AI model is characterized by comparing the predicted obesity level and the re-acquired first data, and adjusting the weight used to predict the obesity level according to the comparison result. The weight is a ratio of the predicted value based on the first data and the predicted value based on the second data for the predicted value of obesity.

According to another embodiment of the present invention, the second data includes data on endurance, agility, balance, and agility.

According to another embodiment of the present invention, classified category information and/or analyzed obesity risk information is provided to the user in a visualized form on a display. and recommending a diet or physical activity for infants and young children based on the classified category information or the analyzed obesity risk information.

According to another embodiment of the present invention, the first and second data are acquired periodically, and the cycle includes yearly, quarterly, monthly, weekly, etc.

A system for predicting obesity based on infant growth and development data using an artificial intelligence algorithm according to another embodiment of the present invention includes a terminal that provides physical information data and physical activity data including the user's gender, height, and weight, and the above. It is configured to include an obesity prediction server that is connected to a terminal and a network, analyzes the body information data and physical activity data provided from the terminal, predicts obesity at a certain point in time, and provides the obesity information to the terminal, The obesity prediction server includes a storage unit that stores the data provided from the terminal in a database, an obesity prediction calculation unit that learns the data in the storage unit using artificial intelligence (AI) to predict the user's obesity level at a predetermined point in time, and an obesity prediction calculation unit. It is characterized by comprising an obesity information providing unit that provides obesity information predicted in to the terminal.

According to another embodiment of the present invention, an obesity prediction calculation unit includes a first prediction unit that predicts body information at a certain point in time based on the body information data from the terminal, and a first prediction unit that predicts body information at a certain point in time based on the body information data from the terminal. a second prediction unit that predicts body information at a point in time; a weight determination unit that determines a weight given to each of the predicted values of the first prediction unit and the second prediction unit; and the obesity level at the predetermined time point is determined by the first prediction unit and the second prediction unit. It is characterized in that it can be predicted based on the predicted value from the second prediction unit and the weight from the weight determination unit.

Another embodiment of the present invention includes a computer-readable recording medium embedded with a program that operates to perform the obesity prediction method according to the present invention, or a program stored in the recording medium.

According to the embodiment of the present invention as described above, the obesity level or risk of obesity of an infant at a certain point in time can be predicted using physical information data such as the infant's height and weight, and physical activity data such as data on endurance, agility, balance, and agility. By providing a method, infants and young children at high risk of obesity can be identified early on.

In addition, the present invention enables more accurate prediction of obesity by adjusting the weight of the predictor variable from the predicted value of the infant's obesity at a certain point in time and the actual obesity derived from the infant's physical information data acquired at the given time.

In addition, the present invention can be applied as an AI screening model for large-scale infant health care projects such as infant physical fitness tests and health checkups, and can also be used as an obesity risk clinical decision-making system and patient management system for hospitals such as children's hospitals and growth and development clinics. ,

1 is a flowchart showing a process for predicting obesity in infants and young children according to an embodiment of the present invention;
Figure 2 is an exemplary diagram of an information source that can obtain body data according to an embodiment of the present invention;
Figure 3 is a distribution chart of body mass index according to an embodiment of the present invention;
Figure 4 is a distribution chart in which the average body mass index in Figure 3 is normalized to 100,
Figure 5 is a distribution chart showing normalized flexibility items,
Figure 6 is a distribution chart showing normalized endurance items,
Figure 7 is a distribution chart showing the normalization of agility items,
Figure 8 is a distribution chart showing normalization of balance items,
Figure 9 is a distribution chart showing normalized agility items,
Figure 10 is a distribution chart showing the normalization of the sum of the normalized values of Figures 5 to 9 converted to 100 points;
Figure 11 is a table showing the correlation between physical information data and physical activity data;
Figure 12 is a diagram showing an obesity prediction process according to an embodiment of the present invention;
Figure 13 is a diagram illustrating an example of providing analysis data on information about the height of an infant or young child to a user according to an embodiment of the present invention.

The terms or words used in the present invention described below should not be construed as limited to their usual or dictionary meanings, and the inventor may appropriately define the concept of the term in order to explain his/her invention in the best way. It must be interpreted with meaning and concept consistent with the technical idea of the present invention based on the principle that it can be done.

Therefore, the embodiments described in this specification and the configurations shown in the drawings are only preferred embodiments of the present invention and do not represent the entire technical idea of the present invention, and therefore various equivalents and It should be understood that variations may exist.

Hereinafter, a method and system for predicting obesity in infants and young children according to a preferred embodiment of the present invention will be described in detail with reference to the attached drawings.

1 is a flowchart illustrating a process for predicting obesity in infants and young children according to an embodiment of the present invention.

First, the infant's physical data is obtained and converted into a database (S10). Physical data includes physical information data including gender, height, and weight, and physical activity data such as flexibility, endurance, agility, balance, and agility. The body information data may further include gender, age (age), as well as any body data that can be measured using an InBody device, such as BMI, body fat, muscle mass, and thigh circumference. In this embodiment, the physical activity data includes data on flexibility, endurance, quickness, balance, and agility, but is not limited thereto.

The physical information data and physical activity data of infants and toddlers must be databased in advance for statistical analysis or the creation and learning of a growth prediction AI model, which will be described later. If the physical data cannot be obtained directly from infants or young children, information as shown in FIG. 2 can be used.

The body information data and the body activity data are normalized (S20).

An example of normalization is explained using Seoul City infant examination data for three years from 2017 to 2019. The number of samples was 19,513, and the analysis targets were age (age in months), height (height), weight, flexibility, endurance, agility, balance, and agility. If there was no data for any of the above analysis targets, they were excluded from the analysis. Flexibility was assessed by the degree of back bending, endurance was assessed by the time maintained in the V-shaped posture, agility was assessed by long jumping in place, balance was assessed by the time standing on one foot, and agility was assessed by the round trip time of a predetermined section. However, it is not limited to this, and there may be various evaluation methods.

Figure 3 is a distribution chart of body mass index (BMI), and Figure 4 is a distribution chart in which the average of the body mass index is normalized to 100. The average age is 65 months, height is 111 cm, weight is 19.5 kg, and BMI is 15.8. Actual body mass index needs to be considered because the standard distribution varies depending on the number of months and gender, but was not considered in this example.

Figures 5 to 9 are distribution charts showing five items of physical activity data, flexibility, endurance, agility, balance, and agility, normalized to (0,1). In the case of agility, it was inverted so that higher scores were superior. Quintiles (divided into 5 sections of 20% each) are also displayed. Figure 10 is a distribution chart showing the sum of the normalized values of flexibility, endurance, agility, balance, and agility converted to a score of 0 to 100 and normalized to (0,1).

Figure 11 is a table showing an example of the correlation between body information data and physical activity data. As seen in the table above, physical information data (age, height, weight) shows a relatively high correlation of 0.41 to 0.76. In other words, as age increases, height and weight increase, and height in particular shows a higher correlation. Among physical activity data, endurance, agility, balance, and agility, excluding flexibility, show some correlation with physical information data, such as height and weight. Based on this correlation, as will be described later, it can be understood that the final obesity prediction value is determined by determining the weight of each of the obesity prediction value predicted with body information data and the obesity prediction value predicted with physical activity data.

In this embodiment, the normalized value of each data is loaded into an AI model that classifies it into 5 levels (very good, excellent, average, caution, needs improvement), and after executing the classification procedure (S30), the average result for each age group is, Average grades and individual results can be provided in a visualized form. Additionally, appropriate physical activity or diet may be provided as recommended information according to the above classification.

The method of analyzing the physical information data and the physical activity data can be performed using an optimal categorization method and statistical analysis method using AI.

Calculate the level of physical development compared to the age-specific average using statistical methods, and convert the sum of the five normalized data on physical activity ability into a score of 0 to 100 to determine the quintile (divided into five sections of 20% each) ) is analyzed.

Analysis of AI data involves classification into categories according to physical activity ability scores and activity abilities such as strength, flexibility, balance, agility, and agility. In addition, because the existing data segmentation technique uses only evaluation information, it has limitations related to additional user problems and new clinical information problems. In order to effectively reflect additional document information and reference information, the existing PMF (Probabilistic matrix factorization) method is used. Document information analysis can be done efficiently by incorporating a CNN (Convolutional neural network) model into the model. In order to more effectively understand the meaning of each information (word) in the data, a CNN is constructed using Distributed Representation, which has recently been proposed as an effective method, rather than using 1-of-V Representation of words. The Softmax function refers to an exponential function normalized to a value between 0 and 1 used for multinomial classification using probability distribution, and the DNN method increases accuracy by adding multiple hidden layers. The SGAN method improves accuracy by using a small amount of initial labeling data and fake data.

Learning performance is improved by preprocessing through normalization for each item of the collected data, and the cost function to find the optimal solution for classification is the cross entropy of the actual value L and the predicted value H, Use to find the weight that has the minimum entropy and classify it into several groups. Multinomial classification uses the softmax function, The index with the maximum value between 0 and 1 is selected and classified as multiple. SGAN is trained so that the generation model G is minimized and the discriminator model D is maximized using the cost function V(D,G) as shown in the formula below, where refers to the probability value of sampling real data. means the probability value for data sampled from random noise using Gaussian distribution, D learns to distinguish between real and fake, and G trains fake so close to real that D cannot distinguish between real and fake. By learning to generate them, the performance of D and G is increased, ultimately improving the grouping accuracy of the data.

Next, the normalized data is used to predict obesity at a certain point in time through a growth prediction AI model based on infant growth curves (S40).

The obesity prediction method will be described in more detail with reference to FIG. 12, which is a diagram illustrating the obesity prediction process.

First obesity is predicted at a certain point in time (e.g., after n months) using physical information data such as height and weight, and second obesity is predicted at a certain point in time using physical activity data such as endurance, agility, balance, and agility. predict. The final obesity level is predicted by assigning weights to the first obesity level and the second obesity level, respectively. The weights in the first measurement can each be assigned a default value of 0.5. A second measurement is performed at a predetermined point in time, for example, n months after the first measurement. Physical information data and physical activity data are also obtained during the second measurement, and the first and second adiposity degrees after, for example, m months are respectively predicted from the physical information data and physical activity data. The weight in the second measurement is determined by comparing the predicted adiposity value in the first measurement with the actual adiposity value derived from the height and weight actually measured in the second measurement. This weight reset can be understood as training a growth prediction AI model for more accurate predictions. After this, measurements and predictions can be repeated in the same way. In this embodiment, it is described as predicting the obesity level, but it is also possible to predict the height and weight and derive the obesity level from the height and weight.

Each time the measurement and prediction are repeated, the rate of increase in obesity and the risk of obesity are analyzed (S50) and provided to the user (S60).

The information provided to the user includes not only the rate of increase in obesity and the risk of obesity, but also each measured/predicted value of the infant's height and weight, and recommended information such as physical activity and diet. FIG. 13 is a diagram illustrating an example of providing analysis data on information about the height of an infant or young child to a user. In this embodiment, measured values are shown up to the third time, and predicted values are shown at the fourth time.

In another embodiment, an image model of an infant or child may be created using the predicted body information data of the infant, and the image model may be visually provided to the user.

The system according to the above embodiment is connected to a terminal that provides body information data and physical activity data including the user's gender, height, and weight, and the terminal through a network, and the body information data and body provided from the terminal. It may be configured to include an obesity prediction server that analyzes activity data to predict obesity at a certain point in time and provides the obesity information to the terminal. The obesity prediction server includes a storage unit that converts and stores data provided from the terminal into a database, an obesity prediction calculation unit that learns data from the storage unit using artificial intelligence (AI) to predict the user's obesity level at a predetermined point in time, and predicts the obesity level. It may be configured to include an obesity information providing unit that provides obesity information predicted by the calculation unit to the terminal. In addition, the obesity prediction calculation unit includes a first prediction unit for predicting body information at a predetermined point in time based on the body information data from the terminal, and a first prediction unit for predicting body information at a predetermined point in time based on the physical activity data from the terminal. 2 a prediction unit, and a weight determination unit that determines a weight to be assigned to each of the predicted values of the first prediction unit and the second prediction unit, and the obesity level at the predetermined time is the predicted value from the first prediction unit and the second prediction unit. It can be predicted based on the weight from the weight determination unit.

The method for predicting the obesity level of infants and young children according to another embodiment of the present invention may classify previously collected physical data of infants and young children into each group and build an obesity model for each group. After measuring the infant's physical data, it is determined which of the above groups the infant belongs to, and the obesity degree of the child can be predicted by using the obesity degree model for that group. In one embodiment, this obesity model can be created in advance, stored on a server, and downloaded from the server upon a client's request.

In an embodiment of the present invention, each component may be implemented by various means, for example, hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, an embodiment of the present invention includes one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), and FPGAs ( Field Programmable Gate Arrays), processor, controller, microcontroller, microprocessor, etc.

In addition, in the case of implementation by firmware or software, an embodiment of the present invention is implemented in the form of a module, procedure, function, etc. that performs the functions or operations described above, and is stored on a recording medium readable through various computer means. can be recorded Here, the recording medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the recording medium may be those specifically designed and constructed for the present invention, or may be known and available to those skilled in the art of computer software. For example, recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROM (Compact Disk Read Only Memory) and DVD (Digital Video Disk), and floptical media. It includes magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program instructions such as ROM, RAM, flash memory, etc. Examples of program instructions may include machine language code such as that created by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. Such hardware devices may be configured to operate as one or more software to perform the operations of the present invention, and vice versa.

In the above, specific preferred embodiments of the present invention have been described, but the present invention is not limited to the specific embodiments described above, and the technical field to which the present invention pertains without departing from the gist of the present invention as claimed in the patent claims. Anyone skilled in the art can make various modifications, and such modifications fall within the scope of the claims.

The present invention as described above can be applied as an AI screening model for large-scale infant health care projects such as infant physical fitness tests and health checkups, and can be used as an obesity risk clinical decision-making system and patient management system for hospitals such as children's hospitals and growth and development clinics. there is.

Claims (14)

A method of predicting obesity based on infant growth and development data using a growth prediction AI model, the method
Obtaining first data, which is physical information data including gender, height, and weight for infants and toddlers, and storing it in a database;
Obtaining second data, which is physical activity data for infants and toddlers, and storing it in the database;
normalizing and categorizing the first and second data;
Predicting obesity after n months through the infant growth curve of the growth prediction AI model based on the normalized first and second data;
reacquiring the first and second data after the n months and storing them in a database;
Learning a growth prediction AI model based on the predicted obesity level and the re-acquired first data;
Predicting obesity after m months using the learned growth prediction AI model; and
Analyzing the risk of obesity based on the predicted obesity level;
A method of predicting obesity based on infant growth and development data using a growth prediction AI model comprising: According to claim 1,
The predicted value of obesity is calculated by calculating a predicted value based on the first data and a predicted value based on the second data, and summing the respective predicted values according to their respective weights,
The step of training the growth prediction AI model is a growth prediction comprising comparing the predicted obesity level with the re-acquired first data, and adjusting the weight used to predict the obesity level according to the comparison result. A method of predicting obesity based on infant growth and development data using an AI model. According to claim 2,
The weight is a ratio of the predicted value based on the first data and the predicted value based on the second data for the predicted value of obesity. A method of predicting obesity based on infant growth and development data using a growth prediction AI model. The method according to any one of claims 1 to 3,
The second data is a method of predicting obesity based on infant growth and development data using a growth prediction AI model, wherein the second data includes data on endurance, agility, balance, and agility. The method according to any one of claims 1 to 3,
Predicting obesity based on infant growth and development data using a growth prediction AI model, further comprising providing the classified category information and/or analyzed obesity risk information to the user in a visualized form on a display. method. According to claim 5,
A method of predicting obesity based on infant growth and development data using a growth prediction AI model, further comprising recommending a diet for infants and toddlers based on the classified category information or analyzed obesity risk information. According to claim 5,
A method of predicting obesity based on infant growth and development data using a growth prediction AI model, further comprising recommending physical activity for infants and toddlers based on the classified category information or analyzed obesity risk information. The method according to any one of claims 1 to 3,
The first and second data are acquired periodically, and the cycle includes annually, quarterly, monthly, weekly, etc. A method of predicting obesity based on infant growth and development data using a growth prediction AI model. A system that predicts obesity based on infant growth and development data using an artificial intelligence algorithm,
A terminal that provides physical information data and physical activity data including the user's gender, height, and weight; and
An obesity prediction server that is connected to the terminal through a network, analyzes the body information data and physical activity data provided from the terminal, predicts obesity at a certain point in time, and provides the obesity information to the terminal.
The obesity prediction server,
a storage unit that converts and stores data provided from the terminal into a database;
An obesity prediction calculation unit that learns the data from the storage unit using artificial intelligence (AI) to predict the user's obesity level at a certain point in time, and
An obesity information providing unit that provides obesity information predicted by the obesity prediction calculation unit to the terminal;
A system for predicting obesity based on infant growth and development data using an artificial intelligence algorithm comprising: According to clause 9,
The obesity prediction calculation unit,
a first prediction unit that predicts body information at a certain point in time based on the body information data from the terminal;
a second prediction unit that predicts physical information at a certain point in time based on the physical activity data from the terminal;
A weight determination unit that determines the weight assigned to each predicted value of the first prediction unit and the second prediction unit.
Including,
The obesity level at the predetermined point in time is predicted based on the predicted values from the first prediction unit and the second prediction unit and the weight from the weight determination unit. Obesity level is calculated based on infant growth and development data using an artificial intelligence algorithm. A predictive system. According to claim 10,
The weight determination unit predicts obesity based on infant growth and development data using an artificial intelligence algorithm, wherein the weight determination unit adjusts the weight by comparing the predicted obesity at a given point in time with the obesity degree from body information data obtained at the given point in time. A system that does. The method according to any one of claims 9 to 11,
A system for predicting obesity based on infant growth and development data using an artificial intelligence algorithm, wherein the obesity information providing unit further provides recommended information about diet and physical activity based on the predicted obesity information. A computer-readable recording medium having a built-in program that operates to perform the method according to any one of claims 1 to 3. A program stored on a computer-readable recording medium operative to perform the method according to any one of claims 1 to 3.